... The ballet of the good city sidewalk never repeats itself from place to place, and in any one place is always replete with new improvisations... - Jane Jacobs
Serendipity is a strange interruption, an idea arriving unannounced, a chance encounter with an old friend in a new city. We describe these moments as luck, sometimes grace, but they rarely emerge from nothing. More often, they are visible traces of something deeper: systems, designed or accidental, that quietly raise the odds.
In cities, dense neighborhoods compress distance, mixed-use streets layer different kinds of activity, and short blocks multiply the number of paths people can take. A well-placed bench or a lively corner extends how long people linger, increasing the chance of an unexpected exchange. We become participants in a structured improvisation, what Jane Jacobs called the sidewalk ballet.
But cities do more than simply maximize encounters, they preserve the conditions for wandering. Parks and meandering streets resist pure efficiency. They invite detours and create space to wander. A perfectly optimized city, one that routed everyone along the fastest possible path, would reduce friction, but also eliminate the idle moments in which serendipity tends to emerge.
The city, in this sense, is a kind of probability engine: one that doesn’t dictate who meets whom, but quietly shapes the odds that they might.
If cities organize encounters in physical space, recommender systems shape them in digital space.
A suggestion surfaces, a post lingers, a passing mention takes hold. These signals do not deliver us anywhere in particular. They accumulate and tilt decisions at the margins -- what neighborhood to wander into, or who to reach out to after a long silence. The effect is less like being routed along a path than like being gently nudged, again and again, until certain trajectories become more likely than others.
Before we moved to New York, we visited two of our friends in Brooklyn and found a nice restaurant to try. By seemingly pure chance, our other friends Philip and Emma happened to be there. I couldn't help but wonder what fleeting post they saw that made them want to try that place that evening.
Later that same night, I was at the same friend's apartment and Philip shared his location with me -- he happened to be at a bar across the street (again, seemingly pure chance). I walked over and as I entered, he turned around, grinned, and said "Share my Location is like a social network!" He was half-joking, but he was also describing an entire category of software that didn't exist yet.
What appeared initially as coincidence began to look more like composition. Not a single system producing an outcome, but several operating at once. The restaurant was not chosen in isolation, nor was the encounter entirely accidental. It emerged from a stack: conversations, recommendations, ambient signals, shared contexts, and finally, a thin layer of explicit coordination. Each on its own was insufficient. Together, they made the meeting feel inevitable in retrospect.
Our coordinates are no longer purely physical. They are triangulated -- part geography, part algorithm, part the ambient signal of other people's presence. Two people arriving at the same place at the same time may look like chance, but the convergence usually has a longer tail than either of them can see.
Serendipity, in this sense, is not defined by the absence of structure, but the presence of multiple structures intersecting. The city is the stage, dense enough for paths to cross. Physical and digital networks circulate signals that shape intent before movement begins. And increasingly, a third layer - lightweight, ambient coordination - allows those paths to bend toward one another in real time.
There is an older version of this third layer. Before platforms formalized it, it lived in a more diffuse form, less system than behavior. People told each other where they had been, where they were going, who they had seen, what others were thinking. Not as coordination in any strict sense, but as a kind of social weather report: incomplete, informal, gossip-adjacent, but unusually good at shaping where people ended up.
Gossip, at its best, wasn't just about information, but a way of routing presence -- of shaping who ended up near whom. Though it historically often carried cruelty, when done well, it carried context and care. It worked because the village was small enough that both could travel together.
What it could not do was scale. Past a certain size the context thinned, the care dropped out, and we replaced the whole apparatus with coarser machinery: broadcast, feeds, rankings. Each of these scaled reach at the expense of the thing gossip quietly did well.
I wonder whether the next layer of the stack might recover this kind of care and attention? Agents that know us well enough to hold our context, and are oriented enough toward our people to carry care between us. Not a system that routes us along the fastest path to our friends, but one that notices when a friend is two blocks away and nudges, the way a neighbor once might have, in passing, over a fence.
The city taught us that serendipity is a property of well-designed spaces. Recommender systems and social networks taught us it is also a property of well-circulated signal. What comes next may teach us it is a property of caring at scale -- the oldest technology we have, finally able to travel.
To subscribe to the newsletter, fill out your email at the bottom of the page.