In Praise of Serendipity
On the joy of finding what you weren't looking for, and how to get lost on purpose.
In 1754, Horace Walpole coined the word “serendipity” in a letter to a friend, borrowing it from a Persian fairy tale about three princes who made discoveries “by accidents and sagacity.” He meant the wandering that puts us somewhere unplanned, and the readiness to recognize what we'd stumbled into when we got there. Some things arrive sideways and turn out to matter more than what we came for, and he thought that deserved a word.
We still have the word. What we’ve lost is the wandering.
The systems we use to find things now are built almost entirely around intent. We arrive with a question, leave with an answer, and the whole apparatus is engineered to make that transaction as fast and frictionless as possible. Lost in that transaction is the detour that becomes the discovery, the neighboring idea that reframes the question we arrived with, and the thing that couldn’t have been searched for because we didn’t yet know it existed.
What serendipity actually is
Serendipity requires two things at once: the conditions that make an unexpected encounter possible, and the prepared mind that knows what to do when it arrives. Walpole understood this, which is why his definition included both accidents and sagacity. The find matters, but so does the person who recognizes it as a find.
In information terms, serendipity is the discovery of something relevant, useful, or formative that we weren’t looking for and couldn’t have searched for, because we didn’t yet know it existed. It’s the book pulled from the shelf because the spine caught the light that turns out to reframe a subject we thought we understood. The footnote followed on a whim that opens into a field we didn’t know was adjacent to ours. The article that arrives sideways and answers a question we’d been carrying without knowing we had it.
What distinguishes these finds from ordinary discovery isn’t just surprise. It’s the specific sensation of recognition: the feeling that this thing was waiting and that we became capable of finding it only by moving through everything that came before it. That’s what happens when a prepared mind encounters the unexpected in an environment designed to make such encounters possible.
On the joy of finding what you weren’t looking for
Some finds arrive in the middle of looking for something else, in a library aisle we had no particular reason to be in, and even on a page we turned to without expectation. The world offers something we didn’t order, didn’t anticipate, and couldn’t have requested - something that turns out to be exactly what we needed before we knew we needed it.
This kind of find produces a different feeling than successful retrieval. When we search for something and find it, we feel efficient. When we stumble onto something we weren’t looking for and recognize its significance, we feel met. There’s a quality of gift to it, of the world participating in our thinking in a way that purely intentional search never produces.
The accidental find tends to attach itself to our thinking differently than the retrieved result, connecting to knowledge we didn’t expect it to touch, surfacing later in contexts we couldn’t have predicted. We carry these finds in a different way. They become part of how we think rather than something we looked up once and moved on from.
That’s the joy of it: not just the surprise, but what the surprise makes possible. The question we didn’t know we had. The connection we couldn’t have planned. The afternoon that went somewhere we didn’t expect and came back with something we’ll carry for years.
The architecture of accidental discovery
Marcia Bates, one of the foundational theorists in information behavior, developed what she called the berrypicking model in 1989. Her argument was deceptively simple: real research doesn’t move in a straight line toward a predetermined answer. It wanders and reconstitutes itself around discoveries that reshape the original question. The question we start with can shift considerably as the search develops. The unexpected discovery functions as a structural feature of how understanding develops, not a distraction from the process.
Bates was writing at the dawn of digital search, when information systems were being designed around a single assumption: that research was a retrieval problem, a matter of matching queries to answers. What she was documenting was something more generative and less controllable: the way real intellectual work moves through a landscape rather than toward a destination.
The spatial arrangement of physical libraries turned out to support exactly this kind of movement. Dewey and the Library of Congress classification systems organize knowledge so that related subjects live next to each other on the shelf, and walking to find one book means walking through an entire neighborhood of related thought. The encounters that happen along the way aren't accidents of shelving. They’re a consequence of how the system is organized, and researchers have documented them as a distinct and valuable mode of discovery that digital search doesn’t replicate.
Researchers Carol Kuhlthau and Ross Todd found that the experience of searching shapes what we take from it, separate from the outcome itself. Stumbling onto something produces different cognitive engagement than retrieving a targeted result, and those accidental finds tend to attach themselves to our thinking in unexpected places, connecting to knowledge we didn't expect them to touch. The journey shapes the learning in ways that can't be separated out after the fact.
