AI Killed the Founding Team, Not the PM Job

Two product leaders shipped a company without an engineer

Reforge published a case study worth sitting with: two product leaders built a working SaaS company in nine months with zero engineers on the team. Not a prototype, not a Figma file with a waitlist. A live and supported product, with paying users, shipped by people whose job titles have never included “writes the code.” What died in that nine months wasn’t a role, it was the founding-team archetype, and that has direct consequences for how equity, hiring, and diligence should work now.

The obvious lesson everyone reached for was the wrong one. The story got read as “PMs don’t need engineers anymore,” and a smaller, angrier crowd read it as “AI is coming for the PM job too.” Both miss the point.

For 20 years, the default shape of an early-stage company has been a pair: a technical cofounder who builds and a business cofounder who sells. Venture capital treated that pairing as load-bearing. Accelerators screened for it. Equity splits assumed it. The Reforge case is interesting because it shows what happens when the load the archetype was carrying disappears.

The narrow read undersells what happened

“PMs can build now” is true and boring. It treats the story as a productivity anecdote, the kind of thing that shows up in a tooling newsletter and gets forgotten in a week. Two people used AI coding tools, moved fast, cut scope, and shipped. Nice. Next.

That framing misses the structural point. The interesting variable was never whether a PM could hand-assemble a React component. It’s that the number of humans required to convert an idea into a running business dropped, and it dropped past the threshold where you need a specialist to hold the technical half of the company together.

The narrow read also assumes engineers were the bottleneck being removed. They weren’t. What got removed was the coordination between two people who each held half the company in their heads. The technical cofounder held feasibility. The business cofounder held demand. The entire ritual of the founding team, the long dinners, the equity negotiation, the “do we actually work well together” anxiety, existed to fuse those two halves into one decision-making unit. When one person can hold both halves because AI carries the feasibility load, the fusion problem stops existing. You don’t need to negotiate with yourself.

The archetype was a hedge against scarce code

Here is the part startup culture hasn’t internalized. The technical-cofounder requirement was never about talent worship. It was a hedge against a specific scarcity: code was expensive, slow to produce, and impossible to fake your way through. If you couldn’t build, you couldn’t test whether anyone wanted the thing, and you’d burn your seed round paying an agency to build the wrong product slowly.

So the market solved it the way markets solve scarcity. It priced the scarce input into the cap table. A technical cofounder got 40 to 50 percent of a company not because writing code is intrinsically worth half a business, but because that person removed the single biggest source of early-stage risk: the risk that you could never cheaply find out if you were wrong.

That scarcity is gone, in the specific sense that matters for company formation. The cost of producing a testable version of an idea has collapsed toward the cost of describing it clearly. And describing an idea clearly enough to build it is a product skill, not an engineering one. The hedge is still priced into the culture. The thing it was hedging against has left the building.

This is the same pattern that shows up every time a scarce input gets cheap. When distribution was scarce, the media company that owned the printing press held the power. When compute was scarce, owning the data center was the moat. The asset that commands the premium is always the scarce one, and the premium evaporates the moment the scarcity does. Code is now the printing press after the internet.

What replaces it is the operator-generalist

The unit that replaces the founding team is the solo or duo operator-generalist: someone who can hold demand and feasibility at once because AI carries the parts that used to require a second brain. This isn’t a temporary arbitrage that closes when everyone catches up. It’s a new default shape.

The consequences ripple straight into the mechanics of company-building. Equity splits built around “technical cofounder gets half” describe a world that no longer sets the price. Hiring plans that front-load three engineers before the company has a repeatable motion are hedging against scarcity that isn’t there, and they burn the runway that used to justify the hedge. The first hire in a lot of these companies won’t be an engineer at all. It’ll be whoever removes the current bottleneck, and the current bottleneck is almost never “we can’t produce code.”

The sharpest change is in how investors read a team. The old diligence question was “is this team complete?”, meaning, do they have their technical half. That question is now close to meaningless. A complete team is one operator with taste, judgment, and the ability to convert ambiguity into a shipped test. The diligence question becomes “how fast does this person learn from what they ship,” because the constraint has moved from building the thing to knowing which thing to build. Feasibility is cheap. Judgment is the scarce input now, and judgment is what should command the premium. This is the same migration from abundant execution to scarce judgment that determines who stays valuable as AI commoditizes the mechanical middle of the work.

The strongest objection, and why it holds up anyway

The obvious counter: AI-built SaaS is fine for a CRUD app with a Stripe button, and it falls apart the moment you need real infrastructure like hard scaling, security, or a genuinely novel technical core. Fair. There are companies whose entire reason to exist is a hard engineering problem, and those companies still need engineers who own that problem end to end. But that objection defends a narrower and narrower territory every quarter, and it doesn’t touch the thesis. The claim was never that engineers stop mattering. It’s that the founding-team archetype stops being the default. Most software companies are not solving a novel infrastructure problem in month one. They’re finding out whether anyone wants the thing, and that discovery phase, the phase where the founding team was supposed to earn its structure, is exactly the phase AI has hollowed out. The deep-engineering hire still happens. It just happens later, as a deliberate choice against a known problem, not as a reflexive hedge stapled to the cap table on day one.

What to actually do about it

For operators: stop looking for a technical cofounder to complete you, and start asking whether the second person adds a capability you can’t buy. If the answer is “they write the code,” that’s not a cofounder anymore. That’s a role you can fill later, from a position of knowing what you actually need.

For investors: retire “is the team complete” as a screen. It selects for the archetype, not the outcome. The team that shipped a company in nine months with no engineers would have failed the completeness test on day one, and it would have been the wrong call. The trait to underwrite is a person’s speed of learning from things they’ve shipped, because that’s the only input that stayed scarce.

The founding team was the market’s answer to a question that no longer gets asked. The companies that win the next decade will be the ones that stop answering it first.