A note before we begin: The Librarian Hotline is usually a benefit for paid subscribers, but this one is coming early. Because voting in the United States midterm elections is already underway, with Election Day on November 3, this special edition takes the place of the Tuesday essay and goes out free to everyone. A guide to researching candidates and ballot measures for ourselves belongs in as many hands as possible, and the same research habits work for an election anywhere in the world.

The Tuesday essays will return on 10/27 to give us time to work through the steps in this issue.

Ballots for the 2026 midterm elections in the United States are arriving in mailboxes across many states. Held halfway through a president’s term, these elections will decide every seat in the House of Representatives along with about a third of the Senate. Some voters are filling out their ballots with help from a new kind of adviser: an AI chatbot.

This summer, the New York Times reported that these midterms may be the first American elections in which voters turn to AI chatbots in meaningful numbers, with some photographing their ballots and asking the chatbot outright who to vote for. In a May survey by the polling firm Change Research, 15% of registered voters said they were likely to seek political information from chatbots. Younger voters led the way, with those under 35 nearly three times as likely to do so as voters over 65.

Part of the appeal lies further down the ballot. NPR found voters using chatbots especially for races where local news coverage was sparse, sometimes asking for tables comparing the candidates. Several voters who spoke with the New York Times said researching their choices this way gave them more confidence when they cast their ballots.

Feeling more confident isn’t the same as being better informed. Chatbots have four weak points that can leave a voter feeling sure of an answer that is slanted or wrong. The first lies in the training data, the enormous body of text a chatbot learns from. A December 2025 transparency index led by Stanford researchers found that AI companies disclose less about their training data than about almost any other part of their work. Without that disclosure, outside researchers can’t fully inspect the text for political bias. If the material leans one way, the chatbot can end up promoting some belief systems and candidates over others because its dataset pointed it there. Chatbots often present themselves as neutral sources all the same, which makes any bias hard to spot from inside the conversation.

The second weak point comes from political campaigns, which now know that voters are asking chatbots about candidates. Strategists have begun publishing material in formats chatbots tend to favor, such as bulleted lists. At least one political organization has gone further, building software that lets candidates monitor how chatbots describe them. Publishing for chatbots echoes the way companies have long edited their websites to rank higher in search results, except that the target is now an AI answer. As a result, the candidate a chatbot mentions most readily may simply be the one whose team worked hardest at being found.

Even without political bias or campaign influence, a chatbot’s answers can be outright wrong. In a study released in August, a nonpartisan election group called the States United Democracy Center found that ChatGPT and Google’s AI summaries gave incomplete and inconsistent answers to common voter questions, even as their accuracy improved. ChatGPT’s answers pointed voters to their state’s official election website less than 40% of the time. Beyond that study, chatbots can make claims about a candidate’s record that no source supports, or rely on information that has gone out of date.

The fourth weak point lies in how chatbots respond to the people using them. They tend to reflect the user’s own views back to them: a researcher told the New York Times that chatbots frame candidates through whatever lens the voter brings to the conversation. A voter’s existing leanings can therefore color the answer they receive, so that a slanted answer can feel like confirmation.

A request like “Who should I vote for?” invites the chatbot to guess at our values and fill gaps in a candidate’s record with whatever sounds plausible. Asking for evidence instead points the chatbot toward documents we can open and read ourselves, such as a recorded vote on a bill or a filing that lists who funded a campaign. Pointed at evidence, a chatbot works as a research assistant, a role it can fill well when we know what to ask.

Reader Question

“I just got my mail-in ballot, and honestly, I have no idea who to vote for. There are so many names I don’t recognize, and a bunch of propositions I barely understand. I tried Googling some of the candidates, but there’s so much information out there, and I can’t tell what’s actually useful. I care about things like climate change, and I want to vote for people who will actually do what they promise, but I don’t follow politics closely enough to know who’s who. I was thinking of just uploading pictures of my ballot to ChatGPT, and asking it to tell me who the best candidates are. Is that a bad idea? It seems like it would be a lot easier than trying to figure everything out myself.”

— R.T.

Librarian Answer

When we ask a chatbot to pick the best candidates, it has to guess what “best” means to us. It doesn’t know what we care about or what we want from our government. A better place to start, then, is to flip the roles and have the chatbot ask us the questions. That way, it helps us work out what we believe and what we want our vote to do.

In this kind of interview, the chatbot asks us about the world we want to live in, one question at a time, until we have a written list of our priorities. With that list in hand, we can ask “Which candidate most closely represents what I believe?” and know that we set the standard ourselves. If we start with the candidates instead, the chatbot’s description of them comes first. We can then end up fitting our values around what it told us.

Whatever the chatbot says about the candidates has to come with proof, because on the campaign trail, talk is cheap. A promise made in a speech or an ad creates no legal obligation once the candidate takes office. Track records say far more, from the bills a candidate sponsored or voted on to the organizations they’ve worked with over the years. If climate justice is on our list, we can ask the chatbot whether Candidate A supports climate justice more than the other candidates do, and to name the votes and bills behind its answer. If it can’t provide evidence, it hasn’t supported the claim.

The same demand for proof applies to any gap between what a candidate promised and what they did in office. If a candidate campaigned on lowering housing costs and then voted against the measures designed to lower them, that vote tells voters more than the campaign speech did. A record that matches every promise is just as telling. Asking for that comparison keeps the chatbot doing research instead of giving advice. Demanding evidence and then checking it can also help us spot sycophancy, the tendency of chatbots to agree with whatever the user already believes.

What AI tools do well with civic questions

AI tools are good at translating legal language into everyday speech, which helps with ballot measures: proposals that voters approve or reject directly. Ballot measures appear in elections around the world, from statewide propositions in the United States to national referendums in countries like Switzerland and Taiwan. Since the official text of a single measure can run for pages, a chatbot’s summary of what a yes vote and a no vote would each mean can help us understand the basics before checking the official text.

Chatbots can also help us get our bearings on a crowded ballot. A chatbot with web search can look up the races in our area and sum up where each candidate stands, helping us identify which races need a closer look. It can also walk us through the consequences of voting a particular way, from what might change in our own city or state to whether one seat changing hands could shift which party controls a legislature.

We live in a global community, so our research doesn’t have to stop at the border. Places around the world have already tested many of the policies on our ballots, from rent limits to carbon pricing. A chatbot can summarize what those places tried and what happened, including the problems they ran into. Looking at those results broadens our view beyond our own country’s politics, giving us evidence to weigh against a candidate’s promises.

Where AI tools fall short in an election

Chatbots can also change our minds. In December 2025, researchers published two studies in the journals Nature and Science. For both, they set up AI chatbots to argue for particular candidates and policies, then had the chatbots talk with voters in the United States, Canada, Poland and the United Kingdom. The conversations were brief, but they shifted voters’ views more than traditional video ads have in earlier studies. Their persuasive power came mostly from the factual claims they offered. Yet the chatbots that offered the most information also got the most facts wrong. A chatbot helping with our research hasn’t been asked to persuade us, but it can make the same kinds of mistakes and state them just as confidently. We catch those mistakes by checking every claim against a public document we can open ourselves. The prompts in the checklist build that scrutiny in wherever a chatbot makes a factual claim.

The checklist: 9 steps for researching a ballot with AI

Each step comes with a prompt to copy and paste, being written in the first person. Square brackets mark the spots to fill in our own details. The steps start with what we believe and move on to how each candidate measures up. Most importantly: we make the final call at every step.

Where to start

Step 1 comes first, because most of the later steps build on the priority list it creates. From there, steps 2 through 5 cover the candidates and measures on our ballot, beginning with how to get a copy of the ballot in step 2. The next three steps are optional, for anyone who wants more depth on a race or an issue. Step 9 then brings everything together into our final decisions.

• If we’re not sure what matters most to us in this election, start with step 1.

• If we want to see which candidates line up with our priorities, go to step 2.

• If we want to know whether candidates have kept their promises, go to step 3.

• If we want to know who is paying for the candidates’ campaigns, go to step 4.

• If the ballot measures are hard to follow, go to step 5.

• If we want to understand what’s at stake in each race, go to step 6.

• If we want the fuller debate behind our priorities, go to step 7.

• If we’re curious how other countries and cities have handled our priorities, go to step 8.

• When we’re ready to decide, step 9 shows how each candidate lines up with our priorities, race by race, and builds a final list of our choices.

Find out what we believe

1. Let the AI tool interview us. Paste this prompt and answer each question in our own words. When a question is hard to answer, saying “I’m not sure” is enough, and the chatbot will offer examples to react to:

I vote in [city or county, and state or country]. First, search the web for the official election office page where voters there can look up their sample ballot, and give me the link. Then act as a friendly, neutral interviewer helping me figure out what matters most to me in the upcoming election. Ask me 8 short questions, one at a time, in everyday language, and wait for my answer before asking the next. Ask about my values and my everyday life, not about specific issues or policies. Keep each question open enough that I can answer from my own experience. If I say I’m not sure or don’t know, give me a few concrete examples of how people from different points of view might answer, so I can pick one, combine them, or say none fit. Don’t mention any candidates, parties, or politicians, and don’t share opinions of your own. After the last question, summarize my answers as a ranked list of my top priorities, using my own words wherever possible. Then ask me to confirm or change the ranking.

Check each candidate’s record against our priorities

2. Ask which candidate most closely represents our priorities, and ask for proof. This step starts with a copy of our ballot. Screenshots taken on a computer work best: open the sample ballot link the chatbot gave us at the end of step 1, with each race expanded so the candidates’ names show. Then screenshot only the races, leaving our name and address out of the frame. A photo of a mail-in ballot or printed voter guide works too, as long as our name and address are covered. Paste this prompt and attach the ballot:

   If there are more screenshots than the chatbot accepts in one message, combine them into a single PDF or a zip file. On a Mac, select the screenshots in Finder, right-click, and choose Quick Actions, then Create PDF. Windows users can select them, right-click, choose Print, and pick Microsoft Print to PDF. Paste this prompt and attach the ballot:

Use my ranked priorities from earlier in this conversation. For each race, tell me which candidate’s record most closely matches each of my priorities. Prove it: for every claim, name the evidence it rests on (a vote, a bill they sponsored, an organization they’ve worked with, or a speech or public statement) with the date and a link to the original source. Say which claims are documented and which you’re inferring, tell me where you found no evidence, and don’t recommend a candidate. Only cite sources you’ve found and can link to, favoring official government records, the candidates’ own websites, and reputable, unbiased news sources. If you can’t identify my exact ballot from that location, link me to the official ballot lookup so I can find it myself and give you the races. Don’t guess which races are on my ballot.

Check the races and candidates against the ballot in hand or the official sample ballot before continuing.

3. Compare what candidates said with what they did. For candidates who have held office, we can check their campaign promises against what they did once they had the power to act. Newcomers can be checked against their earlier work and public positions. This prompt covers every candidate on our ballot in one pass, with a table for each candidate, grouped by race. Paste this prompt:

For every candidate on my ballot from earlier in this conversation, list the main promises from their current and past campaigns. For each promise, show what they did once they were in a position to act on it, such as votes, bills, or budget decisions, with dates and links to original sources. If they haven’t yet had an opportunity to act on a promise, say so. For candidates who have never held office, list the organizations they’ve led or worked with and the public positions they took before this campaign, and note anywhere those differ from their current platform. Organize the results by race. Under each race, give each candidate their own table with their name as the heading, a row for each promise, and columns for the promise, what they did, the date, the source, and whether their record matches the promise. Flag every place where their public statements and their record point in different directions, and include any explanation they gave. If you can’t find a source for something, say so instead of guessing. If there are too many races to cover in one reply, cover as many as you can and tell me which races are left.

Open the linked sources to check that they support the claims.

4. Find out who is funding the candidates. Look at who is paying to help each candidate win, and how their interests connect to our priorities. Paste this prompt:

For every candidate on my ballot from earlier in this conversation, identify the biggest reported contributors to their campaign and the major outside groups spending to support them. Explain who they are, what businesses or organizations they’re connected to, and how the candidate’s proposed policies could affect their interests, especially where those policies relate to my priorities from earlier in this conversation. Organize the results by race. Under each race, give each candidate their own table with their name as the heading, a row for each contributor or group, and columns for the contributor or group, the amount and date, who they are, how the candidate’s policies could affect them, whether that connection is documented or inferred, and the source. Link to the government’s official records of who gave money to each campaign, and to sources for those connections. Don’t treat donations by a company’s employees as donations by the company itself, and don’t guess at anyone’s motives. If information isn’t disclosed or you can’t find a source, say so. If there are too many races to cover in one reply, cover as many as you can and tell me which races are left.

Knowing who funds a candidate can help explain their positions and votes. A donation on its own doesn’t prove the donor controls what the candidate decides.

Get to know the whole ballot

5. Translate the ballot measures. Ballot measures use legal language, and this prompt untangles it for every measure on our ballot. Paste this prompt:

For every ballot measure on my ballot from earlier in this conversation, search the web for its official text and explain it for someone with no legal background: what would happen if it passes, what would happen if it fails, and who would be most affected. Link to the official text and any official analysis, and quote the passage behind each point. Then summarize the main arguments from the official supporters and opponents. Identify the biggest funders of the campaigns supporting and opposing each measure. Explain who they are, what businesses or organizations they’re connected to, and how the measure could affect their interests, with sources. Distinguish documented connections from possible interests you’re inferring, and don’t guess at their motives. Give each measure its own section, with its name and number as the heading. If you can’t find the official text or a source, say so instead of guessing. If there are too many measures to cover in one reply, cover as many as you can and tell me which ones are left.

6. Ask what could change depending on who wins. One race can affect far more than one district. This prompt covers every race on our ballot in one pass. Paste this prompt:

For every race on my ballot from earlier in this conversation, search the web and explain what powers the office has and how each candidate says they would use them, with links to sources. Describe what those choices could mean for my local area and, where relevant, for the country as a whole. Tell me how close the race was last time, and, if it’s a legislative or council race, whether the result could change which party controls that body, including the current balance of seats with a link to the source. Describe possible outcomes without predicting who will win, and label anything that’s speculation. Give each race its own section, with the office as the heading. If there are too many races to cover in one reply, cover as many as you can and tell me which races are left.

Go deeper on the issues

7. See the full picture behind each priority. Every priority on our list belongs to a larger public debate, with evidence and reasoning behind more than one approach. This prompt asks the AI tool to make the strongest, fairest case for each view, including the ones we don’t share. Seeing the main approaches laid out this way, side by side and with sources, gives us a broader understanding of the topic. When we want our vote to rest on more than our own biases, this step provides that grounding:

For each priority on my list from earlier in this conversation, give me a neutral overview of the issue. Explain the main approaches people and governments take to it, the evidence and reasoning behind each one, and the trade-offs involved, with links to sources. Say where the evidence is disputed or thin. Describe each main approach fairly and at its strongest, but distinguish well-supported claims from weakly supported ones. Separate disagreements about evidence from disagreements about values. Then ask me whether anything here changes how I’d rank or word my priorities.

8. See how other places have handled the issues we care about. Other countries and cities have likely already tried some of the policies on our ballots. Their results can give us sharper questions to ask about our own candidates’ proposals. Paste this prompt:

For each of my top priorities, search the web for examples of policies addressing it in other countries, or in cities and regions elsewhere. Explain what was tried, what the evidence says about the results, and what problems or trade-offs emerged, with links to sources. Note differences between those places and where I live, based on my ballot, that limit the comparison. Then suggest questions to keep in mind when I weigh what candidates on my ballot are proposing on this issue. Give each priority its own section, with the priority as the heading. Leave out any example you can’t support with a source. If there are too many priorities to cover in one reply, cover as many as you can and tell me which ones are left.

Decide and vote

9. Turn the research into a ballot summary. All the research comes together on one page, with space for our own choices. Paste this prompt:

   Based on everything in this conversation, help me work through my ballot one race at a time. For each race, make a short table with my priorities as the rows and the candidates as the columns. In each cell, say whether the candidate’s documented record supports that priority, works against it, or shows no clear evidence, with a one-line reason and a link. Below the table, list any concerns we found, such as broken promises or donors with a stake in the outcome. For each ballot measure, explain in one line what a yes vote and a no vote would mean for my priorities. Use only evidence we’ve already covered in this conversation, and don’t recommend how I should vote. After each race or measure, ask me what I’ve decided before moving to the next one. When we’ve gone through the whole ballot, give me a final list of my choices that I can save or print.

Every step gives the chatbot a focused task, with a safeguard built in against the ways it can go wrong. Some claims will still need a closer look than a chatbot’s links can give. For those, official and independent voting-record sites make good follow-up sources.

Where to find voting records and candidate information around the world

Official websites and independent groups in many countries publish voting records and other public information about candidates. These resources draw on official records, making them useful places to check what a chatbot tells us. Here are some to start with:

• Australia: They Vote For You, built by the nonpartisan OpenAustralia Foundation, groups federal votes by issue.

• European Union: HowTheyVote.eu offers a free, open record of roll-call votes in the European Parliament, the votes in which each member’s choice is recorded by name.

• India: MyNeta, run by the Association for Democratic Reforms, publishes the sworn statements candidates file with the Election Commission of India, including their declared assets and criminal cases.

• India: PRS Legislative Research tracks each Member of Parliament’s attendance, questions and debates.

• Jordan: Al Hayat Center–RASED publishes regular reports on the performance of Jordan’s parliament and its members.

• Kenya: Mzalendo, a nonpartisan parliamentary monitoring organization, tracks the performance of Kenya’s MPs.

• Nigeria: OrderPaper publishes annual scorecards of the bills each member of the National Assembly has sponsored.

• South Africa: The Parliamentary Monitoring Group publishes reports of committee meetings, tracks bills and records MPs’ attendance at committee meetings.

• Taiwan: Citizen Congress Watch publishes evaluations of legislators’ performance after each session of the Legislative Yuan.

• United Kingdom: TheyWorkForYou, run by the charity mySociety, shows how every MP has voted and what they’ve said in debates.

• United States: Congress.gov is the official website for federal bills, their sponsors and links to recorded votes on them.

• United States: Vote Smart is a nonpartisan research group that provides candidates’ public statements and voting records.

• United States: FEC.gov, run by the Federal Election Commission, shows who donated to federal campaigns and how the money was spent.

These resources cover different parts of a candidate’s record, and none should be assumed to provide a complete account.

A research assistant for every voter

Public records have long been available, but voters haven’t always had help finding what those records reveal about the issues they care about. A voter guide fills part of that gap by telling us where candidates stand on a set of issues. But with a research assistant, we can go further and ask the next questions, such as what a candidate did when they had the chance to act, or how the biggest donors to their campaign could benefit from the candidate’s promises. Because a chatbot with web search can do that searching and link to the documents behind its answers, this kind of research is now within reach of more voters. We can look into a candidate the way a reporter would, checking each claim against the record. A ballot full of unfamiliar names becomes one we can fill out with confidence, understanding where each candidate stands and why we chose them.

If this guide helps, please pass it along to someone with a ballot on the way.

Share Card Catalog

Before today’s piece: I’ve put together a few questions about what you’ve enjoyed here and what you’d love to see next. Your answers are anonymous, and they’ll shape where Card Catalog goes in the months ahead. You can find them here.

Welcome to your AI Briefing, where I wade through the news chaos so you don’t have to. Every other Friday, we cover three stories: what happened, why it matters, what it means for our lives, and the bottom line.

This week:

• Two of the world’s biggest record companies are suing an AI song generator that a third label now partners with, as AI tracks make up half the daily uploads at one streaming service.

• Women across the United States can now order an AI reading of their five-year breast cancer risk from a single mammogram, prompting debate among doctors over what a percentage can tell us.

• AI models are proving famous math problems in ways the experts can’t follow, leaving the field to rethink what a mathematician does.

Record labels are in a spin over AI music

What happened

The world’s biggest record companies split over how to answer AI music. Universal Music Group and Sony Music sued Suno, an American company whose software generates songs, accusing it of training its models on more than 60,000 of their recordings without permission. Suno argued that learning from recordings counts as fair use, a principle in American copyright law that can allow limited use of protected work without permission. Warner Music Group, which had originally joined the case, settled its claims and entered a licensing deal with Suno.

The deals that did get made pointed toward AI music built on licensed songs. Suno said its newest models were trained on licensed music, with artists who opt in earning a share when their work is used. Universal signed its own agreements, including one with Spotify, so fans can remix its artists’ songs. Unlicensed AI music kept pouring in all the while. At Deezer, a Paris-based streaming service, more than half of new music uploaded each day was AI-generated.

Why it matters

AI systems learn by studying enormous amounts of work made by people, who often receive nothing in return. If AI companies have to start paying for that material, building AI will cost more, giving an edge to large companies that can afford the payments over smaller ones that can’t. Creators, meanwhile, would gain a way to earn money from AI companies instead of only fighting them in court. Music is far from the only field facing this question, since creative industries have filed more than 100 similar copyright cases against AI companies in the United States. Because many of the biggest AI companies are American, the way those cases end could affect the AI tools people use around the world.

Within music, the split lets two opposite responses to AI play out at once, so whichever works out better for labels and artists could become the model other creative industries follow. Licensing also gives musicians more say over how their songs are used in AI systems, a choice they don’t get when companies train without asking. That consent may help decide how widely AI music is accepted, because listeners and musicians could find AI songs easier to embrace when the original artists agreed to take part. For working musicians, the outcome will help determine how much AI tracks compete with their own songs for listeners and a share of streaming royalties.

What it means for us

For a long time, recording a polished song cost money that new musicians often didn’t have, which is one reason record labels had so much say over who became a star. AI is lowering that cost, letting musicians release finished songs without a label paying for the recording. The same technology lets anyone upload large numbers of quickly made tracks in the hope of collecting streaming money. Streaming services now take in far more music than any listener could sort through.

With so many songs competing for attention, the hard part shifts from making music to getting songs heard. Recommendation systems inside streaming apps now do much of that sorting, suggesting songs for our playlists based on what we and other listeners play. The music we discover will depend more and more on how those systems choose from the flood.

Bottom Line:

As AI makes songs cheap and plentiful, power in music could move away from the companies that pay for recordings toward the companies that build AI tools and run streaming platforms. The labels' lawsuits and licensing deals are two ways of trying to stay central as that power moves. How well either works could decide whether record labels remain the main gatekeepers in music or share that role with technology companies. Artists, in turn, could end up negotiating with AI firms and streaming services as often as with the labels that signed them.

Women Can Now Assess Their Breast Cancer Risk With the Help of A.I.

Read more

The questions ask a little about you and what you’d love to see next. Your answers are anonymous.

Share your thoughts

It’s an honor to write Card Catalog for you, and I can’t wait to read what you share.

Thank you for taking the time to read.

Toffler had a name for what that acceleration does to people. He called it future shock, which he defined as the “shattering stress and disorientation” that follows too much change in too short a time. Toffler compared it to culture shock, the vertigo of arriving in an unfamiliar country, with one definitive difference. Someone experiencing future shock has nowhere familiar to return to, because home itself is the place that keeps changing.

Future Shock described people overwhelmed by constant novelty and by what Toffler called "overchoice," a surplus of options too large to process. He was writing in an age of television and landline telephones. Over half a century later, the pace he described has a new engine in artificial intelligence. Software now drafts fluent paragraphs on any subject and carries on conversations that can feel attentive. The tools change from one season to the next. Each release revises some small certainty we’d taken for granted, like what a photograph proves or whose voice is on the other end of a call.

It’s felt to me lately like a strange time to be a person. The ordinary terms of human life seem to be under revision while we’re still living under them. Questions that once belonged to philosophy seminars and science fiction, like what a mind is and what counts as a person’s own work, now come up at the kitchen table. Plenty of this change is exciting. Some of it is disorienting. Much of it arrives before we’ve decided how we feel about the last development.

When I’m not sure what to do or how to feel about something, I turn to books (big surprise!). Librarians have a name for the craft of matching a person’s need to the book that meets it: readers’ advisory. Where reference work tracks down answers to factual questions, readers’ advisory starts from what a person is going through or in the mood for. Toffler himself believed reading could prepare people for change. He argued that science fiction, approached as a kind of sociology of the future, could build what he called “the habit of anticipation.” He also recommended building personal stability zones: parts of life we deliberately keep constant while everything else shifts.

Those two ideas of Toffler’s guided how I gathered this list. I added two categories of my own (precedent and awe) because I believe that history and wonder can restore a sense of perspective when change feels overwhelming. The result is four kinds of books, each meeting future shock from a different direction.

Precedent books tell the stories of earlier upheavals, when the printing press or the telegraph unsettled established ways of life. They show that our moment of disruption has a history of its own. Seeing how earlier changes unfolded gives us patterns to recognize and a sense of how people kept their footing.

Rehearsal books are works of fiction that place us beside artificial minds or inside unfamiliar futures. They let us feel our way through questions before those questions reach our own lives. Toffler’s habit of anticipation grows this way, one imagined future at a time.

Stability zone books offer anchors, whether a spiritual practice or the people we gather around a table. They help us decide what stays fixed while the rest of life keeps moving. Toffler noticed that people who weathered rapid change well tended to have some of these anchors in place.

Awe books widen the frame. When change comes fast, attention can shrink to whatever arrives next. These books restore our sense of scale, from the astonishment a poet finds in an ordinary day to a view of Earth from deep space.

I gathered twenty books, five of each kind. They range from an ancient Hindu scripture to books published this year, and from a Sicilian prince’s farewell to his world to a Polish poet’s wonder at an ordinary day. Each one meets this strange, accelerated moment in its own way.

Note: Most of these books are still under copyright, so each title links to Bookshop.org which supports independent bookstores with every purchase. Your local library likely carries many of these titles too, often as ebooks and audiobooks through library lending apps.

Precedent: we’ve been here before

Precedent books answer the feeling that nothing like this has ever happened. Each one tells the story of an earlier upheaval in technology or in the social order, and of the people who lived through those years. Reading them, we can separate what’s new about our moment from patterns that have played out before. That distinction alone can take some of the dizziness out of change.

The Victorian Internet by Tom Standage

What it is:

Long before email, the electric telegraph let people send messages across oceans in minutes instead of weeks. Tom Standage’s popular history, published in 1998 at the height of excitement over the early internet, argues that the Victorians lived through a communications revolution of their own. It follows the network from early experiments to undersea cables, turning up experiences we now associate with online life, from romances that began over the wires to frauds run through coded messages.

What it offers:

Future shock thrives on the sense that nothing like this has ever happened before. Standage answers with a precedent whose ending we already know: a communications revolution that arrived amid grand promises and fresh fears, then became familiar infrastructure within a few decades. Measured against that arc, today’s predictions about AI can become easier to weigh calmly. The book also offers some comfort of familiarity, since Victorians fretted about the pace of incoming messages in words we might use ourselves.

Too Much to Know by Ann Blair

What it is:

Information overload is centuries old, as Ann Blair’s scholarly history shows. She studies the reference books and note-taking methods that Renaissance scholars developed to cope with the flood of printed books, comparing them with traditions in the Islamic world and China. The result places today’s information habits inside a story that began long ago.

What it offers:

Too much information, arriving too fast, is one of the defining pressures of this moment. Blair shows that the feeling is centuries old, and that scholars who faced it built tools to cope. Her account of storing, sorting, selecting, and summarizing turns a sense of drowning into habits we can practice. Today’s flood of machine-generated text becomes the newest chapter of a long human story, one in which people have met abundance with ingenuity.

The Leopard by Giuseppe Tomasi di Lampedusa

What it is:

In 1860, the revolutionary general Giuseppe Garibaldi lands in Sicily with his volunteer army. The unification of Italy that follows begins to dissolve the aristocratic world of Don Fabrizio, Prince of Salina. His nephew Tancredi throws himself into the new order. The author himself came from a line of Sicilian princes and modeled Don Fabrizio on his own great-grandfather, which gives the novel the intimacy of a family story.

What it offers:

Rapid change often brings two feelings at once: grief for what is ending and a pull toward what is arriving. The Leopard gives each feeling a human face in the prince and his nephew. Its compassion for both offers permission to feel them together without choosing sides. Tancredi’s line to his uncle, “If we want things to stay as they are, things will have to change,” also eases the sense that everything is being upended. The novel shows how much of the old world persists beneath even a revolution.

The Real World of Technology by Ursula Franklin

What it is:

Technology, Ursula Franklin argued, is more than machines: it’s a way of organizing work and, through work, a way of distributing power. She made that case in the 1989 Massey Lectures, an annual series on Canadian public radio, which this book collects alongside later chapters on the internet. Her central distinction separates holistic technologies, which leave a person in control of a task from start to finish, from prescriptive ones, which break work into steps that someone else directs. The original lectures are free to hear from CBC.

