When everything becomes buildable

Software development is turning into a factory. Not a person with an AI assistant, but an end-to-end system where agents take the signals (a bug, a request, a metric that moved), carry them through building, testing, reviewing, and shipping, then watch what happens and feed it back in. The humans don’t leave; they move from being in the loop to being on it. Increasingly, engineers aren’t the ones building the software; they’re building the factory that builds it. Whatever the exact shape it settles into, the direction is clear enough: the cost of turning an idea into working software is falling rapidly toward near zero.

The work I do took its shape in the opposite world. There has never been a clean definition of product management; it bends with the company, its size, its stage, whether the culture treats the PM as a founder-in-miniature or a project manager with a nicer title. I’ve long since made my peace with that. But under every version of it sits the same fact: building was slow and expensive, and someone had to decide what got built. Roadmaps, prioritization frameworks, the ceremony of the quarterly plan, the whole art of saying no: strip the vocabulary away and most of it is machinery for rationing a single scarce resource, engineering time. We got good at choosing which few of a hundred ideas deserved the build, and that skill mattered precisely because the build was the bottleneck. The factory removes the bottleneck, and a surprising amount of the job turns out to have been about the bottleneck.

For anyone who likes to build, this reads as liberation, and a great deal of it genuinely is. But the cost of building was also the thing that forced us to think. Because it was expensive, you had to make the case before you made the thing: write it down, argue it, get it clear enough that other people would spend months of their lives on it. Cheap building removes that forcing function. You can now build before you understand, and the factory will happily produce a working, plausible, beautifully tested answer to a question nobody asked. An abundance of supply, it turns out, isn’t the same as an abundance of value. For most of my career, “does it work” was the hard part, which made it a decent stand-in for “is it worth anything.” When working is free, it stops being a proxy for anything at all.

So the scarce thing moves. When you can build almost anything, the constraint is no longer capacity; it’s knowing where you’re actually trying to go. The word everyone reaches for is taste, and taste is real, but it’s the smaller half of it: the part that picks the better of two things already in front of you. Vision sits upstream of that, deciding which things deserve to be in front of you at all. With it, fast iteration is a superpower: a hundred cheap shots at a target you can actually see. Without it, speed just means failing more often, and the failures teach you little, because you were never aiming at anything in particular. Failing itself is fine, and nearly free now, as long as something is learned, and learning is the part that compounds once the code no longer does. Speed without vision starts to look less like iteration and more like a faster way to stay lost.

If I’m honest, the whole shift is something I’m glad about, because it hands the work back to building. I spent years drifting toward whatever seat kept me closest to the making, and I never shook the belief that the surest way to know if an idea is any good is to build a rough version and look at it. For a long time that instinct had to be rationed like everything else. Now, if you carry some vision and some taste and you happen to love the act of making, the era is arranged in your favor: the thing you enjoy moves back to the center, and much of what you once tolerated to earn your way to it falls away. If you loved building before, you’ll almost certainly love it now, the same pull with a lot of the friction stripped out. What changes is how you build, not whether: less cutting every piece by hand, more composing and steering the factory that cuts them for you. The catch is the factory itself, still early and rough at the edges: some companies will be far better at standing one up than others, a few won’t manage it at all and will fall behind faster than they expect, and building your own still takes real investment before it returns anything. But it’s a bet I like, because underneath the uncertainty it puts the thing I care about most back in the middle of the work, and only asks me to learn a new way of doing what I was going to do anyway.

Monetization, where I’ve sat for a few years now, gets a head start on this, and I don’t think that’s an accident. Monetization has always been run as a portfolio of bets: form a thesis, ship the smallest version, read the numbers, iterate, and make your peace early with most experiments losing. That posture, test-and-learn as the default rather than an occasional event, is exactly what the factory is about to make normal everywhere. The stranger consequence is that the instinct doesn’t just transfer; it scales up. When the cost of a swing collapses, the rational move isn’t only more swings at the same size; it’s bigger ones. The bet that used to be too expensive or too risky to justify becomes a sprint you can afford to lose. So the discipline that taught me to find the smallest change that moves the most is, for the first time, also free to go looking for the largest.

Teamwork changes shape too, maybe more than the individual job does. A lot of what we call collaboration is really handoff management: the spec thrown over the wall to design, design over the wall to engineering, engineering to QA, each seam a place where intent leaks out and time drains away. The factory automates the seams, and the relay race becomes something closer to a small, dense huddle. I’ve written that a PM who can build, next to a designer and a developer, is already a terrific trio; what shrinks is everything that used to sit between them. Teams get smaller and hold more of the problem in their heads at once, and coordination stops being the tax it always was. What they need from each other changes accordingly: less handing-off, more agreeing on where to point the thing, which puts the premium on shared judgment and a common picture of where you’re headed, the two things that stubbornly don’t automate, for now.

None of this makes product less necessary; it strips it down to what it always was underneath.