Engineering Luck

This essay was written by AI, but I thought it was good enough to publish.

Most people treat luck as weather. It arrives or it doesn't. You can't make it rain.

This is wrong, or at least incomplete. Luck has a surface area, and surface area can be engineered.

Marc Andreessen once distinguished four kinds of luck. The first is blind luck — the meteor lands on your house or it doesn't. The second comes from hustle; you move fast enough that you collide with opportunity others miss. The third comes from preparation — you've studied the domain deeply enough to recognize a lucky break that looks like noise to everyone else. The fourth is the strangest: your reputation and character attract specific opportunities to you, opportunities that couldn't have found anyone else.

Only the first kind is truly outside your control. The other three are built.

Consider how many important outcomes trace back to something that felt like coincidence at the time. You sat next to someone at a conference. A friend forwarded an email. You read a paper on a topic adjacent to your own and noticed a connection the authors missed. These events feel random. But the probability of each one was a function of choices you'd already made: which conference you attended, how many friends knew what you were working on, how broadly you read.

The math here is simple and underappreciated. If each "lucky" encounter has a 1% chance of changing your trajectory, and you engineer your life to have 500 such encounters instead of 50, you are not 10x luckier. You are almost certainly lucky at least once. The person who had 50 encounters had a 40% chance of never getting lucky at all.

This is why the advice to "put yourself out there" is more mechanistic than it sounds. It is not motivational fluff. It is a statement about probability distributions.

Someone recently asked me, before an event, what the point was. "Is it just networking, or what's the end goal?" The honest answer is: I don't know. That's the whole point. You cannot pre-compute which encounter will matter because the value of a connection depends on future states of the world that neither of you can predict yet. The person you meet over bad coffee in March becomes the person who introduces you to your cofounder in November — but only if both of you showed up in March with no particular reason to. Peter Thiel and Max Levchin met at a random talk at Stanford. Neither went looking for the other. The expected value of attending that talk, computed in advance, was close to zero.

The right framing is not "what do I gain by going?" It is "what might I foreclose by not going?" The cost of attendance is a few hours. The cost of absence is unknowable, which is exactly what makes it dangerous. You cannot miss what you never encountered.

There is a pattern in the Bay Area that illustrates this well. The people who end up working together on the most important things almost always met each other in their first six months after arriving. Not because those months are magic, but because that is when people are most open, most hungry, most willing to go to the thing they have no reason to attend. They say yes to everything. They have no filter yet, and the absence of a filter is precisely what lets the improbable encounters happen. The people who arrive later, or who arrive and immediately optimize — who attend only the "right" events with the "right" people — end up in a smaller, more predictable world. They meet exactly who they expected to meet. Which means they learn exactly what they already knew.

The decay is gradual and invisible. After six months you have a routine. After a year you have a social circle. After two years you have a network, and networks are self-reinforcing: they connect you to more of the same. The window where you are maximally exposed to the unlike, the unexpected, the not-yet-legible — that window is brief, and most people close it without realizing it was open.

There are specific, unglamorous ways to increase the surface area:

Write in public. A blog post, a technical write-up, an open-source contribution. Each one is a beacon that runs 24 hours a day, surfacing your thinking to people you've never met. Paul Graham has said that writing doesn't just communicate ideas; it generates them. The compounding works in both directions.
Tell people what you're working on. Not in a self-promotional way. Just factually. Most people have no idea what their friends and colleagues are actually doing day to day. Information asymmetry is the default state. Breaking it is free.
Work on things that compound. Skills compound. Relationships compound. Reputation compounds. A year spent building something that teaches you nothing and introduces you to no one is a year with near-zero expected luck, regardless of the outcome.
Stay in the game. Survivorship is the prerequisite for all other forms of luck. The people who seem disproportionately lucky are often just the ones who didn't quit during the long middle stretch where nothing appeared to be working.

There is a subtler point here about preparation. Pasteur's "chance favors the prepared mind" is usually quoted as encouragement to study hard. But the real insight is about pattern recognition. When Alexander Fleming noticed the mold killing bacteria in his petri dish, dozens of other researchers had seen the same thing and thrown the dish away. Fleming's luck was not the mold. His luck was the years of prior work that made the mold legible to him.

This means depth matters as much as breadth. The person who dabbles in everything has high encounter rate but low recognition rate. The person who goes deep in one area has high recognition rate but low encounter rate. The ideal is a T-shape: broad enough to encounter anomalies, deep enough to know which ones matter.

Nassim Taleb would frame this differently. He'd say you want to maximize your exposure to positive Black Swans — events with unbounded upside and bounded downside. Buy a lot of cheap lottery tickets where the tickets are metaphorical: small bets, side projects, conversations, experiments. Most will fail. The ones that work will more than compensate.

The engineering framing matters because it shifts the locus of control. "I was unlucky" is a statement about the past that implies helplessness about the future. "I had insufficient surface area for luck" is a diagnosis with a prescription.

None of this eliminates the role of genuine randomness. Some people are born into circumstances that make every encounter easier or harder. The starting conditions are not equal and pretending otherwise is dishonest. But within whatever constraints exist, the variance in luck-surface-area between people is enormous, and most of it is self-imposed.

The luckiest people I know are not optimists. They are engineers who happen to be building probability machines instead of software.