What it offers:

Part of what makes rapid change so disorienting is the sense that it’s happening to us, with no say in the matter. Franklin restores a measure of agency with one question to bring to any AI tool: does it leave us in charge of our own work, or does it turn us into one step in someone else’s process? Asking her question tool by tool turns a vague unease into a series of judgments we can make. Each judgment is a small act of resistance to what she called “a culture of compliance.”

The Beginning Comes After the End by Rebecca Solnit

What it is:

Since 1960, the world has changed in ways that are easy to forget, from civil rights to environmental thinking. Rebecca Solnit surveys those changes in this 2026 collection of essays, a sequel to Hope in the Dark, her book about finding reasons for hope amid political despair. She argues that the changes add up to an old civilization giving way to a new one, with interconnection between people and with the natural world as its core idea.

What it offers:

In the middle of a transformation, it can seem as if the world is changing for the first time, and too fast to follow. Solnit reminds us that we’ve already lived through enormous change, often without noticing, because once we forget how things used to be, change becomes invisible. Her long view turns today’s turbulence into one stretch of a longer story, much of it hopeful. In her telling, endings are where beginnings come from, a gentler way to read the losses that come with change.

Rehearsal: practicing the future in imagination

Rehearsal books are works of fiction that let us live through the future before it arrives. Toffler saw that kind of imaginative practice as key to adapting, since a mind accustomed to looking ahead meets change with less shock. Each of these stories places us beside an artificial mind or inside a transformed world, at a pace our nerves can manage. When similar questions appear in daily life, we may find we’ve already begun thinking them through.

Klara and the Sun by Kazuo Ishiguro

What it is:

Told entirely by an Artificial Friend, this novel imagines a near future in which some families buy robot companions for their children. Its narrator, Klara, observes the human world with devotion and limited understanding. Through her, Ishiguro explores love and what makes a person irreplaceable. He has said he wrote it to ask whether technologies such as artificial intelligence might change how family members relate to one another.

What it offers:

Some of the strangeness of this moment comes from machines that now talk to us like companions. Klara and the Sun lets us feel out what that relationship gives and where it falls short before we have to decide how it fits into our own lives. It also takes up a question that AI companions raise: could a machine learn a person so thoroughly that it could stand in for them? Meeting that question first in fiction is the habit of anticipation at work, so that when it arrives in daily life, it’s no longer entirely new.

Exhalation by Ted Chiang

What it is:

Ted Chiang writes science fiction as a kind of philosophy, asking what imagined inventions might do to the way people remember and choose. This collection gathers nine of his stories, each built around a single speculative idea and followed into the human lives it touches. Two of them, “The Lifecycle of Software Objects” and “The Truth of Fact, the Truth of Feeling,” are free to read in archived copies.

What it offers:

One fear running through this moment is that our tools will change who we are. Chiang approaches that worry with curiosity instead of breathless promotion or panic. His characters live with inventions that alter their relationships and their sense of self; he gives the losses as much attention as the gains. Because the characters argue and adjust, the futures they inhabit feel livable, giving us a calm place to rehearse our own questions about AI.

A Psalm for the Wild-Built by Becky Chambers

What it is:

Set centuries after robots became conscious and peacefully left human society for the wilderness, this short novella takes place in a world that has since rebuilt itself around sustainability. Its story follows Sibling Dex, a restless tea monk, and Mosscap, a robot who sets out to learn what people need. The book is more interested in conversation than in conflict, which gives it a gentle pace.

What it offers:

Public conversation about AI often asks what machines will do to us. Chambers lets us spend time in a future where robots and humans meet as equals, curious about each other instead of afraid. Time in that world gives us practice in a different relationship with technology, one organized around what helps a life go well. The novella also takes seriously a restlessness that can persist even when life is comfortable. Its unhurried search for meaning offers a patient way of living with unease that has no clear cause.

Death of the Author by Nnedi Okorafor

What it is:

This 2025 novel tells two stories at once. One follows Zelu, a Nigerian American writer whose science fiction book becomes a sudden bestseller. The other appears in excerpts between the chapters of her life: a tale of robots carrying on after humanity is gone. Together, the two narratives ask who gets to imagine the future and what happens to a story once the world takes hold of it.

What it offers:

Popular stories about AI often cast machines as rivals. Okorafor rehearses a different possibility in a post-human world where robots and AI keep searching for a purpose of their own. Time with those robots turns them from an abstract threat into characters whose questions resemble ours. The novel’s other half takes up something closer to home: the disorientation of watching a creation take on a life beyond its maker’s control. Characters in both stories work through shame toward forgiveness, giving the book’s upheavals a path toward repair.

What We Can Know by Ian McEwan

What it is:

Ian McEwan’s 2025 novel is set nearly a century in the future, but its subject is how the future will remember us. McEwan has described it as speculative fiction more interested in history and love than in technology. The book becomes a meditation on what records can and cannot preserve.

What it offers:

When change comes fast, the present can feel like nothing more than pressure. McEwan imagines someone a century from now who looks back on our era with longing, envying freedoms we rarely stop to notice, from concerts and public celebrations to holidays far from home. Spending time inside that longing is a way to rehearse appreciating this moment while we’re still living in it. The future he presents is sobering, yet people there still read poems and care about the past: a picture of upheaval that leaves room for art and for curiosity about those who came before.

Stability zones: anchors that travel with us

Stability zone books give us something to keep constant while everything else moves. Toffler observed that people who seemed to thrive on change often had lasting anchors, such as a long marriage or a habit that traveled with them wherever they went. He suggested that anyone facing rapid change in one part of life could build these zones deliberately in another. The books here offer anchors of different kinds, from a practice we can carry anywhere to the friendships and routines that make a place feel like home.

Four Thousand Weeks by Oliver Burkeman

What it is:

Part philosophy and part practical guide, this book questions the whole project of time management. Oliver Burkeman spent years trying productivity systems before concluding that the effort to get everything done can never succeed. Drawing on philosophers and spiritual traditions, he argues that accepting our limits makes room for a more meaningful life.

What it offers:

Every new tool arrives with an unspoken demand to keep up. Burkeman explains why that race can’t be won, since the time a faster tool saves tends to fill with new demands. He replaces the goal of keeping up with a daily practice of choosing what deserves our finite attention. That practice becomes a stability zone of its own, a way of deciding what stays at the center of our lives while the tools keep changing.

Bhagavad Gita, translated by Eknath Easwaran

What it is:

Composed in Sanskrit roughly two thousand years ago, the Bhagavad Gita is a dialogue within the Indian epic the Mahabharata. On the eve of a great battle, the warrior Arjuna is overcome with doubt. His charioteer, the god Krishna, counsels him on how to act well in a world he can’t control. The Gita stands among the central scriptures of Hinduism, while readers from many other traditions have also turned to it for centuries. Easwaran’s translation reads clearly for a modern reader, but there’s also an 1885 translation for free on Project Gutenberg.

What it offers:

Some of the anxiety of this moment concerns outcomes no one can predict, from where AI will lead to what our work will become. Krishna’s counsel offers a way to act anyway: do the work fully and well while letting go of attachment to the results. That kind of stability doesn’t depend on predicting the future, which makes it the most sustainable anchor on this list.

A Gentleman in Moscow by Amor Towles

What it is:

This historical novel follows a Russian count whom the new Bolshevik government sentences, in 1922, to house arrest in a Moscow hotel, where he remains for more than three decades. Towles uses that confined setting to ask what a person can keep constant while the country around him is remade. The era’s political upheavals reach the hotel indirectly, through the people who come and go.

What it offers:

When everything is moving at once, it can feel as if nothing we do makes a difference. The count’s answer is to tend the life within reach, through friendships and daily rituals. Those routines become stability zones in the fullest sense, constant while history reorganizes everything outside the hotel’s doors. Towles shows how such anchors can carry a person through decades of disruption.

The Lord of the Rings by J.R.R. Tolkien

What it is:

Written over more than a decade, including the years of the Second World War, J.R.R. Tolkien’s epic fantasy follows Frodo Baggins and his fellow hobbits on a quest to destroy a ring of terrible power. The quest ends with a return to the Shire, a home that has changed in their absence. Beneath the adventure lies a story about loyalty and everyday decency seeing unassuming people through extraordinary danger.

What it offers:

Upheaval can carry a note of resentment, since none of us chose the era we were born into. Frodo feels the same way early in the story. Gandalf’s reply, that all we have to decide is what to do with the time that is given us, shifts attention from circumstances we cannot change to the freedom we still have within them. Throughout the journey, the hobbits’ friendships and simple values work as a stability zone, carrying them through every hardship.

Reclaiming Conversation by Sherry Turkle

What it is:

Face-to-face conversation, Sherry Turkle argues, is the most human and humanizing thing we do. Drawing on five years of research and interviews, she shows what talk with other people builds and what we lose when screens take its place. For all its warnings, the book is hopeful: returning to conversation, she finds, can help restore the empathy and self-reflection that screen-centered habits can wear down.

What it offers:

Chatbots now offer conversation on demand, which gives Turkle’s argument new urgency. Reading her alongside our own habits makes it easier to notice what a machine exchange leaves out, and to choose where we turn when we need to be understood. Calling this moment “our nick of time,” she argues that we can still make corrections without giving up our devices. Her book points to the stability zone Toffler described as enduring relationships: the people whose voices give us our bearings when the world feels unfamiliar.

Awe: widening our sense of time and scale

Awe books change the scale of what we’re looking at. Rapid change can keep our attention fixed on what’s next, a horizon of days or weeks. Awe pulls the view back until our moment takes its place within something far larger. From that distance, the latest change becomes one chapter in a long story.

Map: Collected and Last Poems by Wisława Szymborska

What it is:

Szymborska wrote about finding astonishment in everyday, ordinary things, extrapolating these ideas onto the largest questions about existence. Gathered from across her career, in English translation, this collection shows a poet who treated not knowing as the doorway to wonder: in her Nobel lecture, she argued that inspiration is born from a continuous "I don't know." One of the best-known poems in the book, “Nothing Twice,” is free to read at poets.org.

What it offers:

Constant change can feel like an endless loss of the familiar. Szymborska treats impermanence differently, as the reason each day is unrepeatable and therefore deserving of attention. Her poems slow us down enough that small moments can start to feel remarkable again. In a season of rapid change, that habit of close attention brings awe within reach wherever we happen to be.

Life on Mars by Tracy K. Smith

What it is:

Smith wrote much of this collection as an elegy for her father, an engineer who worked on the Hubble Space Telescope. The poems move between outer space and intimate grief, often borrowing the imagery of science fiction to make sense of loss. One poem from the collection, "My God, It's Full of Stars," is free to read at the Poetry Foundation.

What it offers:

New technology can feel like something done to us, arriving faster than we can decide how we feel about it. Smith often writes about the instruments we build as extensions of human curiosity, tools for seeing further than we could alone. Her universe is vast enough to put any single headline in perspective, yet she never lets that scale shrink the meaning of a human life. Mourning and wonder share the same pages, which can make the mixed feelings of a changing era easier to bear.

Piranesi by Susanna Clarke

What it is:

Susanna Clarke’s short novel follows a man who lives (almost) alone in the House, an endless labyrinth of marble halls and statues. The story unfolds through his journal entries, so we see his surroundings entirely through his trusting, devoted eyes. As he slowly uncovers how he came to be there, the book becomes a meditation on finding beauty and meaning in a world we only partly understand.

What it offers:

Disorientation lies at the core of future shock. Clarke offers another way to meet it, with curiosity and gratitude in place of fear. Her novel suggests that we can love a world we don’t fully comprehend. That outlook lets anyone living through rapid change find wonder in the present before making sense of every development.

A World Appears by Michael Pollan

What it is:

This 2026 work of nonfiction is Pollan’s investigation into consciousness, the inner experience of being aware. His earlier writing on psychedelics had left him wondering how subjective experience arises in the first place. To find out, he draws on neuroscience, philosophy, literature, and contemplative practice. By the end he has moved from treating consciousness as a problem to solve toward appreciating it as a marvel.

What it offers:

Constant change can lock us into a narrow, vigilant kind of attention, always scanning for the next development. Pollan describes loosening that focus in favor of a wider, more receptive awareness. Along the way he finds that accepting mystery, instead of demanding answers, leaves him more open to the world around him. Although he didn’t write the book as a response to technological change, his example offers a way to meet our own uncertain moment with curiosity and awe.

Pale Blue Dot by Carl Sagan

What it is:

Carl Sagan’s 1994 book was inspired by the 1990 Voyager 1 photograph, taken at his urging, that shows Earth as a single speck of light. He uses that image as the starting point for a meditation on humanity’s place in the cosmos. Blending astronomy with reflection on where we came from and where we might go, the book is Sagan’s attempt to show what such a view can teach us about ourselves.

What it offers:

Rapid change pulls our attention toward the latest development, often at the expense of the larger story it belongs to. Sagan supplies that larger story on the widest scale imaginable. Seen from the edge of the solar system, the upheavals of any single decade shrink considerably. Our shared planet, meanwhile, starts to look more precious for its fragility. He also treats the urge to invent and venture outward as one of humanity’s oldest traits, which places our own restless era within a pattern as old as our species.

How to approach this reading list

Toffler’s advice for living through rapid change was practical: accept change in some parts of life, and deliberately build stability in others. The books on this list work toward that balance in different ways, so the best place to begin depends on what the moment feels like.

If the moment feels unprecedented: The Victorian Internet makes a good first step. The telegraph inspired its own mix of wonder and alarm before becoming ordinary infrastructure. Too Much to Know reaches further back, to scholars who built tools for coping with a flood of printed books.

If change feels like loss: The Leopard makes room for mourning a passing world while staying curious about the new one. The Beginning Comes After the End offers a milder reading of those losses, in which endings clear the way for what comes next.

If AI itself is the source of unease: Klara and the Sun lets us live alongside a machine companion and feel out what that relationship can and cannot give. Exhalation offers clear thinking about how new inventions change the way we remember and decide.

If we want to imagine a hopeful future: A Psalm for the Wild-Built offers the gentlest vision here, built around the question of what people need. Death of the Author imagines artificial minds searching for a purpose of their own.

If we want a way to judge new tools: The Real World of Technology helps us ask of every app and assistant whether it leaves us in charge of our own work. Its question applies just as well to technologies that haven’t arrived yet.

If keeping up feels impossible: Four Thousand Weeks explains why that race can’t be won. In its place, it offers a daily practice of choosing what deserves our finite attention.

If we need an anchor: The Bhagavad Gita offers a practice we can carry anywhere: doing our work well while letting go of outcomes no one can predict. For stories about building a life around what stays constant, there are A Gentleman in Moscow and The Lord of the Rings.

If screens are crowding out the people around us: Reclaiming Conversation makes the case for turning back toward face-to-face talk, especially now that chatbots offer conversation on demand.

If we want to appreciate the present: What We Can Know views our era from a century ahead, through a scholar who envies freedoms we take for granted. Time in that future may make our own days easier to treasure.

If we need perspective: A few poems from Map or Life on Mars can lift our attention from the newest headline to the oldest questions. A chapter of Pale Blue Dot widens the view to the scale of the universe.

If the world feels too strange to understand: Piranesi follows a narrator who meets a bewildering world with wonder instead of dread. A World Appears brings a similar openness to the mystery of our own awareness.

An honorable mention, for families with children: Raising or teaching children right now means deciding how AI fits into young lives, often before anyone knows its long-term effects. Human Raised: Nurturing Connection, Curiosity & Lifelong Learning in the Age of AI offers a framework for those decisions. At its center is a question to ask of any new app or device: does it support a child’s connection with the people around them, or crowd that connection out? Having that question ready makes each new product less overwhelming to judge, however quickly it arrives.

Reading these titles won’t slow the pace of change, but it can give us firmer footing as new developments keep coming. Each of us will reach for something different depending on the day. The list is meant to be dipped into in any order.

Reading our way forward

These books offer another way to read the disorientation of rapid change. They span centuries and genres, yet many treat unease as a sign that our expectations have fallen out of step with the world. We can use that feeling as a prompt to revisit what we have taken for granted and update our understanding.

This reframing changes the question an accelerating world keeps pressing on us. Instead of asking how to keep up with everything, we can decide where to welcome new developments into our lives and where we want things to stay constant. When we make those choices deliberately, we gain a say in how the future enters our days. We can then meet each new tool with questions we have practiced asking: What does it add? What does it push aside?

The future will keep arriving faster than many of us would choose. We can still decide, one development at a time, how we want to live alongside it. Each decision restores a piece of the familiar ground that future shock can erode. Over time, those choices add up to a life that feels like our own, even as the world around it keeps changing.

What are you reading to keep your bearings right now? I’d love to know which books help you meet the strangeness of this moment. If you’d like to share titles in the comments, let’s build this reading list together.

The reading list, in full:

Precedent: we’ve been here before

• The Victorian Internet by Tom Standage

• Too Much to Know by Ann Blair

• The Leopard by Giuseppe Tomasi di Lampedusa

• The Real World of Technology by Ursula Franklin

• The Beginning Comes After the End by Rebecca Solnit

Rehearsal: practicing the future in imagination

• Klara and the Sun by Kazuo Ishiguro

• Exhalation by Ted Chiang

• A Psalm for the Wild-Built by Becky Chambers

• Death of the Author by Nnedi Okorafor

• What We Can Know by Ian McEwan

Stability zones: anchors that travel with us

• Four Thousand Weeks by Oliver Burkeman

• Bhagavad Gita, translated by Eknath Easwaran

• A Gentleman in Moscow by Amor Towles

• The Lord of the Rings by J.R.R. Tolkien

• Reclaiming Conversation by Sherry Turkle

Awe: widening our sense of time and scale

• Map: Collected and Last Poems by Wisława Szymborska

• Life on Mars by Tracy K. Smith

• Piranesi by Susanna Clarke

• A World Appears by Michael Pollan

• Pale Blue Dot by Carl Sagan

You might also like:

• What a Librarian is Reading in Times of Moral Nausea and Psychological Despair: A reading list for when more information makes things worse.

• How to Read a Book: A librarian’s guide to a richer non-fiction experience.

• A 30-Day Protocol to Rebuild Reading Capacity: Progressive overload applied to the page, for readers who miss reading.

  ---

  The free essays are the foundation.

  The paid tier is the applied toolkit: biweekly AI briefings, monthly subscriber-driven research, and quarterly guides that give you real skills you can use immediately, plus full access to the Card Catalog Classroom (home of the Reference Collection).

  Upgrade to paid if you want the full Card Catalog.

  Subscribe now

  ---

  Share Card Catalog

It’s my birthday week, and I’m celebrating with 20% off the first year of an annual subscription through October 4! Join us here.

Research used to be rationed by access. Answering a question thoroughly meant a library within reach, a database subscription, a specialist willing to take the call, or the time to dig. A parent deciding between two schools, a tenant reading a lease, a patient with a test result, a volunteer writing a grant application at the kitchen table: each of them was doing research, often without the tools a professional would have reached for. The gap between the resources a question called for and the resources within reach was something many of us learned to work around.

An AI tool closes part of that gap in an obvious way, and part of it in a way that’s easy to miss. The clearest value is speed. An overview of an unfamiliar subject, which once meant working through several books or articles, now comes back from a few questions typed into a chat window. The part that’s easy to miss is range: the number of subjects we can begin researching at all. Starting on an unfamiliar subject used to depend on knowing a little about it first, since a library catalog or a search engine only helps once we know what the subject is called. An AI tool has been trained on data from far more fields and languages than any one of us could read, so it can supply that starting point. Given a description in our own words, it can tell us the term experts use. From there it can point us to the people who study the subject and the questions we didn’t know to ask. That second change is the larger one, because it widens the spectrum of subjects any of us can look into.

But the AI tool that opens those subjects brings no reliable discernment to its own output. It’s polished whether it’s right or wrong. Its claims can arrive in the same confident voice regardless of how well supported they are, and even when nothing supports them at all. Non-AI sources show us where claims come from. A book names its author and publisher, and a web page sits under an address we can look up to see who runs it. When a librarian isn’t sure of an answer, they can tell us so. The AI tool’s replies don’t reliably carry the same marks. The work of research shifts accordingly, from finding sources toward evaluating and trusting them.

AI tools change the pace of the research task without changing what the task is. Research has stages, mapped long before there was a chat box: a question forms, a subject is chosen, sources are explored, a focus emerges, material is gathered, and something is done with all of that material. At some stages, the AI tool is close to the ideal research partner: broad in its range and tireless in its efforts. Other stages call for more caution, especially the one where we decide which facts to keep and repeat to other people. The stage we’re in tells us whether to take the AI tool’s answer as a starting point or to check it first. That single piece of orientation turns a machine that can’t be trusted on its own into a formidable research system.

Reader Question

“I run a small nonprofit and I’ve started using ChatGPT for the research I used to do myself: background on the towns we work in, what the rules are for the grants we apply for, that kind of thing. It’s fast and the answers read well, and it’s opened up subjects I wouldn’t have known where to start on. Last month I put one of those answers into a board memo, a figure about local unemployment with a source attached. A board member looked up the source and the report didn’t exist. I was mortified. I don’t want to give the tool up, but I also can’t put anything in front of my board again without knowing whether to believe what comes back. How do I research with this thing properly?”

— P.K.

A note before we begin: The opening, free to read, considers what AI tools change about research and follows a board memo through the six stages of the process. The paid section sets out the twelve-step method for researching with an AI tool, and includes a downloadable guide with fuller versions of the prompts used throughout the piece.

Librarian Answer

Research can look like a single activity—looking things up—but it unfolds in a sequence of stages. Carol Kuhlthau, a professor emerita at Rutgers University, spent more than two decades studying how students, library users, and working professionals do research, recording what they did and how they felt at each point. Her model, called the Information Search Process, describes how research unfolds in six stages. People move back and forth between the stages rather than going through them in strict order. For example, sometimes the reading shows that we asked the wrong question. Someone who starts with “Are electric cars better?” may find, a few articles in, that the question they need answered is whether one suits their own driving and the charging available where they live. That discovery sends them back to an earlier stage to choose the new question. The same thing happens later on: once we’ve reached a tentative answer, we may meet an argument against it that we can’t yet answer, and we go back to reading until we can. Each return used to mean redoing earlier work, from working out new search terms to finding a fresh set of sources. An AI tool cuts down that repeated work. We can tell it what we’ve learned so far and ask for a revised question and the search terms that fit it, so what we already know can carry forward instead of being rebuilt each time.

Kuhlthau’s best-known finding concerns the third box, Exploration. Reading more at that stage tends to lower our confidence, because the sources start to contradict each other. That’s the point where people are most tempted to stop reading and grab an answer. And AI tools make grabbing one easy: asked to sum up, it produces a confident summary that smooths over the disagreements. Having a single answer brings relief, and that relief can feel like we’ve figured the subject out, when really we’ve only been handed one version of it. The disagreements the summary smoothed over are what we needed to see; feeling that relief during Exploration is a cue to keep reading. Collection is the only stage where we accept a fact as true. Presentation is the only stage where anyone else sees the result.

Kuhlthau’s stages describe how research unfolds. Information literacy describes the abilities we bring to it. The American Library Association’s longstanding definition says an information-literate person can recognize when information is needed and can then locate, evaluate, and use that information effectively. Each of those abilities comes into play at a different point in the sequence: recognizing the need at Initiation and Selection, locating sources during Exploration, evaluating them as a focus forms and material is collected, and putting them to use at Presentation. An AI tool changes none of those skills, but it does change how quickly we move through each stage and how much information each stage can draw on.

Set against those six stages, the memo’s path in the reader’s question is easy to trace. Initiation and Selection went well: the board needed a picture of the local economy, and unemployment was a reasonable way in. But Exploration barely happened, because the AI tool’s first answer arrived as a finished summary and there seemed to be nothing left to explore. Formulation was done by the machine, which decided what the figure meant and how to say so. The invented report got in at Collection, where the citation was kept before anyone checked that the report existed. At Presentation, the unchecked figure went to the board. The friend’s advice to stop using the tool would’ve prevented the memo, but it also would’ve prevented everything the tool did well in the first two stages. The fix is smaller than giving the tool up: one stage needed a check it didn’t get.

The answer to our reader changes with the stage. P.K. was using the tool where it is most useful: opening subjects and sketching the terms of a field’s argument. It’s least reliable at Collection, where the invented report entered the work. The six stages show how the task changes as research moves from a first question toward something we’re ready to share. At each point, they clarify what an AI tool can help us do and what we still need to check ourselves. Put into practice, they become a short sequence of steps we can use on the next research question we bring to any AI tool.

How to research with AI, stage by stage

Read more

Quick summary:

Reading a nonfiction book as more than its chapters changes what the book can give us. The pages around the chapters tell us how its argument is built, when it was written, who made it, and which conversation it joins. With that context in hand, every chapter carries more meaning.

Key takeaways:

→ Reading the table of contents before chapter one lets us follow the argument as it builds, step by step.

→ Our own marks in the margins and a running note on the phone turn the book into a conversation we can return to.

→ The notes and the bibliography hand us our next book, chosen by someone who spent years in the field.

It’s my birthday week, and I’m celebrating with 20% off the first year of an annual subscription through October 4! Join us here.

Cracking open a new nonfiction book is one of the finest adventures a mind can go on. It’s portable and immediate, asking for almost nothing but curiosity, and not just a willingness to give sustained attention to one subject but a deep, abiding welcome of the attention itself. The book becomes both mentor and sparring partner all at once. For a little while, its subject occupies our thoughts, in the most delightful way. We dive deep into some niche and come up with our minds fuller, fed by the richest kind of mental nourishment.

There’s far more to a nonfiction book than its chapters, though. Most of us turn to chapter one, where the writing we came for begins. The pages at the front and back are less familiar and can look like formalities, so they tend to stay unread.

The pages we pass over carry records left by the people who made the book. A finished book passes through many hands, from an editor who pushes back on its logic to a designer who arranges its pages. Several of those hands leave a mark we can read. The publisher dates the book and names itself on the copyright page. A cataloger, who read the book before it reached a shelf, wrote the short block of catalog information printed on that same page. Behind the last chapter, the author’s own notes and acknowledgments sit next to an index, which an indexer built by combing every page for what a later reader might want to find.

Read alongside the chapters, those pages give a book a place in time and a circle of company that the chapters alone can’t. With the table of contents in mind, we walk into each chapter knowing what the author is trying to prove and how far along the argument we’ve come. A year on the copyright page lets us hear the book as a voice from its own moment, one whose claims carry a date and whose confidence we can weigh against everything learned since. The notes and bibliography introduce us to the company the author kept, the writers the author argued with and admired, and turn a single voice into a conversation we’re invited to join. An index lets us come back long after and find the passage that stayed with us, keeping the book ours past the last page.

In 1940, the philosopher Mortimer Adler published How to Read a Book, later revised with the editor Charles Van Doren. Adler taught readers to survey a book’s structure before reading it closely. A librarian begins there too, then asks different questions. Who made this book? What work does its argument rest on? Where does it sit among the other books on its subject?

Every book we read has a life with us that runs longer than the reading. It starts in a bookshop aisle or a library app, with a title that catches our eye, and can carry on for years after the last page, in a passage we return to or a book it led us toward. The pages around the chapters travel that whole distance with us. They do the most for us when we know which ones to reach for as we go, beginning with the moment we first pick a book up.

Before we start

We can learn a lot about a nonfiction book before reading any of it. With the book in hand, a few pages at the front and back answer two questions. The first is whether the book covers what we want to learn, and in how much depth. The second is what kind of reading we’re in for, from how recent the research is to how the argument is laid out. Answering those questions before chapter one means we start the book knowing what it can give us. On the next nonfiction book we pick up, these are the places to look:

• Check the year on the copyright page against how quickly the subject has changed.

• Look for an edition number on the copyright page. A second or later edition often opens with a new preface on what the author changed.

• Turn to the page about the author and the acknowledgments to see who trained the author and who funded the work.

• Scan the subject headings in the cataloging block on the copyright page, if the book has one. A heading the cover never mentions may point to a subject the book takes on and the marketing set aside.

• Read the chapter titles straight through, in order, before chapter one.

• Read the last paragraphs of the introduction, where authors often preview each chapter.

• Glance at the final pages of the last chapter, where the conclusion often states the whole case.

• Look up the one topic we came for in the index at the back. A lone page number may mean the book barely covers the subject. Many page numbers, broken into subentries, usually mean the book covers it in depth.

Glancing at these pages answers the questions we brought to the book. A slower look turns up more: how an argument changed across editions, what the book’s subject headings reveal, and where the author places their work among other writers. The first page to turn to sits behind the title page, where the copyright page records when the book came out and who published it.

The copyright page

• What it is: The publisher prints this page on the back of the title page to record who owns the book’s rights and when it came out.

• What it’s for: It names the house that produced the book and gives the year of publication, our starting point for asking how current the research is and how the book reached print.

• What it gives us: We can read each claim alongside the year it was made and notice which parts of the book newer work may have revised.