Search was built to eliminate the space between question and answer, but that space is where serendipity lives. The issue isn’t that digital search performs poorly at unexpected discovery; it’s that unexpected discovery was never part of what it was designed to do. Precision and serendipity require different architectures, and the dominant information tools we rely on were built for precision.
What our tools can and can’t do
Many algorithmic feeds are optimized for engagement with what we already respond to, which means they systematically narrow our exposure to the unfamiliar. The feed learns us, and then reflects us back to ourselves, progressively smoothing away the edges that might have snagged on something new. Search returns what we asked for, bound entirely by our ability to articulate the question, which means we can only find what we already know to look for. Books we don’t know exist are unlikely to surface in our results, and the field we didn’t know was adjacent to ours stays invisible.
AI extends this logic to its furthest point. A well-formed AI response arrives optimized for apparent intent, synthesized for clarity, stripped of neighboring ideas and tangential angles. There’s no browse mode, no wrong aisle, no stray afternoon when we went looking for one thing and came home with another. The answer arrives without the journey, and the journey is where the unexpected lives.
The ability to find exactly what we need has changed what’s possible in research and daily life. But a tool optimized for answering questions can’t tell us which questions we forgot to ask, and it has no mechanism for discovery that happens before something is searchable. As our information lives migrate further toward intent-based retrieval, the conditions for accidental discovery erode unless we rebuild them ourselves.
How to get lost on purpose
Serendipity can’t be summoned directly, but the conditions that produce it can be rebuilt. The practices below are drawn from library science and information behavior research, and they work because they create room for encounter rather than retrieval.
Follow the footnotes.
The citation trail is the original hyperlink, and still one of the most generative paths to unexpected discovery. When a book or article references something unfamiliar, following it mid-read, then following the citations in that source, moves us through a conversation that's been unfolding for decades and leads somewhere the original text never would have.
Browse a physical shelf.
Walking slowly through a section of a library or bookstore without anything specific in mind puts us in a different relationship with information than a search box does. Dewey and the Library of Congress organize knowledge spatially, so the shelf we're standing in front of holds not just what we came for but everything that surrounds it. We came for one thing and left having found something we didn't know existed.
Read the bibliography.
The bibliography of a book that shifted our thinking is a map of the intellectual territory its author traveled. Those sources are a curated list assembled by someone whose judgment we already trust, and following them into their own work opens up a dimension of a subject that keyword search rarely reaches.
Replace algorithmic feeds with RSS.
An RSS reader aggregates content from sources we’ve chosen without ranking or personalizing the results. Everything arrives in chronological order, unmediated, from outlets and writers we deliberately subscribed to. The selection belongs to us rather than to a model predicting what will hold our attention, and nothing is withheld because an engagement engine decided we wouldn’t click.
On staying open to the unexpected
We teach search as technique: how to construct a query, which databases to use, how to evaluate a source. What we don’t teach is the limit built into all of those tools: they can only return what we already know to ask for. Everything we don’t know to ask for requires a different approach entirely.
AI extends that limit further. A language model synthesizes from what we’ve already thought to ask about, which means it’s only as generative as the questions we bring to it. Some of the most generative questions arrive through wandering, through the accidental encounter that reframes what we thought we were looking for, and that’s something current AI tools aren’t designed to do.
The library tradition has long understood that knowledge develops through exposure to the unexpected and through the slow accumulation of encounters that couldn't have been planned. Following a thread without knowing where it leads is how the connections start to form on their own. The practices above aren’t a romantic supplement to serious research. They’re part of how serious research has always developed.
Serendipity has a research literature and a documented role in how knowledge develops. But the reason any of that matters is simpler: the world occasionally offers something we didn't know we needed, arriving from a direction we never would have thought to look. We stumble into it less than we used to, but it's still there when we make room for it.
I'm reminded of a comment made years ago by Eric Hoffer about how he does research. If he is in, say San Francisco and is looking for a particular person but doesn't know their address, he didn't go to the yellow pages or whatever. "This is not my way!" he said. "My way is to stand on the corner of Powell and Market and wait for him to come by. And if you have all the time in the world and you are interested in the passing scene, this is as good a way as any; and if you don't met him, you are going to meet someone else. That is how I do research."
Your namesake, the card catalog, was an excellent example of serendipitous searches. Organized alphabetically by either title or author, one was guaranteed to find things that one was not looking for. That feeling of surprise and satisfaction is lost with digital searches.