serendipity

Two words for two kinds of discovery

Serendipity was coined by Horace Walpole in 1754, after a Persian fairy tale, The Three Princes of Serendip (Serendip being an old name for Sri Lanka), whose heroes kept stumbling onto things they weren’t looking for. The part of his definition that usually gets dropped is the important one: serendipity isn’t only luck. It’s discovery by accident and sagacity, the wit to notice that the accident matters. Fleming’s contaminated petri dish was the accident; recognising it as penicillin was the sagacity. Without the second half you don’t have a breakthrough, just a dirty dish you throw away.

Zemblanity, its opposite, is a much younger word, invented by the novelist William Boyd in Armadillo (1998). Serendip is warm and lush, so Boyd reached for somewhere bleak: Novaya Zemlya, a frozen archipelago in the Arctic. Zemblanity is the knack of making unhappy, expected discoveries by design. You get it when a system is built so that nothing surprising can happen, and the dispiriting, predictable result turns up right on schedule.

The zemblanity engine

Most AI and algorithmic systems are, by default, zemblanity engines. Feeds, recommenders and personalisation models are built to predict what you already want and hand you more of it, which is almost the definition of an expected discovery by design. Optimisation narrows. It strips out the noise and the wrong turns that serendipity feeds on: the misshelved library book, the tangent in a conversation, the search that returns something better than what you asked for.

The same technology can work the other way, and that’s where it gets interesting. Left to run on its own, an optimised system drifts into zemblanity, not out of malice but by doing its job a bit too well. Pointed the other way, AI can be a remarkable engine for serendipity, throwing together ideas from fields a specialist would never see side by side. Which version you get isn’t an accident of the technology. It depends on how the system is built and how it’s pointed, and the easy path, the one every default nudges you down, leads to the confirmation machine. The exploration machine has to be a deliberate choice.

The abundance problem: finding signal in the noise

This is the tension sharpening fastest. AI has made the raw material of discovery, the accident, almost free. Content now arrives faster than anyone could read it, and most of it is fluent, plausible and instantly forgettable. Abundance isn’t the same as richness: when anything is cheap to produce, the average quality of what crosses your path falls even as the quantity explodes.

That flips the problem on its head. For most of history the hard step was finding things at all, getting hold of the book or reaching the expert. That cost is collapsing. The new bottleneck is discernment, knowing which of ten thousand fluent paragraphs is worth keeping. What’s scarce is no longer access. It’s judgement, the prepared mind that can tell which surprise actually counts.

Here’s the uncomfortable part. That prepared mind is necessary but, at this scale, no longer sufficient. No amount of curiosity lets one person read an ocean. Separating signal from noise has outgrown what willpower and good habits can manage by hand, and increasingly has to be helped along by the right instruments. Worse, the instinctive response to overload makes things worse: people reach for heavier filters just to cope, and those filters narrow the world back down to the predictable. More noise leads to more filtering leads to a smaller, more frozen world. You reach the Arctic archipelago by sailing across an ocean of content.

What to do about it

Serendipity now has to be designed for rather than assumed. Worthwhile systems make room for exploration on purpose, keeping some diversity and friction instead of optimising every last drop of surprise away. Curation becomes the high-value work: when generation is free, the real effort is taste, editing, and a willingness to bin most of what you produce. AI is best used to widen first and narrow second, roaming across domains for genuine surprise while human judgement decides what to keep.

The harder truth is that none of this happens by default. It starts with a serendipity mindset, the habit of moving through the world with your eyes open and ready to act on what you find; without that, no system in the world will help you. But a prepared mind does its best work when something is built around it. Against a tide of machine-made content, even the sharpest attention needs leverage, the right structures and support to extend its reach. The people and organisations who keep finding the good, surprising, useful thing are the ones who pair that mindset with deliberate design, treating serendipity as something to cultivate and equip rather than simply wait around to feel. As the noise rises, mindset and system lean on each other more, not less.

Further reading

Marco Balzano, a management researcher at the University of Trieste, is the author of a widely cited survey on serendipity, “Serendipity in Management Studies: A Literature Review and Future Research Directions” (Management Decision, 2022). This maps how the field has studied serendipity and his broader body of work extends it into a fuller framework.