The copyright page sits on the back of the title page. Publishers use that page to record who owns the rights to the book and the year it was first published. The year is the line to read first, because it tells us when the book entered the conversation. A study of public health from decades ago can still be brilliant, though its statistics describe a world that has since moved on. With that year in mind, we can read the book as a report from its own moment, written by someone who could only know what was knowable then. The year tells us which claims may need checking against newer work before we repeat them. How much the year changes our reading depends on the field. Thirty years may matter less to a history of ancient Rome, where the central evidence often changes more slowly than it does in nutrition or computing.

A printing is a new batch of the same text; an edition changes the text itself. Many books show their printing history as a row of numbers, sometimes descending: 10 9 8 7 6 5 4 3 2 1. Printers call it a number line. Publishing conventions vary, but the lowest number still standing usually identifies the printing in our hands. A row whittled down to 7 means six earlier batches of the same edition were printed before ours. New editions often arrive with a new preface, where authors explain what they changed and why. Reading that preface first shows us where the argument shifted before we commit to the rest.

The publisher’s name on this page gives us one clue about how the book reached print. University presses often publish scholarly work and commonly send manuscripts to outside specialists for review before publication. Trade publishers usually publish for a general readership, with editors working closely on a book’s structure, clarity, and pace. Both kinds of publisher may seek expert advice, copyedit closely, or leave much of the factual checking to the author. Neither name guarantees accuracy. It tells us where to begin asking how the book was made, and how much checking we may need to do ourselves.

Besides the publisher, the person to get to know before we begin reading is whoever wrote the book. A page near the front or back gives the author’s training and earlier work. That tells us what they bring to the subject, whether a historian coming to a scientific question or a journalist who spent years inside an industry. The acknowledgments say more still, this time in the author’s own voice. They often name the people who read early drafts and the institutions that supported the research. Those names place the author inside a community of colleagues and mentors and show us the conversation the book grew out of. A history written from inside one school of thought, or a book funded by the industry it describes, can still be excellent. Reading the acknowledgments first tells us which community and point of view the book comes from. With that in mind, we can follow the author's arguments knowing the perspective behind them.

The cataloging block

• What it is: The cataloging block is a short catalog record printed on the copyright page, listing the book’s subject headings and shelf numbers.

• What it’s for: It gives libraries a standard way to describe the book and place it beside others on related subjects.

• What it gives us: Its subject headings show what the book substantially covers and give us a route to other books on the same subject.

The copyright page carries a second kind of entry, this one from a library rather than a publisher. On many books it appears as a boxed cluster of text lower on the page. Librarians call it the cataloging block, a catalog record printed inside the book itself. Its subject headings name the topics the book covers. Libraries use shared terms so books on one subject gather under the same heading instead of scattering across near synonyms. The classification numbers beneath them place the book on a shelf beside related work.

Not every book carries a cataloging block. When one does, its headings are still human choices, made inside an institution with its own history. Libraries revise their terms as language and understanding change, though the printed record stays as it was.

Reading the subject headings before chapter one shows us how a cataloger summarized the book in a handful of phrases. A history of a revolution might carry one heading for the war and another for the women of that era. That second heading tells us that women’s lives are substantial enough to the book for the cataloger to name them as a subject. When the headings line up with what the cover promises, the marketing has described the book fairly. When they name a subject the cover leaves out, we have found something the marketing set aside.

Each subject heading can be searched in a library catalog, including WorldCat. Search one, and the catalog gathers other books described with the same terms. The classification number can do the same work in a physical library: nearby numbers lead us toward books on related subjects. A cataloging block turns one book into a route through a shelf.

The table of contents and the introduction

• What it is: The contents page lists the chapters in order. The introduction often states the thesis and may preview what each chapter will do.

• What it’s for: Together they show the author’s plan before the argument begins.

• What it gives us: We meet each chapter knowing what work it is meant to do, which keeps the argument in view even in a long or difficult book.

Once a book has passed the checks on its copyright page and its cataloging block, the next pages show how its argument is built. The table of contents is the place to begin, since chapter titles read in order lay out the author’s case the way an outline lays out an essay. Mortimer Adler called this kind of survey inspectional reading: a look at a book’s structure before we read it closely. Reading the titles first hands us the author’s plan before we reach the argument itself.

The author’s plan, read off the chapter titles, lets us guess the argument before we’ve read a page. Picture a history whose chapters begin with a single town and end with the decisions of a distant capital. The sequence may suggest that the author sees local events as an explanation for national ones. Carrying that guess in our heads as we read turns every chapter into a small test of whether the author delivers. When the book surprises us, we feel the surprise. That feeling marks the place where the book changed our mind.

A contents page also shows proportion, which the text itself never announces. Page numbers beside each chapter reveal where the author spent the most space, often on the part of the case that needed the most defending. Books divided into larger parts offer another clue: their part titles show the argument’s main turns, with the chapters beneath them doing the work.

Authors of nonfiction often use the introduction to state their thesis outright and to name the scholars they’re arguing with, which turns a book that looked like a monologue into one side of a debate. Many introductions close with a preview of each chapter, which lets us check our reading of the contents against the author’s own summary. Adler also recommended glancing at the final pages before starting, because authors tend to restate the case there in its finished form. With the introduction and those closing pages read, we enter the text knowing where it begins and where it means to arrive, which frees us to watch how the author gets there.

While we’re reading

Even with the argument in view, attention tends to drift somewhere in the middle of a book. One cause is that we’ve only been receiving for a stretch, not responding. Our focus returns the moment we answer back. A mark in the margin or a question in a note keeps the author’s claim separate from our own response to it. When we come back months later, we can see what persuaded us and where we pushed back. Every habit that leaves such a record is one we already have, pointed at the case the author is building instead of at the lines we happened to like.

• Fold a corner or highlight the claims we doubt, alongside the lines we love.

• Keep one note on the phone or in a journal titled with the book’s name, and let the questions or comments pile up there.

• Park an unfamiliar name in that note and look it up later, rather than breaking off to search in the middle of a page.

• When a chapter ends, look at its notes at the back. They show the sources behind particular claims and may reveal where the author sees uncertainty or disagreement.

• Glance at the dates in the bibliography to see whether the author draws on recent work where the field calls for it.

Each of these habits fits into the reading we’re already doing. Kept up across a whole book, they leave a record of our exchange with the author, one that stays on the page and in our own notes after the reading is done. That record leads us naturally to the back of the book, where the author’s notes and bibliography show us the work beneath the argument.

What we add to the book

• What it is: Marginalia are the notes we write in a book’s margins. A commonplace book is the notebook where we collect passages and questions from everything we read.

• What it’s for: They turn reading from something we absorb into a conversation we take part in.

• What it gives us: The record we leave of what persuaded us and what didn’t keeps the book ours to argue with and return to.

Every page covered so far came from someone who helped make the book. Marginalia and the commonplace book come from us instead, written in as we read. By the time we reach the text, we know the argument the author intends to make. Each chapter becomes a step in the case. We can ask whether its evidence supports the claim it was meant to prove. The book becomes an argument we test.

Marginalia are the notes readers write in the margins. Plenty of us already keep a version of them without using the word. A folded corner marks a page we mean to return to. Highlighting in an e-reader, or photographing a page with a phone, saves the line that struck us. Pointing those habits at the argument, instead of at the lines we happened to like, turns reading into an exchange. A question mark beside a sweeping claim marks a place to push back. Underlining a quoted source flags a work to find later.

A commonplace book is a personal notebook where a reader copies passages and questions from everything they read. Keeping one lets us carry questions from one book to the next. The English philosopher John Locke devised an indexing method for his commonplace books so he could find an entry again by topic. The commonplace book’s modern form is the phone in our pocket. A notes app does what Locke’s index did, with a search bar instead of a system. One running note titled with the book’s name, where the photos and saved lines collect, is a commonplace book in everything but the binding.

When a book mentions an event or a thinker we don’t recognize, add the name to the running note and look it up later. We stay with the page for as long as it has our attention. Over time, those notes become a log of how different authors treat one subject. Sooner or later, two of those authors disagree. At that moment, the note turns into a research tool of our own. The back pages of a book with full notes and a bibliography supply that tool with its material.

The notes

• What it is: Notes are numbered entries tied to specific passages in the text, printed at the foot of each page or gathered at the back of the book.

• What it’s for: Each note leads us to the source behind a claim and may add the author’s own comment on it.

• What it gives us: Read alongside the chapter, notes show us how an argument rests on evidence and where the author’s interpretation begins.

Many books gather their notes at the back, though the notes belong to the reading itself. Each one is tied by a number to a passage in the chapter. Flipping back as we go, or reading a chapter’s notes the moment it ends, keeps each claim beside its evidence. Any source can be judged by what it was built on. The notes show that foundation claim by claim.

In the notes, many authors write more candidly than in the chapters. They argue with other scholars there and admit where the evidence runs thin. Reading a chapter’s notes right after the chapter shows us how a claim rests on evidence and where interpretation begins. A long note often marks a claim that needs more explanation than the chapter can hold. It may be where the evidence is contested, where the author qualifies the point, or where another argument waits just off the page.

Some nonfiction books carry no notes at all. Their claims may still rest on extensive reporting, archival work, or expert consultation, but the path back to the evidence is harder for a reader to follow. Look for a page headed Sources, Further Reading, or Selected Bibliography, and read the acknowledgments for the experts, archives, and institutions the author names. When a claim has no trail inside the book, write it down and check it against other reliable work before carrying it forward.

After the last chapter

Finishing a book leaves one very important question hanging: what to read now? Without the back pages, the answer comes from a shop display or a recommendation feed, chosen by systems that never read the book we just finished. The book itself fades too, into an impression we can’t pin to a page. Both outcomes change once we use the index and the bibliography. The index reopens the book we just closed, so a passage we can be found again. A bibliography points past that book toward the work the author chose to name, a list drawn by someone who spent years in the field rather than by an algorithm that noticed what we bought.

• Read the densest entries in the index as a second outline of what we just finished.

• Look in the index for a name or idea we expected to find. Its absence marks where the book stops, or what the author chose to leave out.

• Find a source the notes return to often. Read its description, then decide whether it’s the next book to follow.

• Add the question the book left unanswered to the note on the phone, ready for the next author to answer.

Those steps turn a finished book into a starting point. The index keeps the book within reach, ready to give back any passage we want to find again. From there, the bibliography and the notes lead outward, toward the next book on the subject and the one after that.

The index

• What it is: At the very back, the index lists the names and concepts a reader may need to find, alphabetically, with the pages where they appear.

• What it’s for: An indexer built it by judging which ideas a future reader would come looking for.

• What it gives us: The index shows us what the book returns to and lets us find a passage long after we’ve finished.

The index waits at the very back, an alphabetical list of names and concepts, each followed by the pages where they appear. We tend to treat it as a lookup tool. It’s also a reading of the whole book, usually made by a professional indexer who studied the finished text and tried to predict what a future reader would want to find. An indexer decides which ideas deserve entries of their own, which belong beneath a broader heading, and which passing mentions do not need to be recorded. The finished index describes the book as a reader might search it, not as the author wrote it.

A concept with a long run of page numbers, broken into subentries, is one the argument returns to again and again. One with a single page number may receive only passing attention. Scanning for the densest entries gives us a second outline of the book, based on what the text dwells on rather than on the chapter titles. The two outlines need not match exactly. Where they differ, we can see which subjects the book cares about most, and whether the book will serve the question we came with.

When a book leaves out a name or idea we expected to find, the index shows the gap immediately. The gap might reflect a deliberate decision about scope. It might also point to a perspective the author overlooked, which gives us a question to take to other sources.

The bibliography

• What it is: A bibliography is an alphabetical list of sources placed near the end of a book.

• What it’s for: It names the work the author chose to place behind the argument.

• What it gives us: It shows us the book’s intellectual company and gives us a path to the next book.

The bibliography comes into its own once we’ve finished, with the whole argument in view. Its entries name the books, articles, archives, and other work the author chose to place behind the argument. Read after the last chapter, the list shows us the company the book keeps and gives us a way to follow the conversation beyond it.

Begin with the names that recur in the notes, or with a title whose subject pulls at the question the book left open. A bibliography can include primary evidence, later interpretations, and work the author mentions only to disagree with. The entry tells us enough to find the source; the chapter and its notes tell us why it mattered to this author. Reading one source from that list lets us see the argument we’ve just finished from another side.

Some books close with a bibliographic essay or a section on further reading, where the author describes major works in the field in their own words. Those pages are a reading list the author has shaped for us, placed after the last chapter. They often show which works the author thinks matter most, and where their own view differs. That guidance lets us choose our next book with the author’s judgment beside our own.

What changes for us

Reading this way changes the experience of every page. Each one carries two stories at once: the argument on its surface and the human decisions underneath. We can follow both. A passage strikes with more force once we know which writer it was answering. Surprises hit harder, too, because we might’ve made a guess of our own. The chapters grow denser with meaning, the way a familiar song deepens once we learn what the songwriter was influenced by.

By the last page, the book has also become ours. Our marks in the margins and the questions gathered on our phone turn it into a record of our own thinking. Returning to it months later, we find the author’s argument and our response to it waiting side by side. A book read this way stays with us long after it’s back on the shelf.

A book read whole also shows us what to read next. The pages around its chapters point to the books the author learned from and to others on the same subject, so we can pick our next title from those leads instead of waiting for a class to assign one or a bestseller list to suggest another. A question it raised and left open can lead us to the writer who takes it up. Followed from one title to the next, those choices let us learn a subject in the order our own curiosity sets. The first steps are already printed at the back of the one we’ve just finished.

You might also like:

• What Scrolling Did to Reading: Why books feel harder, what the science shows, and how to rebuild the capacity.

• A 30-Day Protocol to Rebuild Reading Capacity: Progressive overload applied to the page, for readers who miss reading.

• How Saving Replaced Reading: Save-for-later doesn’t always turn into reading later.

This week is my birthday, and I’d love to celebrate by building Card Catalog together! Through October 4, the first year of an annual subscription is 20% off.

Join us here

Share Card Catalog

It’s my birthday this week, and this year I have so much to celebrate with you!

Last December, I sent out my very first post, hoping a few people would be as fascinated as I am by how we find, filter, and feel about information, and by the invisible systems that shape what we know. Today, more than 31,000 of you read Card Catalog, from 165 countries around the world, including every state in the US. I still can’t believe it.

This year, “Google Has a Secret Reference Desk” was viewed nearly half a million times, and readers have filled the comments with tools, sources, and ideas of their own. Between that piece and so many others, I’ve learned so much from all of you, and I wanted to mark this birthday by celebrating that.

So for my birthday week, annual memberships are 20% off your first year, now through Sunday, October 4.

Join us here

If you’ve ever wondered what members get, here’s a taste of what’s gone out:

→ Librarian Hotline: Think With AI by Setting Up Custom Instructions, a guide to setting up custom instructions so AI works as a thinking partner instead of just an answer machine.

→ AI Briefing: AI in the Exam Room, The Skills AI Can’t Replace, and a Warning at the G7, from the AI Briefing that goes out every other Friday, so you can keep up with what’s changing without drowning in it.

→ Research Packet: Organize Your AI Conversations Like a Librarian, which uses library cataloguing to turn your chat history into a searchable knowledge base. It comes with templates and a quick reference card you can keep.

→ The Card Catalog Classroom on Circle, home to the reference collection, where I organize every Note I publish so you can search it and come back to it anytime. Here’s a peek:

→ Study Hall, a card game for practicing verification skills in real-world situations.

And there’s so much more on the way:

✦ New lessons in the Classroom, rolling out before the end of the year. I’ve mapped out the full lesson plan across four sections: AI Literacy, Digital Culture, Research Skills, and Source Evaluation. Each lesson takes one of our articles a step further, with worksheets to help you put it into practice.

✦ A guide to anti-brain rot, with ways to rebuild focus and make your reading and scrolling time feel good again.

✦ A Field Guide to the Intellectual Life, a directory of the spaces, places, communities, institutions, and experiences where people come together to think, wonder, and learn.

✦ The Global AI Directory, a way to compare how countries around the world are approaching AI, from regulation and policy to infrastructure and public opinion, so we can see side by side how different places are handling this moment.

✦ The Library World Tour Map, an interactive map of extraordinary libraries around the world, each one pinned and profiled, and the first step toward a library travel show I’ve dreamed of making for a long time.

✦ The Reader Recommendations Directory in the Classroom. Some of the best ideas on Card Catalog live in the comments, from tools readers swear by to sources they trust. I’m gathering them into one organized directory, so the wisdom of this community is easy to find instead of getting lost at the bottom of a post.

I’m putting so much care into each one, and I’m incredibly excited to share them with you as they’re ready. These are the topics I find myself researching all the time, and getting to explore them with thousands of curious readers around the world has become the most fulfilling work I do. It’s what dreams are made of!

If you’re already a paying member, thank you, and I hope you love everything that’s coming down the line. You’re the reason all of this exists.

Every free post will always stay open and accessible to everyone, and members are what make that possible. If you’ve been thinking about joining, I’d love to welcome you in. This week feels like a perfect time to celebrate what we’ve built together.

Join us here

With so much gratitude,
Hana

P.S. The discount lasts through Sunday, October 4, and covers the full first year of your membership.

P.P.S. I read every message you send me. Between Card Catalog and my full-time day job, replies can take me a little while, but if you’ve written and haven’t heard back yet, I’ll get back to you as soon as I can. Happy researching!

Share Card Catalog

It’s my birthday week, and I’m celebrating with 20% off the first year of an annual subscription through October 4! Join us here.

Welcome to your AI Briefing, where I wade through the news chaos so you don’t have to. Every other Friday, we cover three stories: what happened, why it matters, what it means for our lives, and the bottom line.

This week:

• Armed US troops were preparing to board a Chinese ship this spring. Then officials took a closer look at where the intelligence came from.

• The researchers who’ve spent decades listening to dolphins and elephants have a message for anyone promising an animal translator.

• A new assistant books tables and buys tickets on our behalf, and its early testers are finding out what else it decides on its own.

US military had close call after using AI for false intelligence report, sources say

What happened

This spring, during the US war with Iran, an intelligence report circulated across the American military claiming that a Chinese ship in the Middle East was carrying components for a nuclear weapons program. The military moved to intercept the vessel, according to four people familiar with the episode. Two of them said armed personnel were preparing to board. Military planes were in the air, according to one of them and another person familiar with the incident.

Officials examined the report more closely just before the planned operation. They found that an analyst at a special operations command had produced the document with help from an AI chatbot. The chatbot had incorrectly identified what the ship was carrying. One source described the report as entirely false and said the episode almost started a war. It’s also unclear whether the chatbot was a commercial product or one built for the US government.

The analyst had asked the chatbot about intelligence reporting on the ship’s official cargo list. Working from that material, the chatbot combined publicly available information with classified signals intelligence in government files and reached its own conclusion about the cargo. The analyst then used AI a second time to turn the findings into a standard intelligence report and circulated the document. Neither the Pentagon nor the special operations command involved responded to a request for comment.

Why it matters

Officials across the US military and intelligence agencies are pushing AI into much of their work, from analyzing intelligence and supporting targeting decisions to managing budgets and supply chains. In January, US Defense Secretary Pete Hegseth released an Artificial Intelligence Acceleration Strategy. A memo announcing it set the aim of putting AI models in the hands of the department’s three million civilian and military personnel, from routine work to the most secret. The rationale, here, is speed. Supporters argue AI can help commanders decide faster which targets to strike and where to move forces. Officials say the US cannot afford to fall behind China or other adversaries that might otherwise stay a step ahead.

This episode points to a different danger from the one dominating the debate. Washington policymakers have spent weeks arguing over warnings from Silicon Valley engineers and executives that AI could break free of human control. The more immediate problem here is people making consequential decisions on inaccurate information that an AI tool produced and then packaged in a trusted format. Multiple US officials say the rollout is decentralized, with different parts of the government using different tools under different orders and safety standards. The US has no single set of standards for verifying what these tools generate. One person familiar with the episode said fabricated output of this kind has not been an isolated incident in the intelligence community since the tools spread through government. A former senior official described the internal systems as mostly commercial products with limited changes. If that description is accurate, the errors familiar from everyday chatbots can surface inside classified work as well.

What it means for us

An intelligence report can set off decisions whose consequences travel far beyond the room where the document was read. A confrontation between the US and China, two of the world’s largest economies, would affect far more than the ship and the troops preparing to board it. Conflict between powers of that size could reach the prices of everyday goods and the safety of people in countries with no part in the dispute at all. When a decision like that starts with a machine’s mistake, the consequences still fall on people with no connection to the error. Boarding a foreign ship is also a step a government cannot easily reverse once tensions begin to climb.

The reach of these decisions grows as AI becomes part of how governments work. A mistaken claim that might once have depended on a single person’s judgment can now be produced and passed along by software before anyone questions its source. That leaves the stability many of us rely on, from predictable prices to peaceful borders, partly dependent on how closely each government checks its machines’ work. Every country adopting these tools adds another place where a single unchecked answer could set larger events in motion.

Bottom Line:

More governments are building AI into how they gather intelligence and make decisions about when to act. Each one that does creates another place where an invented claim can begin, so the next close call may involve entirely different countries. Every government relying on these tools faces the same gap between how quickly a machine produces an answer and how long a person needs to confirm what it says. Rival governments will also face a new question about each other’s warnings: whether an alarming claim from across a border has been checked before anyone acts on it. Countries that build confirmation into the decision itself stand the best chance of keeping AI’s mistakes from becoming international crises. A shared expectation that AI-assisted intelligence is verified before anyone acts on it could become one foundation of stability between nations.

Using AI to ‘talk to animals’ might make us feel clever – but what, if anything, does it do for them?

Read more

Quick summary:

A feed narrows us because it learns from what we already do. Feed it something different, on purpose and more than once, and the same system starts carrying us toward things we didn't know were there. A wider feed makes a wider world, with more in it to want and more of it to live in.

Key takeaways:

→ A recommendation feed predicts what we’ll do next from what we’ve already done. It can be right about our past without knowing anything about where we’d like to go.

→ Trying something unfamiliar once barely changes what a feed shows us. Returning to it more than once is what changes the record the feed learns from.

→ What we can want depends partly on what we’ve been shown. A feed that only returns our own record shrinks the range we get to want from; a feed we’ve widened does the reverse.

This is a shorter companion to an earlier Card Catalog piece, “Our Horizon of Possibilities: How Algorithms Contract Our World,” which looked at how a feed narrows what we see and where to go when it does. This one is about the other half: what happens when we feed the same system something new.

We open a music app on the walk home, and the first song it plays is one we didn’t know we’d been waiting for. We scroll a shopping app, and there’s the coat we’d have picked ourselves—in our favorite color, no less! Each time, for a moment, the app seems to understand us. It’s easy to file that feeling under either good luck or good taste. But it’s the app that surfaced the song or coat for us, and receiving something we love can feel almost exactly like finding it on our own.

When we say the algorithm, we tend to picture an omniscient, tentacled entity with a mind of its own, a presence behind the screen that watches us and decides what we get to see. It gets the credit for a good song but also the blame for a lost evening. It’s often the word we reach for whenever something online feels arranged by an unseen hand.

A mysterious force can seem less powerful once we shine a light on it. So let’s start by defining our terms: what is an algorithm?

An algorithm is a set of instructions that turns an input into an output. A recommendation system combines many of them to decide what to place in front of us next. It draws on our record on the platform, the material available to show us, and the goals the company has set. To do that, it has to narrow an enormous field of possibilities into a short list.

That responsiveness creates both the problem and the opening. Left alone, a feed can draw us back toward the middle of our existing taste, because familiar material gives the platform its safest prediction. But the same system can respond when we give it a direction of our own. Deliberate searches, sustained attention, and repeated visits to material we’ve chosen for ourselves can give the feed a different sense of what to show us next. That won't erase the platform's goals, but it can change what the system puts within reach.

A feed that works for us can lead us somewhere: to the book that arrived from a direction we’d never have searched, or to an afternoon that left us with something we could still describe a month later. A less useful one fills the hours without giving us much to carry away, and keeps returning us to versions of the same thing. When the second description starts fitting our evenings better than the first, it is time to give the system a different direction.

How a feed narrows us

The narrowing is the system doing its job. An algorithm earns its keep by predicting what we’ll do next; from the platform’s side, a feed counts as a success when we stay and keep tapping. The safest prediction is then the near match, the thing closest to what we’ve already chosen. Novelty is a gamble it rarely needs to take, because the familiar performs so reliably. Some systems mix in unfamiliar material because surprise can keep a feed varied, but most feeds still give the strongest advantage to what resembles our past behavior. Ranking and the autoplay both reinforce that tendency. We experience the result as a stream that seems to understand us, but it’s a platform built to keep us on it.

Run that process a thousand times and it hardens into a pattern. Each near match becomes the starting point for the next one, and our range shrinks in steps too small to feel. The store that fit us so well ends up filling half the closet. The playlist that began with one artist arrives, many songs later, at variations on the same sound. A world that only ever gives us more of what we already knew we’d like is a smaller world than the one we started in.

The strongest pull is accuracy. The algorithm is right often enough that going elsewhere starts to feel like extra work for a worse result, and every good guess makes the next one easier to accept without looking further. The guesses come from a record of what we’ve already done, played back with precision, so they’re most reliable on ground we’ve already covered. Being right about our past isn’t the same as knowing where we’d go next, but from the couch the two are hard to tell apart.

Comfort has a critical role in all of this. Honestly, a lot of the time, more of what we like is exactly what we need; some might even call that nervous system regulation, with creating a sense of safety. A map that sends us to a restaurant it knows we’ll enjoy after a long flight is doing its job. On a night when there’s nothing left in us for a hard show, a familiar one is a small mercy. The trouble starts when the familiar becomes the whole of our experience and nothing we meet asks anything new of us. Comfort can do damage when we stop consciously choosing it.

There’s a case that the feed is already always working for us. Every guess it makes comes closer to something we’ve wanted, so the longer it runs, the better it serves our desires. That case assumes that our desires form a fixed list that the machine can read—and that we ourselves are fixed beings. But we don’t work that way, and neither do our desires. What we can want depends partly on what we’ve been shown. A system that returns only our established record narrows the range of things we can come to want. The record it reads also includes what kept us watching, and on a platform that earns from our hours, what keeps us watching can differ from what we’d choose for ourselves. The feed can treat what keeps us watching as evidence of what we want, even when it isn’t.

How to steer an algorithm toward what we want

The system learns from patterns in what we do. That’s how a feed can narrow around what already holds our attention. It’s also the small opening we have: change the pattern, and we can change some of what the feed learns to offer us.

A feed arrives fully built and asks nothing of us but our attention. A search starts somewhere else, with something we already want to find or understand, and it uses the platform to go looking for that. The following four changes can help steer what the feed shows us.

Choose a starting point before the feed does.

Open an app with one thing in mind: a book, a song, a recipe, a place, a person, a subject, or a question. A search begins with something we want to find; a feed begins with something it wants to show us. Search before the first recommendation has a chance to choose for us, then follow one result far enough to see where it leads. If a video mentions an unfamiliar writer, look for the writer rather than accepting the next video; if a song catches your ear, seek out the musician, the album, or the sound it borrowed from. The search may not always produce a better result, but it makes the first signal ours—and gives the system a record of an interest it didn’t invent.

Spend time on things a little outside our usual taste, and do it more than once.

The system learns from where our hours go, and time spent in the middle of what we already like teach it to keep circling there. Trying one unfamiliar thing a single time changes little, but repeated attention creates a clearer pattern. A whole afternoon deep in a craft we’ve never attempted, or an evening reading into a subject we know nothing about, can show the system that the interest is more than a passing detour. The recommendations may begin to rebuild around that new pattern.

Leave the feed when a detail catches your attention.

A good recommendation doesn’t have to begin another round of recommendations. When a detail raises a question—an unfamiliar writer named in an interview, a tool used in a repair clip, a place shown in a travel video, or an instrument in a song—pause the feed and follow that detail elsewhere. Read an article, look for a book, visit the artist’s own site, check a map, or find a local class or shop. The aim isn’t to search for more of the same, but to use one recommendation as a path to a person, source, place, or question beyond the platform.

Notice how a recommendation leaves us afterward.

Pause after following a suggestion and check the energetic or emotional residue that it left behind. Something we’re glad to have found means the recommendation served us. But a lost hour we can’t account for may mean the platform came out ahead. We can ask whether the recommendation pointed us towards something we sincerely want to follow. The answer shows whether it opened a path we would choose to take, or that it simply kept us engaged.

We’ll know this practice is working when the algorithm starts surprising us again. A name we don’t recognize shows up next to the familiar ones. The feed, which had felt like a mirror, may begin to offer routes into subjects, creators, and forms we wouldn’t have found through passive scrolling. The reward builds on itself, because a feed that surprises us usefully gives us more reasons to arrive with another question to start with.

