With AI

§Everything in this whitepaper was true before 2023. The reason to write it now is that the ground under it moved.

§For twenty years the expensive part of product was the middle: turning a decision into something a stranger could use. Building took specialists and weeks, so the owner's job was, in large part, managing that: writing requirements, sequencing work, protecting engineers, choosing what to build because building was scarce. Much of the large-company product management canon is about that scarcity.

§It is gone. In 2025 Andrej Karpathy gave a name to what had become obvious, that a person could describe software to a model and have it appear, and the year's fastest-growing companies were the tools that made it so. Lovable, Bolt and their peers reported growth rates that would have been implausible for any software product two years earlier, and what they sold was the collapse of the middle. A founder with no engineer can now ship a working prototype in an afternoon.

§ 11.1What that does to each chapter#

§Read back through this whitepaper with the middle removed.

§Integration matters more. When anyone can build, the signals arrive faster and from more directions, and the person through whom they converge is the only thing stopping the team from building all of them.

§The T-shaped owner's vertical bar shifts. Depth in engineering was once the most defensible expertise for a founder to hold product with. Depth in understanding customers is now the one that cannot be prompted.

§The three promises reweight. Vision was always the owner's. Execution was expensive and is now cheap. Strategy, the list of what the company will not build, becomes the hardest promise to keep, because the cost of breaking it dropped to one prompt.

§User experience is decided at scope and structure, and those are the planes the models are worst at. A generated screen is a surface. It does not know what the job is.

§Engineering and product renegotiate. Appetite still fixes the time; what fits inside it has grown, which makes the question of what to leave out more important, not less.

§Metrics inform, never set, and the temptation to let them set grows with the volume of things that can be built to move them.

§And every failure mode gets cheaper to fall into. Owning everything and deciding nothing now produces twenty prototypes instead of one backlog.

§ 11.2What did not change#

§Someone still has to sit across from a stranger and wait through the silence. Someone still has to tell a room that voted for a feature that the feature goes last. Someone is still accountable, by name, for what gets built. That is the foundation, and it is the same foundation it was. The only difference is that everything built on it is faster, which means the cracks show sooner.

§The rest of this book is what the owner does with the time that building used to take. Product Discovery is next: how to think about what to build, and why smart teams build the wrong thing anyway.