VC and PE is a data problem.

At my first venture firm, the knowledge lived in people.

A read on a founder traveled down the hallway before it ever reached a memo. A thesis hardened slowly, over months, in arguments nobody wrote down. The rules of thumb got handed around like family recipes. Never quite the same in any two partners' hands.

And the calls were fast. A partner would look at a founder, a team, the shape of the early growth, a market that was maybe two quarters early, and weigh all of it in a few seconds against a thousand companies they'd already seen. Below the level of words. I couldn't do that. Nobody can, at the start.

Which is why you learn venture the way you learn a trade: years at someone's elbow, because there is no textbook. You load the library by exposure until your eye is trained.

Crafts have a history. It doesn't end in craft.

For most of human history there was no biology. There was only the trained body. Shennong tasting a hundred herbs, poisoning himself over and over to learn which ones healed and which ones killed. Knowledge bought one trial at a time. For centuries the instrument was a palate and an eye.

Then, inside a single career, a different instrument showed up. Sequencing a genome fell from about $95 million to a few hundred dollars. AlphaFold took a fifty-year problem and handed Andrei Lupas the structure his lab had been chasing for a decade, in half an hour.

The eye didn't go away. It got an instrument.

Private markets haven't had one, and for reasons that are actually good. The data is thin: a few thousand deals a year that mean anything, where other fields count in millions. It's selection-biased in the cruelest possible way: you never learn what the companies you passed on became. You learn about the two that got famous, which is worse. It's non-stationary, the rules mutating every cycle, so the metrics we treated as gospel ten years ago are already being rewritten. And it's slow. A seed check takes ten years to grade. By the time you know whether you were right, you're a different investor in a different market.

People have been chipping at this for decades. What changed is the ground under them.

The exhaust is legible now. Every company throws off a trail long before it throws off revenue, and that trail has finally become cheap to read. A decade of cloud migration piled up terabytes nobody had a way to ask questions of. Asking them is now a rounding error. Hiring pages, commits, job posts, app installs, usage telemetry, developer chatter — all of it moving months ahead of the P&L.

So here's the thing I've come to believe: venture and private equity are getting their instrument, and the transition has already started.

It won't be clean. Every field that gets one goes through a stretch where the trained eye and the new machine disagree loudly and both are sometimes right. New unknowns replace the old ones. Someone will overfit spectacularly and lose a great deal of money proving it.

I find that dangerously exciting. I'd rather be here for it.