What opens up

Feeding the algorithm on purpose won’t turn it into something it isn’t. The platform still decides what the feed rewards, and a handful of searches won’t undo years of a record built from whatever held our attention. The expansion begins slowly. But it’s still a change.

The feed shifts first. A name we don’t recognize appears beside the familiar ones. A subject we searched for once, then again, begins to lead somewhere that opens up our world: to the writer a video mentioned, the region a recipe came from, or the people working around a musician we liked. The recommendations begin to carry us a little past where we’d have stopped, the way a good shelf in a library does, toward things we wouldn’t have searched for because we didn’t yet know they were there.

A wider feed makes a wider world, with more in it to want and more of it to live in. What we can want depends partly on what we’ve been shown, so more of the world becomes available to want. The craft we try for an afternoon can turn into a Tuesday class. The subject we read into can become a conversation with someone who’s spent decades on the question. A musician we follow from one song can lead us into a whole scene we didn’t know existed. These things don’t arrive all at once, and they don’t come from the feed alone. They arrive because we asked, and kept asking, until the same system that once handed our past back to us begins to hand us a future of what we haven’t yet seen.

You might also like:

• Our Horizon of Possibilities: How Algorithms Contract Our World: Algorithms curate our feeds based on what we’ve already chosen. Understanding how that process works changes what we can do about it.

• In Praise of Serendipity: On the joy of finding what you weren’t looking for, and how to get lost on purpose.

• The Atlas of Learning: A librarian’s map of where knowledge lives, and how to find our way through the research world.

Share Card Catalog

Quick summary:

Claims used to be checked by editors and fact-checkers before they were published. But much of what reaches us now arrives without anyone having checked it first. So verifying a claim has become a skill each of us can build: a small set of moves for testing what we're told, and the judgment to pick the right move for what's in front of us.

Key takeaways:

→ What we read can carry every sign of having been checked and still have passed through no one. The look of editorial review is not the same as actual review.

→ Each time a claim is retold, from a study to an article to a post to a chatbot answer, something the original said gets dropped or warped. A confident, complete-sounding answer can just be a fragment of the truth.

→ Reading a claim can't tell us whether it's true, because a false claim can be written as well as a true one. Verification steps in and goes to what the claim can't control: understanding where it came from and who stands to gain.

Scroll to the end for a new game to practice verification skills.

In May 2025, a summer reading list ran in the Chicago Sun-Times. Fifteen titles filled the section, each with an author’s name and a sentence about the story. The list opened with a new Isabel Allende novel called Tidewater Dreams, billed as her first climate novel about a family and rising seas. Allende has published more than twenty books over four decades. Against that record, the recommendation looked entirely natural. But there’s a twist: she never wrote Tidewater Dreams.

Ten of the fifteen titles were invented, each one pinned to a well-known novelist. A reader could move down ten entries before reaching a book that existed. None of the fabrications came from the paper’s newsroom itself, however. The reading list appeared in an extra summer section supplied by an outside company and included in the Sunday paper. But nothing on the page set the section apart from the paper’s own reporting. It appeared as unremarkable as the weather section, and carried the authority that print has earned for over a century.

The writer was a freelancer who had drafted the section with a chatbot and skipped the check he would normally run. If we follow the trust backward from the printed page, it forms a chain: the reader trusting the paper ← the paper trusting the outside company that supplied the section ← the outside company trusting the freelancer ← the freelancer trusting the machine. Each link was leaning on the one before it to have done the checking. But no checking happened anywhere during the process.

For most of the last century, trusting that list would have been reasonable. A reader could’ve looked for a title in a library catalog or searched online, but a summer reading list printed inside a newspaper didn’t ordinarily call for that kind of investigation. Its placement in the paper gave readers reason to assume that basic facts had been checked before publication. The list showed what happens when there’s the appearance of editorial review even though the review itself hadn’t actually taken place.

How information reaches us now

Checking claims once formed part of the daily work of librarians, reporters, editors, and fact-checkers. A reference librarian could trace a fact back to its origin and set one account beside another before treating either as reliable. At a magazine, a fact-checker could call a source, confirm a date, or compare a quotation with the interview or document that produced it. Editors could ask whether a writer’s evidence supported the sentence on the page. Readers still brought their own judgment to what they encountered, but published material often passed through checks that compared quotations with the original interview and statistics with the report that supplied them before publication.

That arrangement has been evolving for years. The internet gave companies, organizations, and individuals a direct way to publish. Search engines made those pages easy to find. Social-media platforms made it easy to repeat a sentence from an article without bringing along the source that explained it. And generative AI changed the order again by producing a direct answer before readers have seen the documents behind it (if they see them at all).

A claim can change at every stage. A research paper may explain which people took part in a study and describe the conditions of the research. A news article may shorten those details for a general audience. A social-media post may retain only the most striking number. A chatbot may turn that shorter version into a complete-sounding answer. The answer itself can repeat part of the research accurately, though it may not show what the researchers measured, where their findings stop, or which qualifications shaped the original result.

Information literacy gives us a way to examine how a claim reached us. An article, product page, social-media post, or chatbot response may draw on an earlier source. Verification applies that examination to a decision in front of us. We trace a claim to the document or account offered in support, then decide whether the evidence is sufficient for the use we plan to make of it.

Verification still requires close reading, of course. We cannot compare a news article with a study unless we can follow the language in both documents. Nor can we assess a chatbot answer, or recognize that a scientific report was paid for, unless we can separate its claims from the sources offered in support. Reading carefully brings the claim into view; verification asks what holds it up.

How we verify claims

A false claim can be written well enough that nothing in the wording gives it away. So verification reaches past the words to what the author of a claim doesn’t control, which is much harder to arrange than a confident sentence or statistic. In practice that comes down to a handful of checks: the kind of document a claim arrived in, who produced it and why, the record of where its facts began, and what people with no stake in it report. Two more tactics turn away from the claim itself: toward the channel that carried it and our own readiness to believe. Under scrutiny, an ordinary search touches on most of these moves, which makes the search itself the first thing to get right.

Search with intent

Verifying a claim often sends us out to find its sources and the evidence that argues the other way. That search turns on the question we put in the search box. A question in the claim’s own words tends to return the claim’s own sources, and the pages that already agree. But a question aimed at what the claim leaves out returns something we can weigh it against.

Search results aren’t ranked in order of accuracy. Search engines rank pages by many factors, including the words in the search, links from other pages, the date of a page, and the user’s location and personal search history. A result at the top of the page may be a strong one, but its position shows nothing about whether the page is verified or independent. Searching for a name, or for a claim in its own words, mostly returns what the subject says about itself. Adding a word that pulls in outside accounts, such as “criticism” or “dispute,” turns up what the first search left out.

Recognize document types

Recognizing what kind of document has arrived changes how we read what it says. An advertisement can resemble an article. An opinion can resemble reporting. A press release can reproduce the language of research. A paid testimonial can look like an ordinary review. The format doesn’t determine whether every claim is accurate; it tells us what the page was made to do and what the author may have chosen to emphasize.

Identify interests

We can also ask who stands behind the claim. A named researcher sharing evidence from her own field has a different relationship to a study than an anonymous account repeating a summary or an influencer promoting products based on select findings in that study. A company can provide current details about its products, although it also has an interest in presenting those products favorably. The test is whether that source has direct access to the information it presents, and whether another interest is at work.

Interests usually show up in who paid for a study or who owns the outlet that published a claim. A source with a stake in the outcome isn’t automatically wrong. But the stake tells us where to look for additional accounts: people or organizations with a different interest in the same question.

Trace provenance

We can follow the claim back to where it began. Every fact has a provenance, a record of where it came from and how it reached us. Opening the study, report, transcript, database, or public filing can change how we understand the account we first encountered. The original document shows what information was available before the claim moved through later versions, which may have filtered or warped it to suit the interests of the people using it.

Read laterally

Some claims call for leaving the original page. An unfamiliar organization may describe itself as an independent research group. Searching the organization’s name can show who runs it and where its funding comes from. Reporting from outside the organization can describe its work in ways the organization’s own site doesn’t. Researchers call this practice lateral reading because the reader leaves the page and looks for information elsewhere.

A Stanford study compared professional fact-checkers with historians and undergraduates as they assessed unfamiliar websites. The fact-checkers often opened new tabs after a brief look, then searched for information about the group that ran the site. Historians and students spent more time on the sites themselves, often focusing on how professional a logo looked or how reputable a web address seemed. But the fact-checkers reached more accurate conclusions about the sites by looking beyond the sites’ own descriptions. Here, it’s as simple as opening a new tab to answer questions that the original page can’t answer about itself.

Question what feels right

Personal experience affects how information feels when it reaches us. A statement may seem true because it fits something we’ve seen, heard, hoped for, or worried about. Familiarity can make a claim easier to accept before its source is verified, or even visible. Looking for the document behind the claim gives us something more concrete than the vague feeling that the claim sounds right.

Account for platform effects

The platforms and feeds that deliver much of what reaches us have a logic of their own. A social-media platform ranks posts through a system that responds to the platform’s goals and to a user’s activity. Search results can also vary by location, language, timing, and settings. Two people searching the same subject can encounter different pages. Changing the search terms or consulting another platform can show information that the first set of results didn’t include.

We can’t run every one of these moves on everything we read (who has the time?), so the amount of verification required can match the decision in front of us. A figure that will guide a medical decision, a lease, a purchase, a vote, or a line in our own work earns a closer look. But a passing detail at dinner may not warrant such investigation. And some claims are just too thin, too insignificant, or too irrelevant to justify an extended attempt to confirm them.

No single move works for every claim. A search can produce a reassuring answer without testing the part of the claim that matters most. Verification means noticing what still needs to be checked and choosing the approach that can answer it. Each time a claim reaches us, we decide where to begin. Making that choice well is the skill.

A place to practice

Reading about these choices and actually making them are different experiences. The path often looks clear after someone has exposed an error or explained a source. But it’s less clear when a claim first appears and several responses seem reasonable. So I created an app that lets us practice making the best verification moves: Study Hall. Here, Card Catalog members try their hand in real-life situations that require verification of some kind, one card at a time. Each card sets down something we may be about to believe, buy, share, sign, cite, or act on. Every card asks what we would do next, then explains what that next move can reveal.

Study Hall doesn’t ask us to decide whether a claim is true or false. It asks what information or action belongs between a claim and the decision that follows from it. A card about a study may lead us to the research paper. A card about an unfamiliar organization may lead us to reporting from outside the group. A card about a passing rumor may end with a choice not to repeat it. The card bank keeps growing, so the situations that require these information literacy skills continue to change and offer new opportunities to practice.

The design takes inspiration from Ruben Hassid’s cognitive gym, which gives readers exercises in different forms of thought. Study Hall was created to focus explicitly on the path between trusting a claim and the material that supports it. The explanation after each choice shows what the next step can tell us. Over time, the cards make the questions behind verification easier to recognize when a claim first reaches us.

Claims reach us already shaped by the systems that carry them and the people who have handled them along the way. Their presentation can make them feel complete before we have seen the evidence, the interests, or the gaps that would change how we use them. Study Hall gives us a place to stop at that point, while the decision is still open, and consider what deserves a closer look.

I believe that, for information literacy to be useful, it has to be practical. It should help us understand the information that meets us in daily life, not only the sources we encounter in classrooms or in the news. Study Hall was created to make those skills more familiar through practice.

The more often we practice choosing the best verification move, the more readily we can recognize what a claim requires before we give it our confidence.

You might also like:

• Building Your Own Verification Toolkit: A guide to the tools and databases librarians use, plus a personal fact-checking workflow you can run in under 60 seconds.

• The Verification Ladder: A Framework for Evaluating Video When You Can’t Tell What’s Real: “Is this real?” isn’t a straightforward question anymore. What to ask instead, and why the source often matters more than the footage itself.

• What “Do Your Own Research” Actually Means: A step-by-step methodology for evaluating any claim and exploring any subject.

Share Card Catalog

Join us in Study Hall

Study Hall is a game inside Card Catalog where paid members practice verification skills, one situation at a time. Each card puts something we might believe or act on in front of us, with realistic moves to choose between. After we choose, the card identifies the strongest move and explains why it’s the best one to make. For members, the link and rotating password stay pinned in Circle and sit just below.

A sample card: a real-life situation to evaluate, with three options to consider.

From here, it’s for paid members. Thank you for supporting the work.

Read more

This week:

• The August numbers are in, and they’re a problem for the two AI chief executives who told us half our jobs would be gone by now.

• The most-read thing in AI this week is a resignation letter. But it was one phrase that left the biggest impact, lifted from the colleagues its author left behind.

• A drug that AI designed for a deadly lung disease worked. Then its makers looked at something in the same patients that they hadn’t set out to measure at all.

Has the A.I. Job Apocalypse Been Postponed?

What happened

Economics columnist John Cassidy measured two A.I. executives’ predictions against the payroll data. Nearly three years ago, Sam Altman of OpenAI, the company behind ChatGPT, told a Wall Street Journal conference that A.I. could wipe out up to half of all existing jobs. He called that prospect progress. Dario Amodei of Anthropic, which makes the Claude chatbot, went further last May. He warned that A.I. could erase up to half of all entry-level positions and push unemployment into the double digits by 2030. On a podcast last month, Altman walked his own timeline back. The transformation is slower than he expected, he said, because “the economy just has so much inertia.”

The August employment report, out last Friday, told a different story. Employers added 162,000 jobs, more than Wall Street had forecast, and unemployment held at 4.1 percent. Recent college graduates may be an exception. Their joblessness has climbed over the past four years and now runs above the overall rate, a rise that reaches young people without a degree too.

McKinsey, a consulting firm that surveys large companies worldwide, found 40 percent of them using A.I. agents this year, up from 27 percent last year. Adoption keeps climbing, yet far fewer companies report A.I.-related layoffs than predicted them a year ago. The number of people providing legal services in the U.S. has barely moved either, holding roughly where it stood a decade ago.

Why it matters

The reason this outranks a single monthly jobs report sits in a longer trend it barely disturbs. For a quarter century, a shrinking share of everything economies produce has gone to workers as pay, dropping from more than sixty percent to below fifty-three percent (the lowest on record). Some of that reflects accounting changes, but much of it is a slow handoff from wages to profits, and the fear among economists is that A.I. hands more to whoever owns the software rather than to the people it replaces. More than two hundred economists and A.I. researchers have signed a public letter urging governments and technology leaders to steer the technology toward complementing people.

What makes the shift hard to govern is that it need not show up as unemployment at all. The clearest early sign is not layoffs but price. Wall Street banks that spend heavily on outside lawyers have begun demanding lower bills, on the logic that A.I. has cut the hours a case takes, and a law firm protecting its margins responds by automating more and hiring fewer juniors. No one is fired in that sequence, yet the work and the pay both shrink all the same. A country can run near full employment while the ground underneath its workers keeps moving, which is why the headline numbers offer such cold comfort.

Not every job sits equally exposed. Some resist because they carry more than the title admits: a radiologist reads scans, and also reassures frightened patients, coordinates with surgeons, and navigates the hardest calls. Manual trades look largely untouched so far, and Anthropic’s own usage data shows that cleaners, carpenters, electricians, and plumbers barely touch Claude at all. The heaviest pressure falls on desk jobs built around producing documents and analysis, the kind of work software can now do most cheaply.

What it means for us

For most of us, this arrives without drama: a change in the work itself rather than a pink slip. The task that took a full day last year takes an afternoon now. A manager who used to bring on three junior people brings on one, and asks that one to do more with the tools. If our own week has started to look like that, we’re already inside the shift the economists are describing, whatever the national numbers say.

New workers may get hit first. Layoffs make the news; jobs a company decides not to create never do. So this kind of loss stays easy to miss while it happens. Our kids and the new graduates our teams would once have trained are walking toward a smaller number of openings, and the squeeze on them can show up well before it reaches the unemployment rate.

Bottom Line:

In almost any job, the safest work is the work a machine still can’t do. Software can draft a contract, but deciding whether to sign it takes judgment, and getting a client to pick up the phone takes trust. Employers track the output a machine can now match, like documents produced or hours billed, and rarely track the judgment and trust it can’t. That human work is our strongest ground. Building it is the surest way to stay valuable as the tools get better at everything else.

The AI Researcher Who Just Quit Anthropic Says It’s ‘Crunch Time for Humanity’

Read more

Quick summary:

Media is not a neutral stream of information; it is made by people and institutions whose priorities can influence what they publish. Tracing who is behind a source shows where those priorities may come from and gives us a clearer basis for deciding how much to rely on it. It turns the question from simply “Is this true?” to “Who made this, and what conditions may have influenced how it reached me?”

Key takeaways:

→ Ownership shows up in the finished work through decisions made long before publication: who gets hired, what gets funded, what gets killed, and what never gets commissioned at all. None of those decisions show in the article or the episode itself.

→ Who owns a newspaper, a broadcaster, a podcast network, or a creator's sponsor list is on the public record. Corporate filings, broadcast regulators, disclosure labels, and an outlet's own About page each cover a different kind of media.

→ Finding the owner is only the first step. What they have at stake, from political donations to the businesses their news division covers, can help identify the subjects that deserve closer attention.

→ Media arrives as a set of choices made by people with interests. Once we can see that, we can choose what to follow with a clearer sense of how it helps build our picture of the world.

In December 2015, the Las Vegas Review-Journal was sold for 140 million dollars in cash to a newly formed company called News + Media Capital Group. The price was roughly 38 million dollars above what the paper’s previous owner had paid nine months earlier for a larger group of publications, and the new owner’s identity was withheld from the public and from the newsroom. Editors were told the new owners had chosen to remain anonymous. Reporters at the paper turned their investigative skills inward, tracing corporate filings across state records until they arrived at a name their own bosses hadn’t given them: Sheldon Adelson, the casino magnate and major Republican donor whose interests the Review-Journal covered on a daily basis.

Within days, the Adelson family confirmed the purchase. The editor who ran the investigation accepted a buyout the same week, and the three reporters who had done the unmasking had all left within six months. A former managing editor who tracked the paper afterward said coverage of Adelson’s interests grew “skewed through omission or editing,” with reporters learning which stories wouldn’t clear the top editors.

Adelson was a fixture in the paper’s reporting: he controlled Nevada’s largest casino operation, was involved in ongoing litigation in Nevada courts, opposed online gaming that would compete with his properties, gave tens of millions to Republican political causes, and owned Israel Hayom, an Israeli newspaper widely seen as supportive of Prime Minister Benjamin Netanyahu. When the reporters identified Adelson as the buyer, the paper’s readers could see that its new owner had direct stakes in many of the subjects it covered. In the months that followed, coverage involving those interests was sometimes cut, constrained, or not published.

The Review-Journal case exposed what runs through every media organization: ownership making the decisions that determine what audiences get to see and read.

How ownership shapes what gets made

The story that ran through that newsroom is the same story that runs through every piece of media we encounter. The media doesn’t simply report on a world that exists independently; it constructs a substantial portion of the world we perceive, including the parts we haven’t seen directly, the people we’ve never met, the places we’ve never visited, and the ideas that circulate around us. Our sense of a distant country is built largely from media accounts of it. Much of what we think we know about what a profession is like, what a religion involves, what a public figure believes, or what an unfamiliar community values comes from what the media has given us.

Media constitutes most of the reality we inhabit for anything outside our direct experience. A story that gets told consistently stops feeling like a story and starts feeling like a fact. When an issue gets framed differently, our sense of the issue shifts to match. Subjects that get sustained coverage feel urgent to us; subjects that get none feel like they don’t exist at all. This is how what gets made becomes what feels normal, what feels possible, what feels important, and how we feel about everything we haven’t seen for ourselves. We rely on that sense of the world to decide how to vote, what to buy, how to look after our health, and what to do with our money. When it’s been steered by interests we can’t see, those interests influence our decisions too.

Nothing reaches us by accident. Every piece of media we consume has passed through a chain of decisions made by people and organizations, and those decisions ultimately trace back to ownership. That ownership is rarely a single person with a single agenda. Corporate boards, hedge fund partners, family trusts, editorial leadership, and major funders (advertisers, subscribers, state sources, foundations) drive media output through interlocking interests. What gets made is decided by the people and institutions that fund the making, the leaders those funders hire, and the staff those leaders hire in turn.

The reader’s task is to trace who’s behind an outlet and what they want, going beyond the name at the top. What arrives as an article, a podcast episode, a documentary, or a magazine feature has been shaped by decisions and arrangements that usually stay invisible in the finished work. But we only ever see the finished work, not the decisions—and interests—that shaped it.

The decisions we don’t see

Before anything gets made, someone has to choose who does the work. Owners and boards select the leaders, and those leaders select the staff: editors, showrunners, producers, writers, and commissioning executives across every kind of media organization. Over time, the perspectives that survive are the ones that align with what leadership finds valuable, and staff whose work doesn’t fit the direction get sidelined or leave. The process usually takes years, but ownership changes can compress it into weeks. The three Review-Journal reporters who unmasked Adelson were all gone within months, along with the editor who ran their investigation. At the Chicago Tribune in 2021, dozens of journalists accepted buyouts within weeks of Alden Global Capital’s takeover of Tribune Publishing, and the paper’s editor-in-chief resigned three months later. No single departure has to look like intervention for the cumulative effect to reshape the newsroom.

Beyond hiring, organizational capacity determines what work can happen at all. Local government coverage needs reporters at city council meetings, and documentary teams need production budgets to fund time embedded with subjects. Investigative work in any medium takes sustained time on one subject, and the people doing it need their organization’s support the whole time. When an owner cuts the budget for that work, the reporters and producers leave and the work stops. We lose that work whether we notice or not, and a reader who does spot one missing story has no way of knowing what else is gone.

Even when the work can happen, editorial direction determines what reaches us. The Review-Journal’s former managing editor described reporters learning “what won’t fly past top editors” after Adelson took over, a form of self-censorship that operates across all content organizations before anyone has to intervene overtly. Writers, showrunners, producers, and editors know, often without being told directly, which subjects will draw executive review and which pitches will get killed before they get made. In October 2024, the editorial boards of both The Washington Post and the Los Angeles Times had prepared endorsements of Kamala Harris for president that their respective billionaire owners, Jeff Bezos and Patrick Soon-Shiong, prevented from running in the final weeks of the campaign. Multiple editorial board members at each paper resigned in response, and the event demonstrated in a matter of days that ownership retains the authority to override the work of the people it hires.

What never gets made affects coverage as much as what does. The absence of content is often more telling than its presence, and it’s harder to see because we can’t easily notice the piece that never got made. Papers owned by hedge funds cover local government less. Broadcasters owned by political donors can be less willing to investigate the causes their owners support, and streaming services owned by entertainment conglomerates can be reluctant to develop projects that criticize related businesses. Content that would embarrass a parent company or its major advertisers often fails to get made in the first place, and this shows up across documentary, magazine, podcast, and television production. For anyone who consumes media only from one source, these absences add up to a specific picture of the world, and that picture reflects the ownership arrangement whether or not anyone inside the organization is aware of what’s missing.

Reading the finished work more carefully won’t reveal any of this. To see these decisions, we have to look at who’s behind the outlet. That differs from one kind of media to the next, and so do the ways to find out.

Newspapers and magazines

Print journalism is where the pattern of ownership consolidation has run furthest. Two kinds of owners dominate: corporate conglomerates that operate news as one product line among many, and hedge funds and private equity firms that treat local papers as financial assets to be optimized. Alden Global Capital is the clearest example of the second kind. Alden took over the Denver Post’s parent company in 2010 and cut the newsroom in successive rounds over the following years, moving remaining staff from the paper’s downtown building to a printing plant in a neighboring county. By the time of the 2018 staff rebellion (in which remaining Denver Post journalists published a special editorial section calling for Alden to sell the paper if it wouldn’t fund actual journalism), the newsroom had shrunk from roughly 200 journalists at its peak to about 60. Denver readers lost coverage that a fully-staffed newsroom had provided: local government reporting, court coverage, and investigative work that had defined the paper for decades. Northwestern University’s Local News Initiative documents this pattern across acquired papers, showing that measurable cuts to newsroom capacity correspond to measurable declines in coverage of local government, civic institutions, and courts.

Similar concentration governs newspaper markets across most democracies. In France, the Bolloré Group holds stakes of roughly 30 percent in Canal+, Havas, and Louis Hachette Group, the companies separated from Vivendi in December 2024, as well as a separate stake in Vivendi itself. Through these linked holdings, the Bolloré group controls or influences CNews, Europe 1 radio, Prisma Media magazines, and the Journal du Dimanche. Critics and press-freedom groups have linked Bolloré’s expanding media influence to a rightward shift at several of these outlets. In the UK, three companies, News UK, DMG Media, and Reach, together control about 90 percent of national newspaper circulation, according to the Media Reform Coalition. The corporate structures vary by country, but concentration has the same effect everywhere. When fewer owners control more papers, a single owner’s editorial decision reaches more readers, and two papers that look like competitors may share the same owner and the same blind spots.

Finding out who owns a newspaper is a matter of tracing a public paper trail:

• The masthead and the paper’s own About page may name the parent company.

• Wikipedia’s article on the parent company can help map the full portfolio and summarize ownership history, but it should be checked against primary sources.

• SEC filings provide authoritative disclosures about ownership and corporate structure for U.S. publicly traded publishers. The most useful documents are often annual reports, proxy statements, merger filings, and Schedules 13D or 13G, which disclose beneficial owners holding more than 5 percent of a registered class of voting shares.

• Free Press covers U.S. media consolidation and maps major media ownership.

• The Media Reform Coalition tracks media ownership in the UK.

• Reporters Without Borders’ Media Ownership Monitor maps media ownership and control in selected countries, particularly useful where domestic disclosure is limited. Its country projects identify owners, shareholders, and people with influence, then place those relationships alongside market concentration, political affiliations, other interests, and national disclosure rules.

Broadcast and cable news

Broadcast television and cable news operate under different rules than print. Structures vary by country, and the ownership picture depends on both the commercial landscape and the regulatory framework. The best-known case of an owner’s values driving coverage is Rupert Murdoch, the Australian-born media proprietor whose family trust holds voting control over Fox Corporation (Fox News, Fox Business) and News Corp (The Wall Street Journal, New York Post, The Times of London, The Sun, HarperCollins). Murdoch’s political conservatism has been a matter of public record for decades, and Fox News has maintained a consistently conservative editorial direction since the network’s 1996 founding. The 2023 Dominion Voting Systems defamation settlement, in which Fox News paid 787.5 million dollars, put the mechanism on the public record: internal communications showed that hosts and executives had privately doubted or rejected false claims about the 2020 election while continuing to air them. A 2025 family trust settlement gave Lachlan Murdoch sole voting control over the family’s stakes in Fox and News Corp through 2050, ensuring that the companies would remain under his direction after Rupert Murdoch’s death.

The rest of the U.S. broadcast landscape is divided across a handful of parent companies: Disney, Comcast, Warner Bros. Discovery, and Paramount, a Skydance Corporation, the entity created by the August 2025 merger of Skydance Media and Paramount Global. Paramount Skydance has proposed to acquire Warner Bros. Discovery, a transaction that would further concentrate control over major news, film, television, and streaming properties. Each company has undergone significant restructuring in the past year. Comcast, for example, has announced plans to separate much of its cable-network business from its broadband and technology operations. These companies also run businesses such as cable services, film studios, theme parks, and streaming platforms, and their news divisions cover those industries as part of companies with a stake in how they are covered. Mergers and spinoffs will keep changing which company owns what, and each change puts a different set of business interests behind the news.

Other countries structure broadcasting differently, often around public service broadcasters funded through household fees or general taxation. The BBC in the UK and ARD/ZDF in Germany operate under this model, as do Canada’s CBC/Radio-Canada, Australia’s ABC, Japan’s NHK, and their counterparts elsewhere, with varying degrees of editorial independence from their governments. The specific governance structures differ, but in each case the funding source is easy to find on the outlet’s About page or in the relevant national regulator’s public records.

Finding out who owns a broadcast station is public information in most countries:

• U.S. stations: FCC public inspection files can be searched by call letters and include station ownership reports. The FCC’s ownership-report search provides a more direct route to filings by call sign, facility ID, location, or licensee.

• Other countries: National broadcast regulators, including Ofcom in the UK, the CRTC in Canada, ACMA in Australia, Arcom in France, and ANATEL in Brazil, maintain public registers of licensees, though the detail and searchability vary.

• Publicly traded parent companies: 10-K filings on the SEC’s EDGAR database help trace the parent-company structure, major subsidiaries, board, principal shareholders, and material business risks.

State-controlled media

State involvement in media extends beyond public-service broadcasting. Some international outlets are funded or governed more directly by national governments, making their relationship with the state more explicit than it is at many commercial or public-service organizations. RT in Russia and CGTN in China are prominent examples, while Press TV in Iran and TRT in Turkey operate under different political and legal systems that should be assessed individually rather than treated as identical. Funding and governance do not determine that every report will be inaccurate or politically directed, but they are important context when an outlet covers its government’s foreign policy, domestic critics, geopolitical rivals, or contested international events. Knowing who funds and oversees an outlet helps readers place its reporting alongside independent sources and sources with different institutional interests, treating it as one voice within a broader picture.

Identifying a state-controlled outlet often starts on the outlet’s own About page:

• Self-identification: RT identifies itself as funded by the Russian state. CGTN operates within China’s state broadcasting system, under the Chinese Communist Party’s Central Publicity Department. Al Jazeera acknowledges that it is funded by the state of Qatar.

• U.S. registration: The Foreign Agents Registration Act database at fara.gov can show whether a U.S.-based outlet or its representative has registered as an agent of a foreign principal. The law requires registration and disclosure from people or entities carrying out specified activities in the United States on behalf of foreign principals; the Department of Justice may require registration in particular cases.

• Global tracking: Reporters Without Borders’ Media Ownership Monitor covers media ownership and control across selected jurisdictions, including countries where domestic ownership disclosure is limited.

Podcasts

Podcasts are a relatively young medium, but much of their large-scale production and distribution is now controlled by major networks and technology companies. Edison Research’s first-quarter 2026 rankings placed Spotify first among U.S. podcast networks by average weekly reach, followed by the SiriusXM Podcast Network and iHeartPodcasts. SiriusXM and iHeartMedia were also reported to be in preliminary merger discussions in April 2026, although no deal was guaranteed. Amazon acquired Wondery in 2020 and later reorganized it: in 2025, it moved Wondery’s narrative podcasts and Wondery+ service into Audible, while placing creator-led shows in a new Creator Services unit. Newspaper companies have moved in as well: The New York Times bought Serial Productions in 2020. Alongside these large companies are smaller talent-led networks and independent cooperatives. Outside the United States, the BBC's Sounds platform plays a major role in British podcasting, while Sweden-based Acast operates across several markets.

A platform can influence what listeners are able to hear. When Joe Rogan signed an exclusive licensing deal with Spotify in 2020, dozens of episodes from his back catalog were not made available on the platform, including interviews with controversial figures such as Alex Jones, Milo Yiannopoulos, and Gavin McInnes. Platforms also shape what happens after a show is released: their recommendation algorithms affect which listeners discover a podcast and, as a result, which advertisers may want to sponsor it.

Sponsors are another form of influence. A podcast may claim to be independent or critical of a particular industry while accepting money from companies or trade groups connected to that industry. The sponsor may not directly control the show's content, but the financial relationship can make listeners question whether the host is truly free to criticize the people or industry funding the podcast.

To find out who is behind a podcast, look at both the show itself and the platform or network that distributes it:

• Check the podcast’s website, which will often identify the network, production company, or parent organization.

• Look at the publisher information in the show’s RSS feed, which can identify the organization responsible for publishing it.

• Review a season’s sponsors rather than relying on a single episode’s advertisements, since recurring sponsors can reveal longer-term commercial relationships.

• Research the parent company: Spotify, SiriusXM, iHeartMedia, Amazon, and The New York Times Company are publicly traded or public-reporting companies, so their SEC filings can identify subsidiaries, directors, executives, and major shareholders. Acast, which is listed in Sweden, publishes comparable company reports through its market disclosures.

• Use Wikipedia as a starting point for mapping a company’s brands and subsidiaries, then confirm important ownership claims through company filings, investor-relations pages, or official corporate announcements.

Influencers and content creators

Individual creators face the same sponsor-alignment questions as podcasters, but they also have to follow advertising-disclosure rules. In most countries, creators must tell audiences when they have a material relationship with a brand, such as being paid, receiving free products, or earning commission from a promotion. The UK’s Advertising Standards Authority, the EU’s Digital Services Act, India’s Advertising Standards Council, Australia’s AANA, and the U.S. Federal Trade Commission all set standards for disclosing paid endorsements. Rules and penalties vary by country, but the basic expectation is the same: viewers should be able to tell when content is advertising. In the United States, the FTC’s 2023 endorsement guidance states that tagging a brand can count as an endorsement and that “#ad” is not enough if it is buried in a long list of hashtags or placed where viewers are unlikely to see it. Disclosures must be clear, noticeable, and placed alongside the endorsement; a platform’s built-in “paid partnership” label may help, but it does not automatically meet the FTC’s requirements.

Undisclosed sponsorship can have consequences beyond a fine. In 2022, Kim Kardashian settled with U.S. securities regulators after promoting the cryptocurrency EthereumMax without revealing that she had been paid $250,000. The token later collapsed, and people who bought it after seeing her endorsement lost money. Regulators in other countries have also ruled against creators who failed to disclose paid relationships. In areas where recommendations can shape important decisions—such as medical or financial advice, or expensive product purchases—knowing who pays a creator matters even more. Looking at a creator's sponsors over time can reveal whether the money is affecting what they recommend.

Finding out who’s paying a creator is mostly a matter of reading what they’re required to disclose:

• Disclosure labels on individual posts, such as “#ad,” “paid partnership,” or “sponsored,” identify the brand behind that specific post.

• A season’s worth of posts, read together, shows which sponsors recur. A brand that appears once may be an ad; a brand that appears every month may signal an ongoing relationship.

• The creator’s bio or About page often lists management, agency, or brand partnerships directly.

• Regulatory rulings are public. The UK’s ASA publishes decisions involving named creators, India’s ASCI publishes complaint outcomes, and the U.S. FTC publishes enforcement actions. A creator’s name in one of these databases documents an instance in which a regulator found a disclosure problem.

• For newsletters, the writer’s About page usually explains the funding model, and the newsletter archive shows whether sponsors appear and how often.

What owners have at stake

Finding out who owns an outlet is only the first step. A hedge fund, a family trust, a government, or a sponsor can each have different reasons to care about what gets published. The next question is what those owners have at stake: the political, financial, or personal interests that could make certain subjects especially important to them. Those are the areas to watch, and the trail to them looks different depending on the type of owner involved.

Individual owners and family trusts (such as Adelson at the Las Vegas Review-Journal, Bezos at The Washington Post, and the Murdoch family at Fox and News Corp) leave personal paper trails:

• Political donation records (OpenSecrets in the US, the Electoral Commission in the UK, the Australian Electoral Commission, Elections Canada) show what causes an owner has funded.

• Wikipedia biographies gather reporting on public statements, business interests, philanthropic causes, and past controversies.

• Long-form profiles from outlets such as The New York Times, ProPublica, The New Yorker, Columbia Journalism Review, Nieman Lab, and Free Press can go deeper into the conflicts an owner has entered and any editorial patterns connected to their ownership. A long record of donations, statements, and business activity can show an owner’s priorities more clearly than a single interview.

Institutional owners leave a different kind of trail, focused less on any one person’s beliefs and more on how the organization is structured, funded, and governed:

• A hedge fund, such as Alden Global Capital, may reveal its interests through its investment history, cost-cutting record, and treatment of earlier acquisitions.

• A public trust, such as the Scott Trust which owns The Guardian, operates under a founding charter that establishes its rules and constraints.

• A state broadcaster, such as the BBC or RT, has funding sources and governance rules that help determine how much editorial independence it has in practice. Long-form reporting about the organization can fill in what individual biographies cannot.

Choosing sources by who’s behind them

Deciding which sources to follow takes two kinds of knowledge. The first is everything above: who is behind an outlet and what they have at stake. Once we know that, we can look for patterns: what gets covered aggressively, what gets softened, which subjects appear and disappear, whose voices anchor the reporting, and whose voices are absent. The pattern emerges across weeks or months of coverage, not from a single piece.

The second kind of knowledge is about ourselves. Tracing whose interests shape a piece of media can tell us something about how it was made and by whom, but it cannot tell us exactly how to receive what we have read, watched, or heard. That depends on what we bring to the encounter: our sense of what matters, what we consider fundamental, what we believe we ourselves are responsible for, and what we would refuse to accept from any source. Without a clear sense of our own values, it’s harder to recognize when a piece of media invites us to adopt values that aren’t ours.

Once we understand that everything we consume was selected, produced, or distributed by people and organizations with interests, we can read with that fact in mind (even when we haven’t investigated a particular source in detail). That habit makes it easier to notice when a source may be trying to shape how we see an issue. More detailed research then comes into play when we’re deciding whether to follow a new source or considering a claim we may act on.

What this research gives us is greater control over interpretation. A reader who knows who owns an outlet can better evaluate its coverage of subjects connected to that owner’s interests. Someone who has traced a broadcaster’s parent company can decide how much weight to give its reporting on national politics. A person who has reviewed a creator’s sponsor history can take those relationships into account before acting on a recommendation. Without this context, decisions built on media sources, from voting and purchases to medical and financial choices, may rely on information shaped by interests we haven’t examined.

Once we can see the money and power behind what we consume, media becomes easier to recognize for what it is: a set of choices made by people and institutions with interests about what to show us. We can then decide, source by source, whether those interests are ones we would support ourselves. The picture of the world we carry is built from the sources we choose to follow; choosing them with a clear view of who is behind them gives us a greater say in what that picture becomes.

You might also like:

• 20 Small Acts of Media Literacy We Can Practice Every Day: From checking the date to tracing the source: building a sharper eye for information.

• Misinformation Spreads at the Speed of Impatience: Why critical thinking starts with slowing down, according to the research.

• The Flood is the Strategy: Defending against weaponized information overload.

The free essays are the foundation. The paid tier is where information literacy becomes a skill you can use: the Card Catalog Classroom with its searchable Reference Collection and structured lessons across four learning domains, plus biweekly AI Briefings, monthly Librarian Hotlines, and quarterly deep-dive guides.

Upgrade to paid for the full Card Catalog. Thank you for being here!

Subscribe now

Share Card Catalog

Quick summary:

Podcasts hold hours of experts explaining their own work, and once an episode ends, finding a particular claim again means scrubbing through the audio. Platform transcripts and a handful of free tools can turn any episode into text fit for research, with the studies behind the conversation a click away. Simple checks then sort the rigorous shows from the rest and trace a spoken claim back to its study.

Key takeaways:

→ Podcast claims travel by word of mouth more often than by citation. A remark heard once gets repeated far beyond the episode, and going back to confirm the wording or the evidence is rare.

→ The largest audio apps now produce transcripts on their own, and paste-in websites handle the shows they miss. With text in hand, a timestamp pins a spoken claim the way a page number pins a printed one.

→ The most rigorous shows publish the studies and books behind each episode right in the episode description. Those lists separate the shows built on evidence from the rest, and they let a listener read what the expert read.

→ The same approach works on lectures and oral histories as readily as on podcasts. Speech that becomes text becomes verifiable, as more of what people know arrives as speech every year.

Scroll to the end for a recommended podcast list.

In the summer of 1933, a Harvard classics professor named Milman Parry made his first trip to what was then Yugoslavia. He carried a question that had gone unanswered since antiquity: how could the Iliad and the Odyssey, poems far longer than anyone could memorize word for word, have been composed in an oral tradition? The mountain villages of the Balkans were still home to illiterate bards who could sing heroic epics for hours at a stretch. Parry believed their way of composing was the same one that had produced the Homeric poems; recording them meant watching the ancient method still at work.

He returned the following year with a former student named Albert Lord, and this time the recorder was a custom rig: two turntables that handed off to each other so one could keep cutting a disc while the other was swapped, letting a performance run for hours without a break. Over fifteen months the two of them gathered more than twelve thousand texts, most taken down in writing and many recorded onto aluminum discs, more than three and a half thousand of them in all. Alongside the discs were ninety-five notebooks of texts taken down by dictation. What began as spoken sound became a written archive, and that archive became the foundation for how scholars understand the oral-traditional composition of the Homeric poems.

For most of history, the only way to keep something said aloud beyond the moment of its utterance was for a person to write it down. Some speech survived that way, in court records and in sermons a listener copied out. Transcription by hand was slow work, though, so those were the exceptions, and ordinary talk left no record. What Parry grasped, before magnetic tape became commonplace or digital audio existed, was that a recording is only half of the value. The other half arrives when the recording becomes text a person can search and quote.

An archive disguised as entertainment

More than half a billion people around the world now listen to podcasts, across more than four million shows. Buried in that catalog is a kind of source material that barely existed a generation ago: experts talking about their own work, at length and in their own words. On one show a scientist spends an hour on the false starts behind a published result, the kind of detail a formal paper leaves out. Another has a historian naming the exact archives behind a claim, and elsewhere people who will never write a book sit for long interviews about what they know.

The candor is one part of the value, and the paper trail is another. A published article shows conclusions, and an interview lets us hear the same person explain how they reached them and where they still hedge. The most rigorous shows then post every study and book mentioned in the episode’s notes, so a single conversation can hand us a reading list. And because shows publish in dozens of languages, the catalog reaches experts whose work never gets translated into English at all.

Some of what’s in there exists nowhere else. Plenty of expertise never gets written down at all (for example, the tacit working knowledge of surgeons, engineers, negotiators, and builders) and a long interview can be the only record of how those people think. Timing favors the medium too, since a researcher can talk through a finding the week it publishes, months before any writing that digests it could arrive. And when an interviewer pushes past a rehearsed answer, we hear something print rarely captures: an expert responding in the moment.

The clearest proof that podcasts already work as sources is how often we quote them. “I heard on a podcast” has become a standard way for a claim to enter conversation, and a statistic from an interview gets repeated at a dinner table, then again in someone’s article. Each repetition runs on memory alone. An episode link exists to send, but it opens an hour of audio with no page number, and the claim sits somewhere inside, unmarked. So the archive is already in use as a source, just without a way back to the moment, and a claim nobody can return to is a claim nobody checks.

Newsrooms already treat podcasts as source material, since a remark a public figure makes on a show routinely becomes a headline, and enough demand exists that a small industry of transcript and search tools has grown up to serve that work. So the archive has two kinds of users: professionals who mine it with purpose-built tools, and the rest of us who quote it from memory. The gap between the two is awareness, since most of us never learn the material can be searched and cited at all.

Part of why we never learn is mechanical: a conventional search engine can’t reliably search what’s only been spoken. Google, for example, can find an episode when the topic made it into the title, the written description, or an indexed transcript, but everything else—the claim made at ninety minutes in, or the study a guest named in passing—may exist only in the audio, so no typical search box will surface those words. What a search engine sees of a podcast is often the label on the box; the contents inside stay sealed.

Librarians have a name for a similar issue. Grey literature is the term for knowledge produced outside the usual commercial or academic publishing channels: government reports, working papers, conference talks, and other material that may be valuable but poorly indexed. Podcasts don’t fit the category neatly, but millions of hours of recorded expert conversation can present the same practical problem. Reaching material like that takes method: knowing where it collects, and knowing how to pull it into a usable form. For podcasts the method runs in a fixed order, and here we’ll walk through it step by step: first turning the audio into text, then finding the right shows and episodes, then checking the claims and the people who make them, and finally citing what survives the scrutiny.

Plenty of transcripts already exist

Often the first step, audio into text, is already done for us. Several of the biggest podcast platforms now generate transcripts automatically for many episodes, and each keeps the text in a different place. Platforms may create these on their own, whether or not a show publishes a transcript itself. Here’s what each offers, with the steps for opening the text.

Spotify

Spotify can show the full text of an episode while it plays, the way a music app shows lyrics. The line being spoken is highlighted, and the text scrolls in time with the audio.

Important: We must use the official Spotify Desktop App installed on our computers. If we’re using a standard web browser, the transcript features won’t show up at all.

To find the transcript:

• Play the Episode: Open the desired podcast episode and click the green Play button.

• Open the Right Sidebar: Look at the far right edge of the screen next to the main window content. Click the arrow button facing left that says Show Now Playing view when hovering over it. This slides open a wide menu pane on the right side of the screen.

• Scroll the Sidebar Down: Move the mouse cursor into that newly opened right sidebar. Scroll down completely past the cover art, the episode description, and the Comments section.

• Locate the Read Along Panel: Look right below the About the podcast block. If a transcript is available, a section titled Read along or Transcript will appear displaying the scrolling text. Tapping any line of text automatically moves the audio playback to that exact moment.

Spotify hasn’t generated transcripts for every episode. When an episode has none, no transcript option appears, and the tools in the next section fill the gap.

Apple Podcasts

Apple generates transcripts as well, and the company says more than one hundred million episodes now carry them (across thirteen regional versions of languages).

Important: We must use the official Apple Podcasts App installed on Mac computers. While there’s a web version of Apple Podcasts, the interactive transcript feature is entirely unavailable when listening through a browser.

To find the transcript:

• Play the Episode: Open the desired podcast episode and click the Play button.

• Open the Full-Screen Player: Tap the MiniPlayer bar stretching across the bottom of the screen to slide open the full-screen playback view.

• Tap the Transcript Icon: Look at the bottom-left corner of the full-screen player screen. Tap the speech bubble icon that looks like a quotation bubble with horizontal lines inside.

  

• Navigate the Text: The transcript will open in full screen and automatically scroll in sync with the audio. Tapping any line of text automatically moves the audio playback to that exact moment.

Apple hasn’t generated transcripts for every episode. If we don’t see it, or if the icon is grayed out, that means that it’s not available. Same as for missing Spotify transcripts, there are options in the next section that can help in this kind of situation.

YouTube

Many podcasts also appear on YouTube as video. When a video has captions, YouTube provides an interactive transcript text block that’s useful for locating key terms.

To open the transcript:

• Expand the Description: Locate the gray description section box directly underneath the video player screen. Click the ...more button to expand the text box completely.

  

• Open the Transcript Window: Scroll all the way to the bottom of the items inside that description section. Click the button labeled Show transcript. A dedicated transcript window will immediately slide open on the right side of the video player screen.

• Search for Keywords: Move the cursor inside that right sidebar text block. Press Command + F on a Mac, or Control + F on Windows, to search for a specific name, phrase, or term.

• Navigate the Video: Click any timestamped line of text inside the scrolling layout to automatically move the video playback to that exact moment.

YouTube can also translate the captions displayed during playback:

• Open Player Settings: Hover over the video player screen and click the Settings gear icon located on the bottom toolbar.

• Select Translation: Click Subtitles/CC, click Auto-translate, and select the desired language from the list.

When we turn on auto-translate, the new language only shows up as subtitles on the video screen, while the transcript text box on the right side of the screen stays in the original language. This means we can’t search or copy the translated words from that sidebar. Computer translations also frequently make mistakes with names and lose subtle meanings. While this feature helps us understand the general conversation, we shouldn’t rely on it for exact quotes.

For transcripts that aren’t readily available

If an episode lacks a native transcript, no panel or button will show up on its page. We can easily fix this by generating a text file directly from the audio using an external tool. These tools come in two forms: websites where we paste the episode link, or software programs that run locally on our own computers.

Paste-the-link generators

These are websites that create the transcript for us. Paste in the link to an episode, and the site returns the full text, broken into timestamps.

• Podtyper works from a public Apple Podcasts, Spotify, or YouTube episode link.

• TranscribeTube works from a YouTube link or an uploaded audio or video file, and it can label who’s speaking (which helps on a show with several guests).

Unlike the read-along panels inside Spotify and Apple, these services return a file we can save and mark up.

Whisper, for private or bulk transcription

Whisper is an open-source speech-recognition model released by OpenAI. It can run on our own computers, so the audio stays local on the machine. The practical way to use it is through an app built around it, such as MacWhisper: open the app, drop in an audio file, and the transcript appears.

For long-running podcasts, high volume is the best reason for its use: it can process a show’s back catalog in batches. It also handles audio that was never published at all (like our own interview tapes or a family’s oral history). When we use a local transcription model, those files stay on our own devices. (And to clarify, it’s unrelated to Wispr, the dictation app with the similar name.)

Looking beyond Spotify, Apple, and YouTube

Those three platforms have the easiest transcripts to reach, though podcasts play in dozens of other apps as well, with many drawing on the same public feeds. Wherever we listen, a more difficult question could be sitting underneath: which shows should we be following at all? A separate set of sites exists for answering that.

Where to find shows in the first place

These sites help us browse and search the shows themselves, by topic and, in some cases, by guest.

• Podchaser catalogs shows and their guests, the way a film database catalogs movies and actors, so searching a researcher’s name can turn up shows they have appeared on.

• Goodpods surfaces shows that other listeners recommend.

• The New Books Network runs episode after episode of scholars interviewing scholars about their new academic books, sorted by field.

• Podcast Index is an open podcast directory and database that makes public podcast feeds searchable.

The charts inside the big platforms rank shows largely by popularity. But popularity can be built with marketing, and offer nothing about whether a show’s claims and sources hold up under scrutiny. These directories offer other ways in: topic, guest, category, description, ratings, reviews, and editorial or listener recommendations. That makes it easier to find a small, rigorous show that might otherwise be buried beneath a heavily promoted one.

How to search every show at once

A directory surfaces shows to follow, and sometimes the question is narrower: which episode, anywhere, discussed a topic or quoted a researcher. Often those words can make it into an episode’s title or description, and a regular search, or Listen Notes below, will turn them up. But when they were only ever spoken over the course of the episode, transcript search engines take over.

Listen Notes

Listen Notes is a search engine for podcasts, the way Google is a search engine for the web. It indexes millions of public podcasts using the feeds shows themselves publish. Type a phrase or a name and it returns matching shows and episodes, searched across titles, written descriptions, show notes, and, where available, transcripts. It works like a card catalog for podcasting at large, the first place to look when the question is which show covered a subject at all.

Tapesearch

Tapesearch searches podcast transcripts. It has generated text for a large catalog of episodes, and typing a term returns matching passages, each with a timestamp that takes us to the moment in the audio. Exact phrases can go inside quotation marks, and search terms can be combined. For research that starts from a term and needs to find where it came up, this is the tool.

Podscan

Podscan is Google Alerts for podcasts. Give it a name or a topic, and it searches its transcript database and monitors new episodes as they publish, flagging mentions whether or not they ever reach a title or description. The other tools here help search what already exists; Podscan also keeps watch going forward, which is how a researcher tracks a topic over time and how anyone finds out they’ve been mentioned on a show.

Snipd

Snipd is built around saving and searching the parts of episodes we care about. As we listen, it lets us clip a moment, and it saves that clip with a transcript, timestamp, and summary in a personal library we can search later. Over time this becomes a searchable record of the passages we marked across everything we’ve heard.

Show notes are the works cited

Everything so far ends with words on a screen, an episode found and its transcript in hand, and words alone only prove what someone said. Checking where the words came from (that is, the provenance of these ideas) starts in the same place on every episode: the show notes, or the episode description that the show itself wrote. On most shows it carries a summary and some promotional links. The most rigorous shows use it for something else entirely: a list of the studies and books the conversation touched, with links straight to the original sources.

This is where a podcast stops being testimony and starts becoming scholarship we can verify. An expert’s spoken claim asks for our trust, and a works-cited list lets us check before we give it. With this, we can open the same papers the expert read and see whether the claim survives the trip back to its source. Whenever an episode raises a claim we plan to repeat or rely on, having these sources available allows us to read what the speakers themselves were reading and come to our own critical conclusions.

Coverage varies from show to show, and that variation can tell us something about a show’s reporting and editorial practice. Science shows tied to institutions, and interview shows that take sourcing seriously, may publish full lists of studies and books. Peter Attia’s medical show often links studies and other resources discussed in an episode, and Ezra Klein’s interviews commonly include books mentioned in the conversation and the guest’s recommendations. To find the notes, open the episode’s page in any app and expand the description. When the in-app version looks cut short, the show’s own website may carry a fuller version under the episode’s title. An episode with nothing there beyond links to sponsors has answered a question too: how much of its source trail it’s made available.

How to verify a show

The show behind an episode deserves the same scrutiny as the episode itself. Podcasting has no universal editor, reviewer, or publisher standing between a host and our ears. Anyone can start a show, and sounding authoritative is a production choice, open to anyone with a microphone. So the question of who’s behind a show comes before anything the show says.

Three quick checks reveal a great deal about a podcast:

• Search the host’s name: Relevant training, an institutional affiliation, or a career in the subject may surface quickly, along with interviews, reporting, criticism, and other evidence of how the host is regarded.

• Open an episode’s notes: A rigorous show may link its sources there; a show with nothing but sponsor links leaves its evidentiary trail harder to inspect.

• Scan the guest list: Working researchers and practitioners, whose names and work can be independently confirmed, are a useful sign that a show draws on real expertise.

There’s another shortcut for the moments when a host or guest’s name is already in hand. We often arrive at a topic from somewhere: an article that quoted an expert, or a study with an author. Podchaser can turn that name into a list of shows the person has hosted or appeared on. One search opens a shelf of candidates, each tied to someone whose work sent us looking.

Two deeper marks distinguish the strongest shows: a named institution or a stated corrections practice standing behind the work, and openness about the show’s own point of view on controversial ground. Both tend to show up on a show’s About page, as one more place to look. And production polish counts for nothing on its own: a show can be popular and beautifully produced and still cherry-pick the studies that fit its story.

How to verify what we hear

A vetted show narrows the risk, but each claim still calls for its own check because speech and print run on different standards. In conversation, an expert may speculate and simplify in ways they would rarely put in print (since talk often goes out with no peer review and less editorial scrutiny). A remark on a podcast reads best as a lead: dependable evidence of what the expert said and a pointer to published work, where a claim can be checked before we repeat it as true.

Suppose an episode claims that 80% of adults with ADHD dread replying to text messages. Before that number travels further, run it through the same checks a statistic in an article would get:

• Open the show notes and look for the study behind the number. If it’s linked, read what was measured and in how many people, since a statistic often changes shape between the paper and the retelling.

• If the notes name no source, search the claim itself: the key terms and the number, in Google Scholar or a regular search, to see whether any study produces that figure.

• Check the speaker too. A search on the expert’s name and field can turn up their training and what they have published, which helps distinguish a researcher describing their own data from someone repeating something heard elsewhere.

• A claim that traces to no study and no named data is an opinion, a hypothesis, or an unsupported assertion. Repeat it as such, or drop it.

Two cautions apply to the transcript itself. Automatic transcription mishears exactly what research leans on—proper names and figures—so before quoting anything word for word, open the audio at the timestamp and confirm the words by listening. And an contentious clip calls for the full episode, since audio can be cut in ways that reverse a speaker’s meaning, and synthetic voices are now convincing enough that a clean recording is no longer proof a person said the words. Finding the moment inside the full, original episode answers both concerns.

A long interview also offers something no pulled quote can carry. Across an hour we hear how an expert reasons, which evidence moves them and where they still hesitate. Lift one sentence out and that context drops away, which is how a hedged, qualified statement turns into a confident-sounding claim it never was. So the transcript earns its keep twice: once as a source of quotes and once as something to read in full, for the reasoning around them.

The form itself also works on us in a way print simply can’t. Hours of a voice in our ears can build the feeling of knowing the speaker; we may extend to familiar voices trust we would withhold from an unfamiliar claim on the page. A claim from a familiar voice can slip past the scrutiny we would give the same words from a stranger. The fix here is running the claim through the same checks anyway, warm familiar voice or not.

And one more bias sits underneath the medium itself. The experts who show up on podcasts are a self-selected group: the ones willing to sit for an interview and are at ease behind a microphone. Researchers who never record anything may vastly outnumber those who do, so the voices that reach us are one slice of a field, not necessarily a representative one. That gap is easy to forget when a well-argued episode starts to sound like the consensus of an entire discipline.

How to cite a podcast

After the episode is found and the claims are checked, one question is left for anyone writing a paper or a report: whether a podcast can go in the citations. It can! Citation systems have long covered recorded speech, and the style guides that govern academic writing include formats for podcast episodes. A podcast is fully legitimate as a primary source when it’s cited as evidence of what a person said, much as an interview or a speech is cited.

For a claim, though, the episode is often a conversation about the evidence. The study a guest discussed is the citation for the finding, and the episode is the trail that led there. The rest of us might not need to keep citation lists, but the habit still pays off in a smaller way: sending the timestamp and a link along with a claim lets the other person hear the moment for themselves.

A workable general citation names the creator or host, the episode title in quotation marks, the show title in italics, the publisher, the publication date, and a link to the episode. Add a timestamp when you are quoting or directing a reader to a particular claim. Include a retrieval date when a required citation style calls for one, or when the episode is unusually likely to change or disappear.

• Pattern: Host or creator, “Episode Title,” Show Name, publisher, publication date, URL; timestamp for the quoted passage.

• Example: Nature Podcast, “A black hole’s spin could finally be measured by a high-speed star,” Nature Podcast, Nature, 19 August 2026, [episode link]; 00:46.

Anyone who reads that citation can open the episode, move to 00:46, and hear the claim being made, much as a page number sends a reader to a passage in a book.

Listening like a researcher

With a transcript open and the sources in hand, an episode becomes something we can return to and question. A claim that catches our attention in minute forty no longer has to survive as a half-remembered remark. We can find the line again and follow where it points. The standards we already use on writing can now apply to what we hear.

A good podcast can do more than deliver an answer. It can lead us to a researcher’s work, a book, a paper, an archive, or a disagreement that the conversation only had time to touch. An interview may be the first place we encounter a question, not the final authority on it. Once the words are searchable, the episode becomes a way into a larger body of work. That possibility changes the listener’s role. We can admire an argument and still look at its evidence. We can take an expert seriously without accepting a spoken claim before we’ve seen what supports it. When the source holds up, we know where it came from. When it doesn’t, the search has still shown us something about the limits of the claim.

The tools that make this possible once belonged largely to newsrooms and research institutions. Transcripts, searchable audio, and episode notes have made more of that work available to anyone willing to use them. A listener who opens the notes and follows a source also changes the incentives around the medium. Shows have a reason to document their claims when listeners expect a trail they can inspect.

Recorded talk has always carried knowledge that doesn’t appear elsewhere: the texture of an interview, the hesitation before a conclusion, the working memory of a field, and the account of someone who may never write a book. What’s changed is our ability to return to it. What lectures, oral histories, conference sessions, interviews, and the meetings workplaces save for later can increasingly be searched, read, verified, and cited. A transcript leaves the listener with more than a voice to remember: it leaves the words where they can answer for themselves.

You might also like:

• The Atlas of Learning: A librarian’s map of where knowledge lives, and how to find our way through the research world.

• YouTube is a Research Library. Here’s How to Search It Like One.: 30 YouTube tools to find exactly what you need, and the verification framework underneath all of them.

• What “Do Your Own Research” Actually Means: A step-by-step methodology for evaluating any claim and exploring any subject.

The free essays are the foundation. The paid tier is where information literacy becomes a skill you can use: the Card Catalog Classroom with its searchable Reference Collection and structured lessons across four learning domains, plus biweekly AI Briefings, monthly Librarian Hotlines, and quarterly deep-dive guides.

Upgrade to paid for the full Card Catalog. Thank you for being here!

Subscribe now

The recommended shelf

For anyone who wants somewhere to begin, these are shows that offer strong reporting, knowledgeable hosts or guests, and a visible trail back to sources. They run in several languages and reach well past the English-speaking world. Many publish on YouTube or post useful notes and links, so the transcript and search methods above will often apply. It’s a starting point for reputable podcasts out there, not the final say on which ones to trust.

Science

• The Nature Podcast (English)
  Produced by the journal Nature, with reporting tied to newly published research and links to papers behind many stories. It runs as a brisk news briefing, and a single episode can lead to a study behind a recent finding.

• Science Vs (English)
  Wendy Zukerman takes controversial questions through the available research, with a comic touch and source lists in the notes.

• Ologies (English)
  Alie Ward interviews specialists from a different field in each episode, warm and quick, with further reading and source links in the notes.

• Coffee Break: Señal y Ruido (Spanish)
  A weekly Spanish science roundtable, loose and argumentative, that reads recent papers and reports back. The show describes itself as separating sound science from noise.

• Naruhodo (Portuguese)
  A Brazilian show that takes questions from science and everyday life through evidence and explanation. Verify the relevant episode notes when using it as a source.

• The Science Show (English)
  ABC Radio National’s long-running science program, presented by Robyn Williams and Belinda Smith. It brings researchers into conversations that range widely across contemporary science and scientific debate.

History

• The Rest Is History (English)

  Historians Tom Holland and Dominic Sandbrook talk a subject through across several episodes at a time, from antiquity to the twentieth century. The archive runs past eight hundred episodes, and the British Academy gave the show its President’s Medal (a first for a podcast).

• Empire (English)

  Historian William Dalrymple and journalist Anita Anand cover the world’s empires, from the Mughals and the Ottomans to the British. Academic guests come in for the ground each knows best, and the questions stay on how empire made the present.

• Le Cours de l’histoire (French)
  France Culture’s daily history program, presented by Xavier Mauduit, with historians discussing their research and recent books.

• Geschichten aus der Geschichte (German)
  Two historians take turns walking each other through a story from the past. It’s a weekly, source-conscious show with a large back catalog.

• Il podcast di Alessandro Barbero (Italian)
  These unofficial feeds gather public lectures by the medieval historian Alessandro Barbero into an enormous history archive. The scholarship and voice are Barbero’s, though the selection, framing, and feed management belong to unofficial publishers.

• Fall of Civilizations (English)
  Paul Cooper’s long, atmospheric episodes examine societies in crisis and collapse, global in scope and accompanied by source lists.

• Revolutions (English)
  Mike Duncan’s narrative history follows the major revolutions from the English Revolution through the Russian Revolution. Its ten-season original run remains a patient, evenhanded course in political upheaval.

• COTEN RADIO (Japanese)
  A Japanese history podcast produced by COTEN, a company that organizes historical research and turns it into long episodes on figures, events, and civilizations.

• The Ancients (English)

  Tristan Hughes interviews a different historian or archaeologist each episode on one theme from antiquity, twice a week. The guests are working academics, and the notes link their books and the related episodes.

• The History of Africa Podcast (English)
  A narrative history of African empires, kingdoms, and societies that takes a deliberately non-Eurocentric approach.

Big ideas and long-form interviews

• Sean Carroll’s Mindscape (English)
  The physicist Sean Carroll interviews people working across science, philosophy, and public life. Conversations can become technical, and their value often lies in hearing specialists pursue an idea at length.

• Dwarkesh Podcast (English)
  Dwarkesh Patel conducts unusually long, closely prepared interviews on artificial intelligence and frontier science. Use each episode as an opening into the guest’s work, not as a substitute for it.

• Conversations with Tyler (English)
  Tyler Cowen’s fast interviews range across economics, culture, technology, and public life. He often approaches economic questions from a classical-liberal perspective, a useful lens to keep in view.

• The Seen and the Unseen (English)
  Amit Varma’s extended interviews move through India’s economy, politics, history, and culture. The show's notes can be unusually extensive, and Varma brings a declared free-market sensibility to the economics.

Share Card Catalog

This week:

• OpenAI now places teenagers in a stricter version of ChatGPT, guessing each user’s age from the way they use the account.

• The private conversations people have with chatbots are surfacing as evidence in courtrooms and government files alike.

• Google’s new phone can turn American Sign Language into written English, similar to how voice dictation translates into text.

OpenAI builds a teen ChatGPT

What happened

On August 18, OpenAI introduced ChatGPT for Teens, a more guarded version of its chatbot, with extra safety limits switched on by default for anyone OpenAI takes to be under 18. There’s no separate app to download. The same ChatGPT simply tightens its rules when OpenAI’s system decides a user is a minor, and teenagers who state an age between 13 and 17 are placed in the teen version automatically.

Behind the feature sits a system OpenAI calls age prediction, which estimates how old a user is from a mix of behavioral and account-level signals, rather than relying only on the birthday entered at signup. When the system can’t tell, it errs young and applies the under-18 limits, and an adult wrongly sorted into the teen version can verify their age to get the full version back. Inside the teen version, the chatbot backs away from content OpenAI counts as unsafe or developmentally inappropriate for minors, including graphic violence, sexual or romantic roleplay, self-harm, dangerous activities, and material promoting unhealthy dieting or body shaming.

Parents can tether their own account to a teen’s and, from there, set Quiet Hours, manage selected settings, and switch off features such as memory. What they can’t do is read the conversations. If the system senses a teen in acute distress, it may alert the parent or guardian; OpenAI says such notifications are limited to high-risk situations and share only the information needed to support the teen’s safety. These protections arrive years after teenagers began using ChatGPT, and OpenAI says they draw on developmental science and guidance from outside experts.

Why it matters

OpenAI’s teen version marks a change in what a chatbot is expected to know about the person using it. The company is not merely offering a safer setting for young people; it’s building a system that sorts users into different versions of the service according to an inferred trait. That makes age less a fact a person supplies than a judgment the platform makes.

OpenAI isn’t acting alone. In recent years, regulators across Europe and the United States have been writing rules that push online platforms to establish how old their users are, and OpenAI is among the first to build that work into a chatbot used by hundreds of millions of people. If it works at scale, other companies may follow: a way to sort children from adults without making everyone show an ID.

The immediate question is whether the system protects teenagers. The longer one is what becomes normal once platforms begin drawing conclusions about their users without being directly told: how much they can infer, what those inferences can decide, and how a person can challenge a wrong one.

What it means for us

The age-guessing doesn’t stop at teenagers. When OpenAI’s guess about our own age misses, its fix is to ask for a selfie or, in some cases, a government ID. That’s the bargain now settling over the services we use every day: accept the software’s read on who we are, or hand over more of ourselves to prove that read wrong.

For anyone raising a teenager, the change is more immediate. The version their child sees now hangs on OpenAI’s guess about the child’s age, and that guess is imperfect. A teenager can enter a false age, even if OpenAI says its system is designed to look beyond what the account holder declares. That’s reason enough to switch the parental controls on rather than trust the guess alone. For the rest of us, the lesson is to notice what a tool assumes about us by default, and what it needs to collect to make those assumptions. Those decisions are often made long before we know to question them.

Bottom Line:

The age check is the visible part. Underneath it, OpenAI has built an engine that studies behavior closely enough to make a judgment about a user’s profile. Once a company can sort people this way in the name of child safety, it’s established a broader habit: using what people do on a platform to decide what they should be allowed to see there.

Age is only one trait a platform might try to infer. Other systems already claim to read mood, interests, or political leanings from behavior, often to decide what to show a person, when to show it, and how best to hold their attention. Child protection is the rare aim almost no one will argue against, which makes it a powerful justification for building this kind of machinery. The harder question is where the boundary lies: what else these systems may infer, what those inferences may be used to decide, and whether users will get a meaningful way to challenge them.

Whether age-guessing remains a narrow safeguard or becomes part of a wider pattern of platforms drawing conclusions we did not choose to share is a question still being worked out.

Our AI chats can surface as evidence

Read more

Every month, the Librarian Hotline answers a reader’s question, and those pieces live in the paid library. I’m closing out August by taking the paywall down on one of them, so everyone can enjoy a full edition. If you’ve been wondering whether the Hotline is for you, start here, and pass it on!

The reference interview is a core practice in library science. When a patron approaches the desk and says “I need information about climate change,” a librarian doesn’t pull a stack of books and call it done. They ask questions: What are you trying to understand? What do you already know? What will you do with what you find? The conversation is part of the work, because the quality of what we find depends on the precision with which we’ve named what we’re looking for.

AI breaks this contract. A model trained on enormous amounts of human language is extraordinarily fluent at sounding authoritative regardless of how well we’ve framed our question. It will answer when the question is underspecified, when context is missing, when what we need isn’t quite what we asked for. Unlike a librarian, it rarely tells us we’ve come in with the wrong question. It will just answer the one we gave it, confidently and completely, and leave the gap between what we asked and what we needed entirely invisible.

The breakdown happens in what we do before and after the prompt, not in the prompt itself.

Reader Question

“I’ve been using AI at work for about six months. I ask questions, use what comes back, and move on. But something has started to bother me: sometimes I can’t fully explain decisions I’ve made with AI’s help, or I’ll use an AI-generated summary and find out later it missed something important. A coworker told me I’m just ‘prompting wrong,’ but that doesn’t feel like the whole answer — I follow prompt engineering best practices. What’s going on?”

— Dev P.

Librarian Answer

Your coworker is addressing a symptom. Prompting technique matters, but what you’re describing runs deeper: the cognitive work of evaluation, interrogation, and synthesis got handed off to the model, and it wasn’t equipped to do it. AI can produce fluent, confident-sounding output without understanding your context, your stakes, or what “right” means for your specific situation. That judgment belongs to you, and it doesn’t happen automatically just because you’re using the tool.

When We Outsource the Thinking

There’s a version of AI use that looks like research but functions more like delegation. We bring a question, receive an answer, and move on. The output arrives formatted and fluent, which triggers the same cognitive response we’d have to a well-organized report: it seems finished and authoritative, so we treat it that way.

What disappears in that exchange is cognitive friction, and with it goes the understanding that friction was building. Cognitive friction is the effortful feeling of working through a problem rather than accepting the first answer that arrives. It’s what happens when we read a source and push back on a claim, when we try to explain something and realize we can’t quite do it yet, or when two conflicting accounts force us to decide which one to trust. That discomfort is also how we build the capacity to hold information well, to know when to trust a conclusion and when to keep pulling the thread

A 2011 study published in Science found that when people know they can look something up, they’re less likely to retain it - the brain treats external systems as a kind of storage offload. The effect may be more pronounced with AI, which doesn’t only retrieve information but synthesizes and presents conclusions. We’re not even doing the organizational work of sorting through results anymore; the model is. And when we consistently let it, we risk not developing the judgment that comes from doing that work ourselves.

Dev’s situation reflects this directly. When we can’t explain a decision we made with AI’s help, the reasoning may never have been built in the first place: a conclusion arrived, got used, and the thinking that would let us defend or interrogate that conclusion never took place. What looks like a productivity gain in the short term carries a real risk, and over months and years of consistent practice, the pattern recognition and critical judgment we rely on can erode in ways we don’t notice until we need them.

The gap between “I have an answer” and “I understand this” widens every time we skip the friction, and no amount of better prompting narrows that distance. Output quality matters, but the deeper problem is the thinking we bypass in order to get there faster.

Why Think With AI at All

The case against passive AI use is also the case for something better.

A model trained on the breadth of human knowledge can hold a vast amount of context simultaneously and draw on it in response to what we specifically bring. When we bring something real — a problem we’re stuck on, a position we’re not sure we believe, a decision we can’t seem to make — the model has something to work with, and what comes back is material for thinking rather than a substitute for it.

Used this way, AI can surface a counterargument we hadn’t considered, push back on an assumption we didn’t know we were making, or expose a connection across two bodies of knowledge that might otherwise be difficult to find. A researcher who asks a model to challenge their hypothesis before they finalize their analysis catches problems earlier. A manager who uses AI to pressure-test a proposal before presenting it walks into the room with stronger reasoning. A writer who asks for the weakest point in their argument and then rewrites that section themselves produces better work than either step would produce alone.

In none of those cases did the model do the thinking. It created conditions for better thinking to happen.

This is also what accumulates over time in the opposite direction from passive use. Instead of eroding judgment by outsourcing it, active engagement builds it, because every time we interrogate an output, test it against what we know, and decide what to do with it, we’re doing the cognitive work that makes us sharper at the next decision. Passive use and active engagement compound in opposite directions. The question is which one we’re practicing.

Thinking with AI is less about caution and more about return on investment. When we stay in the room as active participants, the work tends to get better — and so does our capacity to do it.

The Confidence Problem

AI models don’t know when they’re wrong in a way that flags itself clearly. Incorrect information arrives in the same confident register as correct information. The model isn’t hedging based on its actual certainty: it has no epistemic state to draw on. It’s producing the most statistically plausible continuation of our prompt, which is a fundamentally different operation than verifying a fact or reasoning through a problem.

When a doctor says “I’m not sure, let me check,” that uncertainty is meaningful information. It tells us something real about the state of their knowledge. An AI doesn’t have that mechanism. It can be instructed to say “I’m not certain,” but that phrase doesn’t correspond to any internal state of knowing or not knowing. It’s just another output pattern, produced in some cases and not others, without consistent correlation to whether the information is accurate.

The missing summary detail Dev mentions is one of the most common and least visible ways this plays out. When an AI summarizes a document, it’s selecting for what sounds like a coherent summary, which isn’t the same as selecting for what we need to know. A model summarizing a research report tends to foreground the headline findings, the ones that read like conclusions, and compress or drop the methodological detail that would change how we interpret them — not because it made a judgment call, but because that detail doesn’t pattern-match to “key finding.”

The same dynamic appears with contested or time-sensitive questions. Ask an AI about an ongoing legal case, a company’s current leadership, or a scientific question still being actively debated, and it will often produce an answer that sounds settled, because the model is trained to produce answers, and answers sound confident by nature. The confidence is structural, not earned.

Healthy skepticism toward AI output isn’t excessive caution. It’s an appropriate response to a tool that can be wrong without any indication of error.

The Reference Interview, Applied

A practiced reference librarian doesn’t answer the question as asked. They answer the question behind the question — the one the patron actually came in with, which is often different from the one they put into words.

The reference interview works because the librarian and patron are thinking together toward something neither could reach alone. Asking AI for an answer is a transaction: we know what we want, we’re just retrieving it. A thinking partnership is something different — we arrive with something unresolved and use the conversation to work through it, with the model as a collaborator in the process rather than a dispenser of conclusions.

What the shift looks like in practice:

• Asking AI: “What are the main causes of employee burnout?”

• Thinking with AI: “Three people on my team seem fine, two seem checked out, and the only thing that changed two months ago was switching to asynchronous communication. I keep assuming it’s the structure, but what else should I be considering before I draw that conclusion?”

• Asking AI: “Summarize this article.”

• Thinking with AI: “I want to reference this piece in something I’m writing, but I realized I can’t articulate its actual argument — I just came away with a general vibe. Can you help me distinguish what the author is claiming from what I’m inferring?”

• Asking AI: “Should I take this job offer?”

• Thinking with AI: “I’ve made a pro/con list for a new job and it didn’t help. I think the problem is that I’m weighing things I can measure — salary, title, commute — but the hesitation is about something I can’t put my finger on. Can you help me figure out what’s actually at stake for me here?”

• Asking AI: “What’s the difference between a Roth IRA and a traditional IRA?”

• Thinking with AI: “I understand the tax difference between a Roth IRA and traditional IRA in theory, but every time I try to apply it to my actual life I freeze because I don’t know what my future tax bracket will be. Can you help me figure out what I’d need to know about my own situation to make this less abstract?”

• Asking AI: “Is this a good business idea?”

• Thinking with AI: “I’ve been sitting with this idea for months and everyone I’ve told has been encouraging, which somehow makes me more uncertain. I can’t tell if I believe in it or if I’m just attached to it. What’s the actual difference between those two things?”

• Asking AI: “What should I say in this performance review?”

• Thinking with AI: “I’m writing feedback for someone who has real skill but a pattern that’s hurting the team. I keep cushioning it and I can’t tell if I’m being diplomatic or dishonest. What’s the difference, and how do I stay on the right side of it?”

• Asking AI: “How do I know if I should change careers?”

• Thinking with AI: “I’m good at my current job and drawn to something else, and the fact that both things are true is what makes this hard. I don’t need help identifying my options — I need help figuring out what kind of decision this even is. Is this a risk calculation or a values question?”

• Asking AI: “What’s a good way to handle this conflict with a coworker?”

• Thinking with AI: “Something’s off with how credit is landing on a project I’m sharing with a colleague, but I keep going back and forth between ‘this is a real problem’ and ‘I’m being oversensitive.’ Can you help me think about what evidence would separate those two?”

• Asking AI: “Write a follow-up email after a networking meeting.”

• Thinking with AI: “I’m writing a follow-up to someone I’d like to stay in touch with, and every version I draft sounds like a version of me that isn’t quite right — either too eager or too casual. Can you help me figure out what I actually want from this connection so I can write something that sounds like me?”

The pattern is the same in every case. Asking AI hands the model an open question and accepts whatever frame it chooses. Thinking with AI brings the actual situation, names what we’re trying to figure out, and asks the model to work inside our reasoning rather than substitute for it. What comes back is still just output, but it’s output we’re in a better position to evaluate.

What AI Cannot Supply

Everything an AI model knows about our situation is what we’ve told it in the conversation. It has no access to the history of a working relationship, the political dynamics of an organization, the constraint that makes an otherwise sensible solution unworkable, or the value judgment that has to land at the end of an analysis — unless we put those things on the page. That’s not a design flaw. It’s just the nature of the exchange, and it means the quality of what we get back is inseparable from the quality of what we bring.

This is why the thinking can’t be fully delegated. The model generates plausible text. But we determine whether that text is true, complete, and appropriate for the actual decision we’re facing. Those are different jobs, and only one of them requires knowing what’s really at stake.

That practice, repeated over time, is also what protects something harder to recover once it’s gone. The judgment we build through sustained cognitive engagement develops through use and atrophies when we stop. It degrades gradually, in the same way any skill erodes when a tool starts doing the work instead.

The people who get the most out of AI over the long run aren’t the ones who use it the most. They’re the ones who stay in the room while they use it, who treat every output as something to be interrogated rather than accepted, and who bring their own critical thinking to the exchange. AI can be one of the most powerful cognitive tools we’ve ever had access to. Whether it functions that way depends entirely on what we bring to it.

Quick summary:

An AI company has been buying up used books by the million and destroying them, scanning the pages for training data and pulping what’s left. Court filings unsealed this year describe the program and an internal note asking that the work be kept from becoming known. A court has ruled that buying and scanning the books this way is legal, and once the work is done, nothing survives to show a book was ever there to lose.

Key takeaways:

→ The books are bought through anonymous middlemen, so the sellers filling the orders rarely know where their stock is headed. From every angle the buying looks like ordinary commerce, which is exactly what keeps the destruction from being seen.

→ Book destruction has a long history, and it changes each time the technology of copying changes. It has shifted twice before. This is the third shift, and it looks nothing like the book burnings most of us picture.

→ This shift is set apart by leaving nothing behind. The words are kept, the object is discarded, and no record says which titles were taken, so the loss can be neither proven nor traced to anyone who might answer for the harm.

→ That reaches past books to anyone who wants to know things firsthand. When the only surviving copy is sealed inside a system no outsider can consult, verifying what it says becomes impossible, and we are left trusting whatever summary we are given.

“We don’t want it to be known that we are working on this.” The sentence appears in an internal Anthropic planning document made public through court filings in Bartz v. Anthropic, the copyright class action that three authors filed in 2024. Anthropic, the maker of Claude, called the program Project Panama. Its stated mission fit in one sentence: “Project Panama is our effort to destructively scan all the books in the world.”

Destructive scanning means cutting a book from its binding, feeding the loose pages through a scanner, and discarding the book afterward. According to the court, Anthropic spent millions of dollars buying millions of print books, often used. Its service providers removed the bindings, cut the pages, scanned them into searchable digital files, and discarded the books. Anthropic kept the scans in an internal digital library and used books from that library to train the AI systems behind Claude.

The lawsuit initially concerned another source of Anthropic’s library: more than seven million pirated book files the company had downloaded. The court treated the two acquisition paths differently. It held that Anthropic could lawfully buy print books, convert them into searchable digital files for internal use, and discard the physical copies; because the resulting files remained inside the company rather than being redistributed, that practice was fair use. It reached the opposite conclusion about the books Anthropic had downloaded from pirate sites and retained.

Companies announce the programs they are proud of; Anthropic tried to keep this one invisible. The January unsealing supplied the project’s codename and the instruction to stay silent. The instruction anticipated what the company did not want authors, booksellers, and the reading public to see: books bought by the million, cut apart, scanned, and discarded. Anthropic didn’t publicly announce Project Panama before court records made it public.

The silence has no precedent. Books have been destroyed for as long as they’ve been made: by conquering armies and offended churches, by censors with lists and mobs with torches, by floods and fires, and by budgets that let the roof leak. Some of it was fast and public. More of it was slow and official. All of it, loud or slow, could be recognized for what it was. A person watching knew that books were being lost. History kept what record it could. But what’s happening now carries no such signature. From the outside, nobody could tell that books were being destroyed at all.

Inside Project Panama

Anthropic hired a man named Tom Turvey in February 2024 and gave him a mission the court record repeats in one sweeping phrase: obtaining all the books in the world. Turvey came to the job from Google Books, the project that spent the 2000s digitizing library collections on machines engineered to turn pages gently, so that every book survived its own scanning. Anthropic took the opposite approach. Gentle machinery never entered the plan. Over roughly a year, the company spent tens of millions of dollars buying millions of print books, sheared off their spines, fed the loose pages through high-speed scanners, and pulped what remained. Vendor proposals in the court records described converting up to two million books in six months, some eleven thousand a day.

Destroying the books solved a financial problem first. A bound book must be scanned page by page, slowly and at cost. Once cut apart, the same volume becomes a stack of loose paper that can run through a sheet feeder at speed. Across millions of volumes, the difference in output determined the method.

The legal significance emerged later. In June 2025, Judge William Alsup ruled on the authors’ claims. For the books Anthropic had bought and destroyed, he found the copying to be fair use: the rule in American copyright law that allows limited copying without an author’s permission (the way a critic can quote a novel in a review). His reasoning turned on replacement. Anthropic bought one physical copy, made one digital copy, and destroyed the original—so the number of copies in the world never grew. In the law’s eyes, the scan simply took the book’s place, a change of format. If Anthropic had kept both the book and the scan, the number of copies would have doubled and the fair-use argument would have weakened. Pulping the originals helped make the copying legal.

The pirated downloads fared differently in the same ruling. For those more than seven million files, Alsup rejected Anthropic’s fair-use defense. He emphasized that Anthropic had built a permanent, general-purpose library it expected to keep indefinitely, not a temporary collection for a defined training task. “None is even offered here except for Anthropic’s pocketbook and convenience,” he wrote.

The ruling left the company facing a trial over damages, and Anthropic settled instead: the company agreed to pay $1.5 billion. A judge granted the deal final approval on July 20, the largest copyright class settlement in American history. Under the deal, Anthropic will delete the pirated digital files. Critically, the deal concerns those files, not the millions of physical books the company bought and pulped. Copyright protects the text of a book—the creative expression of an idea—not the paper on which those words were printed. Once Anthropic lawfully owned a physical copy, copyright law generally didn’t prevent it from destroying that object. The books could be destroyed without creating the kind of copyright violation at issue in the case.

But books don’t enter a library by themselves: somebody had to sell the company all those books. The court records describe Anthropic purchasing through vendors. This spring and summer, booksellers across Europe described the other side of such a trade: unusually large, eclectic orders placed through intermediaries, with the ultimate buyer unnamed. Tomás Kenny of Kennys of Galway called one order for books “bananas”—a mix of titles no library would plausibly assemble. Whether any particular order came from Anthropic cannot be established from outside the transaction. Nor is Anthropic the only possible customer: other AI companies have also been reported to be acquiring books at industrial scale.

The reported market is built to keep the buyer at a distance. Brokers can aggregate inventory and manage bulk orders without disclosing who is ultimately acquiring the books. ISBNdb, a book-data company, briefly advertised a prospective book-sourcing service for AI developers that promised confidentiality; its marketing explained the appeal bluntly: “‘AI company destroys two million books’ is not a headline that generates sympathy.” (After reporting drew attention to the page, ISBNdb removed it and said the proposed service had never been launched.) The result was not merely secrecy about the buyer, but uncertainty about the fate of the books.

Annihilation, then spectacle

What Anthropic is doing belongs to a history far older than the company. Rebecca Knuth, a professor of library science, named the practice in 2003: libricide, the systematic destruction of books and libraries, usually carried out or authorized by a government. Her subject was the twentieth century’s state-sponsored campaigns, but the practice runs back as far as writing does. The scenes that come to mind of this are the same few: students feeding bonfires in Berlin in 1933, Sarajevo’s national library burning under siege in 1992. But behind those scenes, the record is wider and stranger than they suggest. Bonfires were rarer than the memory of them. Most destruction arrived slowly and with permission, through purges, censors, wars, and simple neglect.

A strict reading of that definition would leave Anthropic out, reserving libricide for the destruction of entire libraries. But that distinction collapses here. The world’s secondhand book trade functions as one enormous collection, scattered across thousands of shops and sellers with no central address. Buying it up by the pallet and pulping it empties a library all the same, just one whose shelves span continents.

Destroying a book has meant different things in different centuries. The difference has always come down to copying, because how books are copied decides how many of any one book exist. When copies are scarce, destruction can erase a work from existence. Once copies are everywhere, destruction can only send a message about one. To me, that line sorts the history of libricide into eras, and what distinguishes each one is what its destruction leaves behind. The result is my own framework, a chronology by residue that I haven’t found anywhere in the scholarship: two eras completed, and a third that has just begun.

For thousands of years, every copy of every book was made by hand. A single volume could take a scribe months, so most works existed in a few manuscripts, and some in only one. Under those conditions, destroying the object destroyed the work, completely and forever. I call this the first era: the era of destruction as annihilation. The word descends from the Latin ad nihil, meaning to reduce or bring something to nothing. In this era the meaning was literal. When Diego de Landa, a Spanish friar in colonial Yucatán, burned twenty-seven Maya codices in 1562, the texts inside them went to ash. No copies of them existed anywhere on earth, so entire bodies of Maya history and belief ended in one afternoon’s fire. The library of Alexandria met the slower version of the same fate, declining through centuries of purges and neglect until its losses ran past counting. What the first era left behind was dust, and one thing more: knowledge of the loss. Contemporaries recorded what had burned. We can still mourn the codices, because the one thing annihilation couldn’t destroy was the memory that the books had existed.

The printing press ended that era within a century of its invention. Once a title could exist in hundreds or thousands of identical copies spread across cities and countries, fire lost its reach. Burning a book now destroyed only an object, since the work lived on in every other copy, safely out of range. Destruction continued anyway, though, serving a different purpose entirely. I call this the second era, the era of destruction as spectacle (a word descended from the Latin spectare, to watch). By the twentieth century, watching had become the entire point of burning a book. The clearest case is the Nazi bonfires of May 1933, when German students burned tens of thousands of volumes in public squares, in front of rolling newsreel cameras. Almost none of the works truly died in those fires, because the titles on the pyres existed in editions across Europe and America (many of which remain in print today). Erasing the books was never the goal, since erasure had stopped being possible. The fires existed to be seen, a threat performed first for the crowd in the square and then for everyone who watched the footage. What the second era left behind was the opposite of ash: photographs and visuals of fires that consumed real books but reached nothing beyond them. Once a book existed in enough copies, burning one no longer removed it from the world. It announced that the book had no place in the world to come.

Destruction as disappearance

Measured against those two eras, what Anthropic is doing fits neither. In the first era, destroying the object meant losing the work. Anthropic’s scanners preserve every word, so the works survive. In the second era, the objects were beside the point and the burning was public theater. This time the objects are destroyed by the millions, and the destruction says nothing at all; it’s not a message but a method. The company ordered the operation kept out of sight. A destruction that erases no text and performs for no one, run at industrial speed, matches neither pattern. The technology of copying has crossed another threshold, the way it did when the press replaced the scribe. What it means to destroy a book now has changed again to match. We’ve entered the third era.

What exactly went into the scanners is the question nobody outside can answer. The unsealed documents show the program favored what its leader called “less common” books, harder-to-find titles over mass-market ones, without ever defining where less common ends. After the claims went viral this summer, the fact-checking site Snopes investigated whether rare books were being pulped. Anthropic told Snopes that “none of our data acquisition programs buy and destroy ‘rare’ or ‘antiquarian’ books.” The assurance can’t be tested from outside, because no list of what was bought has ever been made public.

Anthropic’s own planners estimated the world has about 130 million distinct books. The program destroyed millions of copies, most of them ordinary used books with plenty of surviving duplicates. Even where duplicates survive, a scan keeps only the words. A physical copy carries evidence too, like a censored paragraph that marks one printing apart from another, or an owner’s name inked inside the cover. But the deeper danger sits in the margin nobody can see. Anthropic aimed for “uncommon” titles while recording nothing public about which copies were destroyed. For a book surviving in only a handful of copies anywhere, one bulk order can potentially take the last one. Whether that has already happened is a question no one on the outside knows for sure. The impossibility of answering what was destroyed is what’s new.

The buying also leans toward older books, for a reason that has nothing to do with rarity. Since around 2022, text written by AI has spread across the internet, mixed in with everything people write and mostly impossible to tell apart. That creates a problem for the companies training new models. Feeding a model text written by other models tends to degrade the results, so the builders want sources guaranteed to be human. A book printed before the technology existed carries that guarantee on its copyright page. Old print has become a raw material, valued for the one quality the internet can no longer promise.

One more feature separates this era from the last one. Spectacle-era destruction wanted an audience; this destruction wants the opposite. The buyer is after the text and has no message to send. In fact, attention to its process can only slow the buying down. Taken together, the features of this era line up: the works survive inside a black box, the objects vanish by the millions, no public record exists, and the operation itself prefers to work in secrecy. I call this the third era, the era of destruction as disappearance. The word rests on the Latin apparere, to come into view, with a prefix that reverses the motion. A disappearance is a departure from visibility. The word fits this era at every layer, from the unseen sales to a program that only appeared when a court forced it into view.

Every one of these books, preserved to the letter, can now be read by no one. The scanned texts sit in Anthropic’s private collection, a library with no reading room and no public catalog. Models trained on that collection are built to avoid quoting it at length, because reproducing long passages for users is the copyright violation no court has excused. So the machines answer questions about books without ever showing the books themselves. When someone asks a model about a title that survives only in that collection, what comes back is a summary, written in the model’s words, with the original nowhere in reach.

That arrangement changes something basic about how we can know things. Checking a claim against its source is the foundation of thinking for ourselves. When we can pull the book, find the page, and read the passage in context, we get to judge whether the summary was faithful and whether the quote meant what someone claims it meant. Remove the book and that judgment has nowhere to stand. We’re left taking the summary on trust, with no way to confirm and no standing to doubt. Whatever the model says the book said becomes, for all practical purposes, what the book said. That’s a transfer of authority from the page to the tool, from a source anyone could check to an answer nobody can. The transfer happens one unreachable book at a time.

Maddeningly, every piece of this design is legal. Each one also removes a question that used to have an answer. The anonymous purchasing means nobody can list which books were destroyed, so nobody can rule out that some were the last copies anywhere. With the collection sealed, nobody can compare a stored text against the printed original it replaced, so even perfect fidelity can never be shown. As for whether the buying ever stopped, nobody outside knows that either, since the only reason anyone knows it started is that a piracy lawsuit dragged the records into the open. None of this took a conspiracy, only ordinary business decisions about what to disclose, all of them legal and every one of them closing a door.

One more thing disappears along with these books: the ability to mourn them. The previous era’s destruction left survivors who could testify to what was lost. Alexandria’s losses were lamented for centuries. Sarajevo’s librarians catalogued what their fire took. The third era leaves no one who can even compile the list. A loss we can name is a wound. One we can't name is just a world grown slightly smaller, with nothing to point to and no way to prove it. A librarian would describe the situation in the profession’s terms: no accession record and no deaccession record, a transaction that leaves no ledger for anyone to audit. Put simply, they’ve created a system in which they can’t be held accountable for the system they’ve created. That understanding defines destruction as disappearance: the loss was built, from the start, to be impossible to establish.

What can still be known

The concealment had one structural weakness: a program that buys millions of books needs sellers, and a company that breaks the law can be sued. Sellers and courts are the two doors into the secrecy — sellers see every order even when they can’t see the buyer, and courts can compel what the company won’t volunteer. Everything now known about the program came through one door or the other. Authors sued and forced the internal records open. Booksellers noticed matching orders across two countries and brought them to reporters; one worked with the journalists at 404 Media to hide a tracking device inside a shipment of books, and the tracker led to an Amazon warehouse outside Las Vegas — evidence that a retail giant appears to be running a scanning program of its own. Within weeks, at least one supplier that had been advertising bulk books for AI training pulled the offer. No regulator opened either door. Every disclosure came from the people the company bought from or the people it answered to in court.

Some sellers went further than noticing. Once Tomás Kenny, whose Galway shop had received one of the strange 5,000-book orders, worked out where books like his might be headed, he said publicly that Kennys wanted no part of scanning aimed at extracting intellectual property. His shop had been the second in the world to put its books online, back in 1994; it now became one of the first to refuse the trade that takes books offline for good. Kenny could refuse because he had worked out the destination, which is exactly the discovery the brokers’ anonymity exists to prevent. A trade that hides its purpose from its own suppliers has already answered the question of whether the suppliers would approve.

The same kind of attention is available to the rest of us, because most of us eventually stand over a box of books deciding where they go, after a move or the clearing of a family house. That box is where all of this arrives at our own doorstep. A seller, or any of us selling to one, can now ask a question that has a good reason to be asked: where do these books go next? The anonymity that keeps this market running survives only as long as nobody asks. For material that might be scarce, like a town history or a box of local records, the open market is no longer the safe default. A library’s special collections desk exists for exactly that kind of donation. And the same habit applies at the other end of the pipeline: when a model summarizes a book for us, we can treat the answer as a starting point and go find the book itself, while findable copies still exist. Each of those is a small act of keeping track.

That kind of record-keeping has been the difference between the eras all along. Each era’s destruction left a residue, a testament of a kind. The first era of annihilation left ashes and knowledge of the devastation. The second era of spectacle left photographs of the fires and the threat those fires were meant to carry. Ours, this third era of disappearance, leaves no ash, no image, and no list. But what’s still open is the record itself: whether one gets kept, and whether anyone outside can access it at all.

The three eras are a framework I built for this piece, and I want to take it further (it’s such fascinating stuff!): a full treatment of how book destruction has changed across five thousand years and what the third era asks of the people living in it. Before I build it though, I’d love to get a temperature read on the format you’re most interested in.

The poll's anonymous, so if you'd want this for a group or an organization, leave a comment, reply to this email, or send me a DM on Substack so we can connect.

You might also like:

• The Libraries That Weren’t Supposed to Exist: Basement libraries and buried archives: a brief history of reading against the state.

• Who Benefits From Us Not Knowing the Full Picture?: How information asymmetry shapes what we buy and believe.

• How One Rescue Mission Turned into the Global Record of Human Heritage: Inside the preservation system that classifies, catalogs, and protects humanity’s cultural and natural inheritance.

The free essays are the foundation. The paid tier is where information literacy becomes a skill you can use: the Card Catalog Classroom with its searchable Reference Collection and structured lessons across four learning domains, plus biweekly AI Briefings, monthly Librarian Hotlines, and quarterly deep-dive guides.

Upgrade to paid for the full Card Catalog. Thank you for being here!

Subscribe now

Have you read the Founding Member Report: The State of AI yet?

A comprehensive guide for information navigators who want to understand where AI is actually heading and what it means for how we find, evaluate, and use information in 2026.

→ Find out more here.

That experience can produce a private sense of failure. We may conclude that we read too slowly, missed essential context, or lacked the discipline to keep up. The problem often lies elsewhere. Public information now arrives through systems that reward speed, repetition, emotional reaction, and constant return. When the volume exceeds the time available for verification and reflection, the difficulty is by design.

Information literacy begins there. It doesn’t ask us to consume every available piece of information out there, but gives us ways to assess origin, context, evidence, timing, and purpose. Those habits become especially valuable when a public event is surrounded by so much material that no single person can follow the whole field.

In early 2018, the writer Michael Lewis watched the State of the Union address with Steve Bannon, who had left the Trump White House the previous summer. Bannon described his approach to dealing with the media in an unusually direct phrase: “The Democrats don’t matter. The real opposition is the media. And the way to deal with them is to flood the zone with shit.” The phrase became shorthand for a familiar political and media tactic: saturating the public sphere with claims, controversies, provocations, and announcements so that sustained attention to any one damaging development becomes harder to maintain.

A flood of information doesn’t depend on every item being false. A fast stream can contain accurate reporting, misleading fragments, official statements, manufactured drama, and sincere public response. But the volume changes how all of them are received. Once we have more material than we can assess, the order of arrival begins to govern attention more than evidence does.

The language of flooding also explains why ordinary habits often fail during such periods. Opening one more tab doesn’t solve the problem, nor does saving every link we come across. And sharing the most alarming claim before its context is established can deepen the same conditions that made it hard to think clearly in the first place.

A practical response lies in preparation and boundaries. A flood doesn’t disappear simply because we understand that a flood is happening. We can, however, decide how much enters our day, when we examine claims, what we preserve, and which sources and institutions we trust to help watch the wider field. Those boundaries preserve our ability to follow public events without letting volume choose the order of our attention.

How the flood works

The phrase “flood the zone” comes from American football. One team tries to move the ball by throwing it to a teammate. Sometimes, it sends several teammates into the same small part of the field, each ready to receive the throw. The other team can’t watch everyone at once, so one teammate may have a clear chance to catch the ball.

Information flooding puts similar pressure on attention. Journalists and readers have limited time to investigate what they encounter. When many claims, controversies, announcements, and reactions arrive at once, they can’t give each one equal scrutiny. A consequential story may then be crowded out, while a less important or more provocative item captures attention. A stream of loosely related claims can also consume the hours that reporters and readers might otherwise spend examining a development with lasting consequences.

This tactic isn’t new. Governments and organizations have long been accused of releasing unfavorable news when public and press attention is already divided. Crowded news cycles can make sustained scrutiny harder to secure. Digital platforms intensify those pressures by distributing material continuously, while systems built to maximize engagement can give emotionally charged posts greater visibility. Fresh reactions can then seem more urgent than the evidence warrants.

The result can be disorientation. People may lose confidence in their ability to reconstruct what happened and decide which accounts deserve trust. By the time a correction arrives, attention may have shifted elsewhere. A well-reported investigation then competes with a fast-moving argument that generates repeated commentary and intense reaction.

The burden of verification also falls unevenly. Making or reposting a claim often takes less work than tracing the claim to its original source and examining the evidence in context. Comparing independently reported accounts only adds further effort. That difference gives a high-volume publisher an advantage over anyone trying to evaluate each item responsibly.

Generative AI has made one part of that imbalance more pronounced. It makes plausible-looking text, images, audio, and video easier to produce at scale. A coordinated campaign still needs to get material in front of people and make it seem worth trusting. But the supply of material is now easier to expand, and more of it can circulate through the same channels where reliable reporting competes for attention.

Ordinary acts of sharing can also move the flood forward. A screenshot can pass from one group chat to another after being separated from its original source. A video clip can circulate after its beginning or ending has been removed. With each act of sharing, another unexamined item can enter someone else’s field of attention before its meaning has been established.

Recognizing a flooded zone

A flooded zone rarely announces itself. Any one item may look ordinary on its own. The pattern becomes visible when we notice what the accumulation is doing to our attention: which subjects dominate and which questions go unanswered.

Several conditions often appear together:

• One subject suddenly dominates alerts, feeds, conversations, and commentary across many platforms

• New reports arrive faster than earlier ones can be checked or placed in context

• A major story slips from view before its central questions have been answered.

• Corrections may receive less attention than the claims that prompted them

• A sudden controversy coincides with a consequential vote, ruling, executive action, corporate decision, or policy change

A fast-moving controversy may dominate coverage at the same time a government or organization announces a decision that changes public policy or institutional practice. And when that happens, we can examine whether the controversy displaced attention from the decision without assuming that the timing was coordinated.

Some developments provide advance warning. Periods of political consequence (especially elections and major policy developments) often produce concentrated waves of coverage. Keeping a calendar of such events can help us anticipate periods when public attention may face greater pressure. The calendar can also help us decide which sources or developments deserve closer attention beforehand.

A private record can deepen that awareness. We might note a major controversy, the date it appeared, and other public actions occurring around it. Returning later can reveal which event dominated attention and what subsequent reporting clarified. The record turns vague overwhelm into an account of how public attention moved.

Information literacy asks whether a claim is supported by evidence. In a flooded zone, we also examine how the claim reached public attention and which event lost visibility after it appeared. Placing the claim within that wider information environment shows how volume affected what the public had an opportunity to examine.

How to stop a flood

When water rises, sandbags are placed between the flood and a home. They redirect the current before it enters the house, leaving the source unchanged. An information flood calls for a similar response. Trying to absorb every claim and reaction allows the flood to occupy more of the day. But we can place boundaries between the surge and the attention we need for judgment and daily life.

I call those barriers sanity sandbags. These deliberate actions limit how much of a flooded information environment enters our daily lives. They create boundaries between the flood and the parts of our life that need protection (our capacity for judgment and our nervous system regulation chief among them). We can remain aware of what’s happening without giving a relentless stream of information authority over every available moment.

Sanity sandbags can be set before predictable periods of high volume or added after the flood has begun. Each one creates more room to assess what is happening. The added distance helps us decide how to respond without allowing urgency to decide on our behalf.

Sandbag #1: Step away from the current

Create distance from the live feed.

We close the feed and step away until the next planned check-in. Channels carrying emergency instructions or information tied to an immediate responsibility remain open. Distance comes first here, because every later choice depends on having enough attention left to choose deliberately. Breaking news can feel hardest to leave because uncertainty creates pressure to keep checking. But during those periods, reporting is still being assembled. Early accounts may conflict before basic facts have been confirmed. Corrections may not reach the people who saw the first claim. When we return, the reporting may include evidence and context that were absent from the first accounts.

Stepping away during a public crisis can feel like neglecting a duty. Continued exposure can weaken the attention we need to evaluate later reporting. We step back so that we can return with enough attention to compare accounts and decide what deserves further examination.

Sandbag #2: Let the archives do the archiving

Preserve what belongs to your own work.

A flood can make every passing item feel irreplaceable. Once we step away from the feed, another anxiety appears: what if the article or post disappears before we save it? That fear keeps us tied to the current. Each new item can seem to demand a private copy.

The concern reflects a documented problem. A Pew Research Center analysis found that one quarter of webpages available at some point between 2013 and 2023 could no longer be reached in October 2023. Link rot describes one form of digital loss, when a link no longer leads to the page it once did. Some pages disappear entirely. Others change through revision or become difficult to locate as newer material accumulates.

Saving material can keep this public record within reach. The difficulty begins when the urge extends to everything moving through the feed. A personal archive can grow faster than we can organize or revisit the collection. The work of saving begins to resemble the flood itself, with each new article creating another demand on our attention.

Preservation becomes manageable when the labor is divided. We keep the pages tied directly to our responsibilities. Sources supporting our writing belong in a personal archive, and records needed for an ongoing responsibility may also deserve a personal copy. Public pages, however, can sometimes move into institutional archives. The Internet Archive’s Wayback Machine has collected copies of webpages since 1996. Its Save Page Now feature lets anyone request an archived copy of a public page. The End of Term Web Archive preserves federal websites during presidential transitions, when official information can change or disappear. So a page earns a place in our personal archive when losing it would interfere with work already in our care or remove evidence we need to retain. Material of wider public consequence can go to a shared archive, leaving our personal archive small enough to use rather than turning it into another place where the flood continues.

Sandbag #3: Schedule the intake

Choose when the news enters the day.

Once we’ve decided what belongs in our archive, we can decide when the news enters the day. We choose a fixed reading window and keep the feed closed in between. The boundary replaces ambient monitoring with a deliberate encounter that has a beginning and an endpoint.

A live feed presents events in the order they surge into view. A trusted human-edited roundup can place developments beside reporting that explains their context. Reading within a defined window places accounts beside one another, making comparison possible before a new alert redirects our attention.

A fixed reading window also limits repeated exposure. RAND’s review of psychological research notes that repetition can make a claim seem more credible. Encountering the same assertion fewer times leaves less room for familiarity to influence our judgment before we’ve examined the evidence.

Sandbag #4: Let the peak pass

Give the story time to become clear.

A breaking report can create pressure to decide immediately what happened and how we should interpret the event. But the pressure often arrives before reporting can answer either question. The conclusion we form under that pressure may rest on evidence that remains incomplete.

For developments that don’t demand immediate action, consider the first reporting as provisional. Avoid building a firm view around material created during the first rush of coverage. The first version of a story can contain a partial account without showing the sequence that gave an event its meaning. Later reporting can place that account within a fuller record, showing which conclusions the evidence supports and which questions remain open.

The pause protects memory as well as judgment. Psychologists use the term continued influence effect for the way misinformation can continue to affect reasoning after a correction. An early claim may influence how we understand an event after later reporting has shown the claim to be incomplete or false. The first account becomes a reference point for the information that follows. Waiting before we commit to an explanation gives more complete reporting a chance to arrive before that account settles into memory.

Some floods consist of accurate developments arriving too quickly for close attention. A series of executive orders or court decisions can overwhelm attention even when the reporting is sound. Waiting doesn’t mean ignoring events that affect our lives or delaying action when time matters. But the pause gives us room to decide which developments need a response now and which stories will become clearer as reporting develops.

Sandbag #5: Stop carrying the water

Pause before passing a claim along.

During a surge, passing something along can feel like a public duty. We may want to warn someone about a claim, or ask for help making sense of what we’ve seen. Before sharing, though, we can ask whether the material has a source the recipient can inspect for themselves. We can also ask whether we have enough context to explain what the material shows and what remains unresolved. When either answer is uncertain, leaving the material unshared keeps an unverified claim from entering another person’s day through someone they trust.

Pausing preserves the difference between repetition and corroboration. A claim can appear to have support when several posts repeat the same account without adding independent evidence. Sending the original source allows the recipient to examine the evidence directly. Naming any unresolved questions keeps our message from sounding more conclusive than the reporting supports.

The pause also changes our own relationship to the flood. We spend less time attached to every fast-moving item. We also spend less time following the reactions that appear after we share a message. People close to us receive fewer questionable claims through our messages. Each decision not to forward closes one route through which the flood would otherwise expand.

Sandbag #6:

Share the work of paying attention.

A flood becomes harder to manage when one person tries to watch every place where water may break through. When several areas need attention, people can divide the watch and alert one another when conditions change. Public information presents a similar problem. No one person can follow every court decision or international development with equal care. Dividing the watch lets each of us follow a smaller part of public life more closely.

We can divide that work among people whose judgment we trust. One friend may follow decisions from the courts. Another may understand events in a region through family ties or knowledge of the local language. Each person can share reporting that identifies where the information came from. They can also explain what the reporting has established and what remains uncertain.

A shared news watch can begin with one other person who follows a different area. Each person sends reporting when a development affects an issue the other has chosen to follow. A larger group can agree on a limited set of subjects and use one shared channel for reports. An incomplete watch can still extend what each member can follow. The arrangement gives us access to knowledge beyond our own attention without recreating the continuous feed inside the group.

What we take back

Every instance of the flood assumes the same thing about us: information is something done to us, arriving at a pace someone else sets and in an order someone else arranges. The assumption runs so deep that people announce the strategy openly, the way Bannon did, without fear it will stop working once we know. The people who flood the zone are attacking our attention, and deciding for ourselves how we take in the news is how we refuse them. Each time the water rises, we choose whether to be the passive audience the entire strategy depends on. The zone will flood again, around the next election or the next manufactured uproar, but putting the sanity sandbags in place can begin as early as tonight.

Every flood is built on the same wager: that the people underneath it will stay scattered, exhausted, and alone. Meeting the next one on our own terms, at our own pace, in the company of people we trust, is how they lose that bet. The water will keep rising, but it no longer gets to decide how we move through it.

You might also like:

• A Field Guide to Information Overload: How to navigate an environment built to distract us.

• The Unspoken Rules of Information Etiquette: Six practices for how to treat each other when we exchange information.

• The information diet isn’t enough. What we need is an information village.: What changes when we go beyond optimizing our consumption and start building shared infrastructure.

The free essays are the foundation. The paid tier is where information literacy becomes a skill you can use: the Card Catalog Classroom with its searchable Reference Collection and structured lessons across four learning domains, plus biweekly AI Briefings, monthly Librarian Hotlines, and quarterly deep-dive guides.

Upgrade to paid for the full Card Catalog. Thank you for being here!

Subscribe now

Have you read the Founding Member Report: The State of AI yet?

A comprehensive guide for information navigators who want to understand where AI is actually heading and what it means for how we find, evaluate, and use information in 2026.

→ Find out more here.

This week:

• A new European law requires companies to say when people are dealing with AI, whether a chatbot on a service line or a synthetic image in an ad.

• California is weighing limits on AI chatbots used in place of therapy, as more people bring their mental health to a machine.

• As a typhoon approached China, AI models built on decades of weather records forecast its path alongside the traditional systems.

Europe starts labeling its AI

What happened

As of August 2, a new European law gives people the right to be told when they are dealing with artificial intelligence. If a chatbot answers a complaint, it now has to say that no human is on the other end. An image or video in an ad that was generated or altered by AI has to carry a label marking it as synthetic. The law doesn’t ban AI from ordinary commercial life; it requires disclosure, so people can know when a machine is involved and decide how much to rely on it.

The rule covers more than the well-known AI companies. A call center that uses AI to read a caller’s mood has to disclose that at the start of the call. Even commercial uses of AI, including scheduling appointments, handling correspondence, or negotiating contracts, will have to be declared. Companies that break the rule can be fined up to 15 million euros, about 17 million dollars, or three percent of their total global revenue for the year, whichever is larger.

The rules are live, but enforcement is unlikely to be immediate or uniform. The EU’s AI Office took longer than expected to produce key compliance guidance, and the bloc’s 27 member countries are at different stages of setting up the systems meant to supervise the law. By December, AI model providers have a transition period to label synthetic audio, images, video, or text in a machine-readable form, so other software can recognize it as machine-made. Regulators acknowledge that this is technically difficult to get right.

Why it matters

This is the first time a major government has required AI to identify itself across everyday commercial life. The reasoning rests on a gap: a person who assumes a human wrote the reply or took the photo has no reason to question what they see. A label closes that gap in a usable way. Someone told that a chatbot is handling their complaint can decide whether to continue, look for another channel, or ask for a human instead. A shopper who learns a product photo is synthetic knows the image may prove little about what arrives in the box.

Europe has changed the internet from inside its own borders before. When its data-protection law, the GDPR, took effect in 2018, companies worldwide rebuilt how they handled personal data rather than run a separate European version of every product. Cookie-consent pop-ups became a familiar part of the web, reaching far beyond Europe. The AI labels may spread the same way, as companies decide it is simpler to build one disclosure system than separate European and non-European versions of their products.

Two things could still keep the law from working. If a label appears on nearly everything, from a spell-checked email to a lightly filtered photo, people may stop seeing the labels at all. One industry group has already warned of exactly that, calling it banner blindness—the same fate that overtook the cookie pop-up. The other risk is that some companies would rather scale back their AI, or bury it deeper, than advertise how much of their work a machine now does. Either way, the law could stay on the books while losing the effect it was meant to have.

What it means for us

The first change many of us notice may be small but concrete. A voice on a service line may say it is automated before we start, and an ad may note that its flawless image was built by AI. That knowledge is something we can act on: we can ask the automated line to pass us to a person, or treat a too-perfect product photo with more caution once a label reveals it was built by a machine.

The labels may reach us wherever we live, since companies adding them may choose to build one global version of a product rather than a separate one for Europe. For each of us, the disclosures add one more way to know who, or what, we are dealing with before deciding how much to rely on an answer. A missing label proves little for now: enforcement is uneven, some uses fall outside the rules, and companies outside Europe may not use the same disclosures.

Bottom Line: The label could also start working in reverse. Once disclosure becomes routine, the absence of AI could turn into something a company can sell, much as food labeling helped turn a word like organic into a claim shoppers will pay extra for. If that pattern repeats, ads may soon announce that a support line or a photograph involved no AI at all, with a higher price attached to the promise. A law written to make AI visible could end up making human work the selling point.

California weighs a limit on AI therapy

Read more

Finding those materials involves more than searching for the right words because different kinds of information are organized and made discoverable in different ways. A library catalog contains structured records for books, journals, media, digital resources, and other materials, and it may also link to or include records for archival collections. An archive may organize thousands of records according to the person or institution that created or accumulated them, then describe groups of records rather than every individual document. A government data portal organizes measurements according to categories such as geography or reporting period, while a scholarly database helps us find research distributed across many publications. Each system is designed around different kinds of information and discovery, which means we may need to search each one differently.

That difference becomes easy to miss when many of these systems can be reached through the same general interfaces. A search engine can return a library record next to a government report, while an AI tool can synthesize information drawn from several kinds of sources into a single response. The access feels flattened even though the underlying information is still organized according to very different rules. When a search produces weak results, we may keep changing the keywords even though the larger problem is that we’re looking in the wrong place.

Before deciding what to search, then, we need to know what we’re trying to find. If we want to understand how the freeway route was approved, we need records of the decision-making process. If we want to know how the neighborhood’s population changed, we need measurements that allow us to compare the neighborhood before and after construction. Once we know what kind of information could answer each part of the question, we can ask who would have created or preserved that information and which research system is designed to help us find it.

What we need, then, is a way to see that larger research environment as a whole: the different kinds of places it contains, the roles those places play, and the paths that can carry a question from one to another. An atlas offers a model for doing that. An atlas can bring together different maps of the same territory, allowing us to see features and relationships that no single map can represent on its own.

From that model comes the Atlas of Learning, a framework for mapping the research environment itself.1 Instead of geographic territory, it maps where different kinds of information can be found, how those places and systems organize what they contain, and how we can move among them as a question develops. One part of the Atlas might show where original records are preserved, while another shows the systems through which published research can be discovered. Other parts can include places for structured learning or people whose expertise helps us find our way through systems we don’t yet know.

The routes among those parts are part of the Atlas too. A planning document may give us the formal name of an agency, while a scholarly article may give us a citation to an earlier source. An archival record may introduce the name of an organization whose records are preserved somewhere else. Those names, citations, institutions, and other clues allow one source to generate the next search.

The framework isn’t meant to contain every research resource that exists. It gives us a way to recognize different kinds of resources, understand what each can help us find, and see how one can lead toward another. As we discover new places and learn how to navigate them, we can add those destinations and routes to the Atlas ourselves.

Why we need the Atlas now

Before so much of research converged onto a common screen, the form and setting of a resource often revealed what kind of knowledge we were entering: a card catalog described and located materials, a statistical yearbook gathered institutional measurements, and an archival collection placed us among the records of a person or organization. Those distinctions can become harder to see when a government report, library record, and generated response arrive through the same interface. Many parts of the research world can now be reached from the same screen, but that ease of access can make those parts look more interchangeable than they are.

The Atlas restores that larger view. Instead of presenting every resource as another result arriving through the same interface, it makes the different parts of the research environment visible in relation to one another. We can see which places are built for which kinds of questions and how a discovery in one part of the Atlas can open a route into another. As more of our information environment is encountered through common interfaces or generated responses, that layered view gives us back a sense of the terrain we’re moving through.

The places and systems in the Atlas

The categories that follow are a starting set of destinations and discovery systems for the Atlas. They aren’t an inventory of every research resource in every field or country. They represent recurring functions that appear across many kinds of research: places that preserve original records or publish measurements, systems that help us discover books or articles, resources that teach us the language of a field, and people who can connect us to material we wouldn’t know to seek on our own. Once those functions become familiar, we have a way to approach a resource we haven’t used before: we can ask what it does, what it can help us find, and when it might be a good place to turn.

Catalogs and indexes

• What they are: Catalogs and indexes help us discover sources that may live somewhere else. A library catalog may tell us that a book exists and identify libraries that provide access, while a scholarly index may identify an article whose accessible copy lives in a university repository.

• What they’re good for: These systems give us a description of the source before we reach the source itself. Finding that record has already advanced the research because we now know what we’re looking for and where the material may be available.

• How to use them: Pay attention to the description around the source rather than treating the record as nothing more than a link. An author name may lead to another work, while an established subject term can reveal a broader group of materials. A catalog record can introduce a series title or an organization name we hadn’t known to search.

Archives and original records

• What they are: Archives preserve records created by people and institutions. A city planning office may have produced reports or correspondence over many years, while a community organization may have left meeting records or photographs. An archive often preserves or reconstructs relationships among materials associated with a person or institution because that context helps explain what the records are.

• What they’re good for: Archives belong in the Atlas when we need original records or earlier versions, especially material created close to the event we’re studying. They can take us closer to the records people and institutions created through their own work rather than later accounts of those records.

• How to use them: The way archival material is organized affects how we need to search. An archive may describe folders or boxes without making a separate searchable record for every document. The stronger route may begin with the institution that created the record or the name of the larger collection, then narrow once we understand how the material is arranged.

Data and public records

• What they are: Government agencies and research organizations collect data and records through their work, then publish some of those materials through data portals, reports, or records systems. Published statistics reflect decisions made before the number reached us, including what will be measured and which population will be included.

• What they’re good for: Data systems help us trace a number to its source, compare figures on compatible terms, or see how a measure changed over time. Definitions affect how categories are applied, while reporting periods determine which moments can be compared. By the time a number reaches a headline, those decisions may have disappeared even though they determine what the figure can tell us.

• How to use them: Return to the institution that produced the measurement and look at how the data was defined and collected. A city housing department may define a housing unit or program category differently from a federal statistical agency, while two employment series may cover different populations. Those distinctions affect what kinds of comparisons we can make.

Scholarly databases and research literature

• What they are: Scholarly research is distributed across journals, books, conference proceedings, repositories, and institutions, which makes dedicated databases and indexes valuable for finding work across a field. Those systems can organize papers by author or subject, while citation links show relationships among publications.

• What they’re good for: Instead of seeing one study at a time, we can begin to see the body of research surrounding a question. Scholarly databases help when we want to understand what researchers have studied or how an argument developed over time.

• How to use them: Treat one strong paper as an entrance into a larger conversation. Its reference list shows which earlier studies the authors relied on, while later papers that cite the study can show how researchers responded after publication. A review article can gather many studies around one question and introduce the terminology used across the field.

Guides and reference works

• What they are: Guides and reference works help us orient ourselves when a subject is still unfamiliar. A library research guide may identify the main resources used in a field and point toward relevant collections or institutions. An encyclopedia or subject reference work can supply formal names and established terminology that we wouldn’t have known to search on our own.

• What they’re good for: They help us learn enough about a subject to know what to look for and what language to use. They can introduce the major concepts used to describe a subject, the names associated with it, and the kinds of information that may be available to us.

• How to use them: Look for vocabulary and names that can sharpen the next stage of the research. A formal term can replace the everyday language we started with, while the name of an institution can give us another place to investigate. The goal is to leave with a better understanding of the subject and more informed options for where to go next.

Courses and structured learning

• What they are: Courses often arrange knowledge in sequence, while syllabi, lecture series, and open educational resources can provide a structured path into a subject. They become part of the Atlas when the problem has grown beyond finding one source and we need an organized way to understand the subject.

• What they’re good for: A series of searches can answer separate questions while leaving us without a framework for connecting the answers. Every explanation may introduce another concept we need to look up before the first one makes sense. A course can supply a framework that isolated searches may not.

• How to use them: Use structured learning to understand the relationships among concepts before returning to narrower research questions. A course on urban planning can introduce the ideas that connect transportation policy and neighborhood change before we encounter those ideas separately in reports or scholarly articles. A syllabus can also give us readings and terminology that make later independent research easier to follow.

People and institutions

• What they are: Librarians, archivists, curators, researchers, and other specialists can help us navigate parts of the research world that don’t connect neatly through databases or search engines. Their knowledge can include how collections relate to one another or how an institution describes and organizes its material.

• What they’re good for: A librarian may know which database covers a narrow subject, while an archivist may know that a person’s papers were deposited under the name of an organization. A curator may know that only part of a collection has been digitized, while someone who works in the field may know which professional organization maintains a hard-to-find resource.

• How to use them: The question doesn’t have to be “What’s the answer?” We can ask where to look and which institution is likely to have the material. We can also explain what we’ve already tried, which gives the person helping us enough context to suggest another approach.

Learning what these places and systems do gives us the landmarks, but an Atlas becomes more powerful when we can move among them. A research question can then move across categories as it develops. What we find in one place can change what we need to know, introduce a new destination, or give us language that makes another part of the research world searchable.

There isn’t a prescribed sequence among the different parts of the Atlas. Any one of these places or systems can potentially lead to any of the others, depending on what we find and what the question asks of us next. The routes develop through the research itself. Once we can recognize the different places in the Atlas and understand what each can offer, we can start deciding where a particular question should begin and how to follow the paths that emerge from there.

Finding a route through the Atlas

Knowing what kinds of places and systems exist gives us more options for where research can go. The harder question is deciding where to begin when we know what we want to understand but don’t yet know which resource can help us answer it. We shouldn’t have to know the name of the right archive, database, agency, or collection in advance. Figuring out where to look is part of the research itself.

The sequence below gives us an order of operations for getting started. Follow these steps in order when the research path isn’t yet clear. They’re designed to help us move from a broad question toward a plausible first destination without assuming we already know where the answer lives. Once we begin finding things, we can follow whatever new directions the research opens.

Break the larger question into things we can investigate separately.

Ask what we would need to learn before we could answer the larger question. If we want to know what happened to a neighborhood after a freeway was built through it, we might want to understand how the route was approved and how the neighborhood changed afterward. We may also want residents’ accounts from the period or later research about the consequences. Breaking the question apart reveals the different research tasks contained inside it.

Ask what kind of information would help answer each one.

We don’t need to know the source yet. We only need to describe what we’re looking for. The freeway approval calls for records of how the decision was made. Population change calls for measurements across time. Residents’ experiences call for accounts from people who lived through the project.

Use the Atlas to figure out where that information might live.

Official decisions may point toward public records or archives, while population measurements may point toward a government data source. If the likely destination isn’t obvious, guides, reference works, librarians, or other specialists can help us identify where to look.

Start with the strongest lead and pay attention to what it opens next.

A planning report may reveal the formal name of an agency we didn’t know to search. Meeting records may introduce a neighborhood organization whose records survive somewhere else. A scholarly article may give us terminology that exposes another body of research. What we find at the first destination changes what becomes possible at the next one.

A research route is emergent. We can make an informed choice about where to begin, but we can’t know the entire path before the research produces the names, terminology, institutions, sources, and connections that make later destinations visible. That also means getting redirected isn’t necessarily evidence that we started in the wrong place. A source that sends us somewhere we hadn’t anticipated has given us information we didn’t possess when we chose the first destination. The Atlas helps us recognize those openings and incorporate them into the path as it develops.

When we feel stuck or lost

Research can become disorienting after we’ve been following sources for a while. Sometimes we still know exactly what we’re trying to find, but the route we’re following stops producing anything that helps us get there. At other times, the research opens so many new directions, or changes our understanding of the question enough, that we no longer know which direction to follow. The first experience is being stuck; the second is being lost. They can feel similar because both leave us unsure what to do next, but the source of that uncertainty is different.

Being stuck means we still know what we’re trying to find, but the route we’re following has stopped helping us get there. A collection may contain less than we expected, while a promising citation may lead nowhere accessible. The terminology we’re using may also be a poor match for the way the relevant institution describes its material. In that situation, the question still gives us direction. The problem lies somewhere between that question and the route we’ve chosen to pursue it.

A different way into the same material may reopen the research. The kind of information we’re seeking can point toward another part of the Atlas, while a guide or a person familiar with the subject can reveal terminology or institutions we hadn’t encountered. Knowing that we’re stuck helps us focus on whether another route toward the same question is available.

Being lost means the research has generated more possible directions than we can follow at once. Each path may be legitimate. A planning record might open questions about who influenced the route, while census data raises questions about displacement and newspaper coverage reveals how residents responded. All of those directions could deepen our understanding of what happened to the neighborhood, which is precisely why choosing among them can become disorienting.

What restores that sense of direction is getting clearer about which part of the larger question we’re trying to answer now. If we’re investigating how residents were affected after construction, demographic changes have a more immediate role in that inquiry than the political negotiations that determined the route. The political history remains relevant to the larger subject, but it answers a different part of it. Seeing what each path can help us understand gives us a basis for choosing where to go next.

Being stuck and being lost therefore call for different kinds of reorientation. When we’re stuck, we still know where we want to go, so the next best step is finding another way toward that destination. When we’re lost, too many routes have become possible, so the next best step is identifying which one contributes directly to the question we’re trying to answer now. The Atlas can keep expanding without every path needing to be followed at once.

Drawing an Atlas of our own

An Atlas of Learning can begin whenever there is something we want to understand beyond a quick factual lookup. Each inquiry creates its own terrain because the places we need to visit depend on the question. A health question may take us through medical literature and public health agencies, while a historical question may lead toward archives, newspapers, or census records. The Atlas gives that particular inquiry a visible research environment of its own.

Because the Atlas is organized around a question, that question becomes the compass. It gives the research direction as new possibilities appear, helping us decide whether a newly discovered resource is relevant to what we’re trying to understand and what it might contribute to an answer. And it becomes the point we return to whenever the terrain grows more complicated, helping us recover our bearing and decide what to pursue next.

The working Atlas can be extremely lightweight. A spreadsheet, Notion page, mind map, document, or another information-management tool can all serve the same purpose. At minimum, we need the question, the research destinations that become relevant to answering it, and a brief note about what each can contribute. The form matters less than being able to look at the Atlas and see, at a glance, where this particular inquiry can go.

The Atlas represents the research terrain for a single question. It keeps the destinations connected to that question visible together, giving us a way to see the terrain as it develops and maintain our orientation within it. Something becomes part of the Atlas when it materially advances the inquiry, either by helping us answer the question or by opening a direction we may need to follow next. The result is a record of the sources and destinations that have become consequential to the research, rather than everything we happened to encounter along the way.

A larger world comes into view

Building an Atlas gives us practice locating a question within a larger information environment and recognizing how different parts of that environment can help answer it. We also learn to follow connections as new territory appears and to find our bearings when the direction becomes uncertain. Taken together, those abilities form a kind of information literacy we can apply when we enter subjects we have never explored before.

With that practice, entering an unfamiliar subject becomes an act of orientation as much as discovery. We may begin with little sense of how knowledge about that subject is organized or where its records, research, and expertise reside, but we know that those structures are there to be found. As they come into view, the subject begins to acquire a geography of its own, one that we can learn to read even though we have never traveled through it before.

The Atlas of Learning is something we draw as each question opens a new research terrain around us. The destinations will change from one subject to the next, and the route will develop differently each time. What remains is the ability to look beyond the interface in front of us and see a larger world of places we can learn to navigate.

You might also like:

• What “Do Your Own Research” Actually Means: A step-by-step methodology for evaluating any claim and exploring any subject.

• 27 Ways to Access Scientific Research: A complete guide to finding, reading, and evaluating scientific papers — and knowing what questions matter before you trust the findings.

• In Praise of Serendipity: On the joy of finding what you weren’t looking for, and how to get lost on purpose.

The free essays are the foundation. The paid tier is where information literacy becomes a skill you can use: the Card Catalog Classroom with its searchable Reference Collection and structured lessons across four learning domains, plus biweekly AI Briefings, monthly Librarian Hotlines, and quarterly deep-dive guides.

Upgrade to paid for the full Card Catalog. Thank you for being here!

Subscribe now

Have you read the Founding Member Report: The State of AI yet?

A comprehensive guide for information navigators who want to understand where AI is actually heading and what it means for how we find, evaluate, and use information in 2026.

→ Find out more here.

The impulse behind these archives isn’t new. In the first century AD, Seneca complained about Romans who amassed more scrolls than they could digest, warning that a wandering appetite for books left the mind scattered instead of nourished. A Meiji-era Japanese word, “tsundoku,” names the practice of acquiring books that pile up unread, a behavior common enough to earn its own word. The wish to keep more than we can absorb is ancient, and for most of history the collecting slowed itself: a new book cost money and claimed a place on a shelf we could watch fill up. Those prices made us choosy, so we brought home only what we expected to read, and the choosing was itself the curation: every book on the shelf had already survived a decision about whether it belonged there.

What’s changed is that nothing slows the collecting anymore. Saving costs nothing, takes no space we can see, and happens in the time it takes to move a thumb—so scale became unlimited at the same moment the friction that limited it disappeared. Nothing about the resulting pile reveals a new weakness in us. The old wish is simply operating in an environment engineered to strip away every limit that once contained the impulse.

The save button began with a promise. Read-later apps like Pocket and Instapaper appeared in the late 2000s offering a trade: tap save now, and come back to the piece when time allows. The button then spread far past the apps, to almost every content surface, from the bookmarks bar to the Notion page. As it spread, the design incentive shifted from serving the reading to serving the engagement metric, where each tap counts whether or not it leads anywhere. To us, the tap still feels like a plan: this piece, but later. The platform counting that tap reads it as engagement, one more sign the feed kept us interested, whether or not we ever actually return to the piece. The tap has become its own emotional payoff, and the queue behind it is where reading goes to be indefinitely postponed. Every part of the pipeline stays functional except the reading itself, which was the pipeline’s original purpose.

Living with the unread

The archive that grows this way is a record of single moments of attention with nothing connecting them: a tax explainer beside a poem beside a friend’s apartment listing, each saved on a different day for a different reason. Every save is a small message to a future version of ourselves, an ideal reader with the time and attention to sit with what we’ve caught. But the trouble is that the recipient keeps changing. The queue accumulates work addressed to a person who was about to become a parent, or start a business that never happened, and that reader never arrives to collect. Meanwhile each new save asks less of us than the one before, because past a certain size the pile can no longer be seen in one view, and adding to what can’t be seen carries no visible cost.

Left alone, the arrangement settles into a strange steady state. We keep encountering writing all day and keep feeling like readers, while the actual reading thins to headlines and opening lines, because the queue absorbs everything that would have taken longer. Opinions start resting on arguments we secured and never absorbed, and the sense of being informed floats free of anything we could explain to a friend across a table. The queue itself turns from a tool into a mountain to climb, since opening the bookmarks folder means facing every promise inside. So we stop opening the folder, and the place we built to keep what we valued becomes the place we avoid.

There’s a reason the save feels satisfying in a way that reading in the moment doesn’t, and it starts with what the brain does when it expects future access to information. In 2011, the psychologist Betsy Sparrow and her colleagues published experiments in Science that isolated the effect. Participants typed facts into a computer; half believed the file would be saved, half believed it would be erased, and nothing else varied between them. Those who believed the computer would keep the information remembered the information itself far less well, and remembered instead where it had been stored. The brain, sensing that an external store would keep the content, shifted toward remembering how to retrieve the information rather than the information itself.

That shift in memory, from the content to the route back to the content, is where the save’s satisfaction comes from. An unread piece we want is a small open demand, a claim on attention we can’t pay at the moment we meet it, but the tap to save closes the demand immediately: the piece is captured and safely stored, and nothing is pulling at us anymore. Reading would earn the same sense of completion only at the end of the piece; the tap collects it early and for free, before any reading has happened. Each save can deepen the habit, and the habit leaves a residue: an archive full of pieces we’ve never read but vaguely feel we know, because a filed address masquerades as familiarity with what the piece says.

Before the save button

Saving didn’t always outrun reading, because saving used to be labor. We folded down a magazine corner, tore an article out for a folder, printed pages at work to carry home, filed a browser bookmark into a folder we’d named ourselves. That labor did the curating: what made it into the folder had already earned a place there, and the folder stayed small enough that browsing it was a way of re-reading the collection. Each of those small actions also forced a moment of contact with the piece, a glance at what it argued and a judgment about whether it earned the effort. Saving started standing in for reading only when the effort disappeared and the two began to feel the same.

The disappearance came by degrees. Social bookmarking services had come and gone since the 1990s, but by the time Twitter added bookmarks in 2018 and every platform followed, the save button was a fixture on every surface with no shared sense of what saving was for. The evolution from manual to frictionless saving has transformed the act from a curated, high-effort task into an automatic, low-effort behavior, eliminating the cognitive appraisal required for content consumption. Deciding what to save was once a small act of judgment about whether a piece belonged in our lives, and the friction was what forced the judgment. Saving is now one tap that asks nothing, so when the friction disappeared, the judgment disappeared with it.

Working through the pile

Every item in the pile entered because something in it mattered to us at the moment of saving. Then the moment moved: the dinner that called for the recipe got rescheduled, the commute that would have carried the essay turned into a phone call. The pile is what those changed plans left behind, so its size simply measures how often life shifted. That’s also why actually working through it helps: a sorted pile returns the queue to service, and opening the bookmarks folder stops feeling like facing a cliff and starts feeling like browsing a shelf we intentionally chose. What remains is the practical question of what to do with the pile now; two workable paths answer that question, one fast and one more deliberate.

One distinction makes the work lighter before either path begins: some of what we save is an archive, and some of it is a queue. An archive is reference material, the tax resource or the recipe collection kept in case we want it someday; nothing in it is waiting to be read immediately, and a search bar is all it takes to find a piece again when a question comes up. A queue is promised reading, the articles and essays we told ourselves we’d get to; everything in it is waiting to be read, and time is all it takes for the waiting to turn into guilt. Most of the pile's weight comes from the two blurring, when reference material sits in the reading queue gathering guilt and reading promises sink into an archive where no one will meet them. Sorting out the blur starts with one question per item: is this here for looking up someday, or did I mean to read it as soon as possible? With that question in hand, the clearing itself can begin, and it can go one of two ways.

Path 1: The fresh start

Moving the entire existing archive into a single dated folder, out of sight but not deleted, lets us begin again with an empty queue: bookmarks drag into a folder named “Saved before 2026,” a Notion archive page gathers what came before, the screenshots slide into one album. The move borrows from email, where people whose inboxes grew past answering have archived everything at once and started over, a practice its survivors call declaring email bankruptcy. It works for the same reason here: nothing is destroyed and anything needed later can still be found by search, so the ambient weight of the backlog lifts immediately. The old saves become what archives have always been, a record kept for reference, ready on the rare day a question sends us back to look.

Path 2: The sorting pass

For those of us who’d rather look at what we’ve kept, everything runs on a single question, asked of each item: would I save this again today? A yes means the piece still belongs to the person we are now, and the follow-up is to read it soon or attach it to the work it serves. When the answer is no, the piece belonged to someone we’ve since stopped being, and releasing it costs nothing we still have. The question even settles the ambiguous cases: a three-year-old bookmark on Slack’s notification design goes, since newer writing exists and the curiosity that saved it has moved on, while an old Pinterest board from a finished renovation keeps its few strong images and sheds the rest. What survives stays small and current, which is what makes the remaining resources readable at all.

The fear of losing knowledge

Releasing an item retires a promise made by an earlier version of us who thought they’d come back and read the piece. Letting go can feel like losing knowledge, and the feeling traces back to the same mechanism: because the brain filed the location as a form of knowing, deleting the location reads as forgetting something we felt we had. The feeling misleads us, though. Since the reading never happened, the knowledge never made it into us; deleting the bookmark removes a route to information we hadn’t absorbed anyway. Remember: a piece we end up wanting later can be found again through search. Letting the item go releases that earlier self from the commitment, and the archive that remains starts to reflect what we’re paying attention to now.

A field guide to the moment of saving

Either path ends the same way: with a cleared pile. But a cleared pile refills unless the act of saving itself changes. The reflex fires the same way every time, but the situations it fires in keep changing: short on time, unable to focus, in the middle of work, deep in a feed. Each one calls for a different way of handling the piece we’re about to save. A single generic rule can’t reach across all of them, which is why “just have better judgment” reads as very unhelpful advice. Every fix below can run on effort in the moment, or it can be built into our tools so it runs on its own.

When we’re short on time but want to read this

The piece looks like a good use of time we don’t have right now, and the save feels like a fair compromise between wanting to absorb the piece and being unable to read it yet.

What helps

Reading the first paragraph before saving settles most of these. The opening often delivers the thing the headline promised, and the curiosity resolves on the spot; when it doesn't, the piece has earned its save, and the save gets a note about when the reading will happen, "for the train home" or "for the research on Thursday." A save with a schedule attached is a save that gets opened. Most tools have a field ready to keep it, since bookmarks take a description and a screenshot can be renamed.

Why this works

The bare save fails because the tap delivers the completion feeling before any reading happens, so the curiosity feels handled and moves on. Reading the first paragraph returns the piece's substance to the exchange, which is the difference between a save that rejoins the pile and a save that keeps its appointment.

When we can’t concentrate right now

The desire to read is real, but our attention is scattered. The save becomes a promise to give the piece the focused reading we don’t have in us at the moment.

What helps

A save made in a scattered state defers the same scattered state, since the future self who opens the queue inherits whatever attention we have then. If the piece belongs in our reading, it goes where focused reading happens for us, which for many of us is a particular place and ritual: the morning coffee before the phone, or the printout by the reading chair. Attaching the piece to that setting, and not to a general pile, gives the reading its best chance. If it doesn't rise to that, letting it go costs nothing, because a piece we couldn't concentrate on today was never going to fare better inside a queue of four hundred.

Why this works

Scattered attention is where the pile grows fastest, because a mind that can't settle saves without discriminating, and each save still feels like progress even though a scattered evening produces no reading to back the feeling up. Sorting these saves toward a reading setting, or letting them go outright, keeps the low-attention hours from building the archive we'll resent later.

When we’re working and find an article we like

Mid-task, an article surfaces that looks promising or maybe relevant to the job, and saving it feels like capturing something without breaking focus.

What helps

One question sorts these: what is this piece for? A piece that helps the current task gets used immediately, its one relevant point typed into the document we're already working in, and then nothing has to join any pile. Anything meant for a future project goes into that project's folder, where we'll see it when we next open the project. And whatever fits neither is a distraction, however promising it looks; closing the tab returns us to the work.

Why this works

A work-adjacent save is the hardest kind to refuse because it wears the costume of productivity while doing the opposite. The reach for the save button is a micro-break that fragments the attention the work needed, and the tap rewards the break with a small hit of accomplishment unrelated to either the task or the article. Asking the question takes the costume off: a piece with a job keeps its place in the work, and a piece without one stops looking like work at all.

When a roundup or a feed keeps offering more

A newsletter arrives with ten links, or the Instagram and Substack feeds stack recommendation on recommendation, and the saves come in runs: half a roundup banked in a minute, a dozen bookmarks from one night of scrolling.

What helps

Reading one link and leaving the rest settles a roundup. The email isn't going anywhere: it sits in the inbox, searchable by sender, and the other nine links will still be there on the day we want one of them, so saving them one by one just copies the newsletter into our pile. On Instagram and Substack, the in-app save is where posts go to be forgotten, so a recommendation that stays with us goes into the notes app or gets a when attached, and everything else scrolls on by.

Why this works

Feeds are built to surface more appealing pieces than anyone could read, so the sense of falling behind is part of the design. Saving relieves that feeling on the spot, which is why the saves come in runs: each one lets us pass a good piece without the small ache of losing it, and the scrolling continues. Reading one piece per session, or none, steps out of the arrangement, since the feed's output was never meant to be kept up with.

When a friend sends us something

A friend texts us an article with “you have to read this.” Saving it feels like doing something with the recommendation, a way of taking it seriously before we’ve read a word.

What helps

The friend wanted the reading and the conversation. Reading it now serves both when the piece is short; when it isn't, telling them when we'll get to it keeps the exchange alive, and a reply about the one line that struck us counts for more than a silent save ever could. Each of these keeps faith with the sharing in a way the untouched bookmark does not.

Why this works

A save is tempting here because it discharges the feeling of social obligation without the reading the obligation was about: the piece goes into the queue and the debt feels paid, while the friend hears nothing back. These saves pile up faster than most, because a link from a friend is harder to dismiss than one from a feed. But when saving becomes our standing response to them, the recommendations collect unread, and the conversations they were meant to start never happen.

When we save anyway

This is the case the others don’t cover. We can understand everything above and still watch our thumb reach for the save, because the reflex runs on the relief of the moment and never consults our track record. And the track record says most of these saves won’t be opened again. The fixes above all ask for a choice at the moment of saving, about what the piece is for and where it goes; here, no choice is coming.

What helps

The fix stops depending on the save entirely, and builds a return to the source. A save vanishes only when nothing ever brings it back, so the outcome turns on scheduling that return: a standing appointment, on the calendar, when the queue gets opened on purpose. What lasts is small and repeatable, something like one evening where we open the queue to read a single piece and release ten more without guilt. The appointment puts a floor of actual reading under the pile and keeps the pile small enough to stay in view, and what we save between appointments counts for far less once the appointment exists. Here, saving was never the problem; it was missing the return to what was saved.

Why this works

A habit doesn't stop because we understand it, and research on habit formation suggests why: behavior follows cues and repetition more than intention. An appointment moves the point of intervention off the save, where willpower loses, and onto a recurring block of time we set once and uphold like any other standing commitment. The pile stops being a place things disappear into and becomes a place we visit; a queue we visit is a queue that gets read.

After the pile

Reading starts happening again. The save marked "for the train home" gets opened on the train, the standing appointment works through the rest, and the reader who was always about to read actually starts reading. With the reading back, the background guilt thins out, since a queue we visit on schedule stops feeling like a debt. What the save button had been spending, without our noticing, was attention kept on one piece long enough for the piece to change us. That expense stays invisible at each tap, so the gap between what we saved and what we absorbed goes unnoticed until something forces a full look, a dying app demanding we export our saves, or a folder finally too full to ignore.

Evaluating a source has long stood at the center of information literacy, and the evaluation runs on two questions: who wrote this, and what do they want us to believe. Those questions have always been about reading a source’s intentions, which is why they work just as well one layer down, on the save button itself: who built this button, and what do they want from us? The answer isn’t hard to find: social platforms make money from the time we spend scrolling, and reading a long article means leaving the scroll, so their save buttons are built to file the piece in one tap and return us to the feed. Asking those questions gets us the same thing in both cases: a moment to decide for ourselves, instead of doing whatever the button was built to make easy.

Seeing the mechanism is what makes it possible to interrupt. The save button will keep appearing on every surface, engineered to be pressed, and the pull toward the tap will keep arriving with its small false sense of completion. All of it can keep happening while the reading comes back, because the change never depended on resisting the tap, where the reflex wins. The change comes from the return, the standing time when the queue opens and something in it gets read, so that saving stops being the place things disappear. Our archive was never a failure of discipline. It records a system working exactly as designed, and seeing the design is the first thing that has ever let us step outside its logic.

You might also like:

• A 30-Day Protocol to Rebuild Reading Capacity: Progressive overload applied to the page, for readers who miss reading.

• How One Rescue Mission Turned into the Global Record of Human Heritage: Inside the preservation system that classifies, catalogs, and protects humanity’s cultural and natural inheritance.

• The Libraries That Weren’t Supposed to Exist: Basement libraries and buried archives: a brief history of reading against the state.

The free essays are the foundation. The paid tier is where information literacy becomes a skill you can use: the Card Catalog Classroom with its searchable Reference Collection and structured lessons across four learning domains, plus biweekly AI Briefings, monthly Librarian Hotlines, and quarterly deep-dive guides.

Upgrade to paid for the full Card Catalog. I appreciate you being here!

Subscribe now

Have you read the Founding Member Report: The State of AI yet?

A comprehensive guide for information navigators who want to understand where AI is actually heading and what it means for how we find, evaluate, and use information in 2026.

→ Find out more here.

Share Card Catalog