Over the past year I started building products alone, with AI, from idea to production. First Bloom, a training app I use every day. Then km7, a voice companion for runners. Real products, real users, no team.
The thing nobody tells you about building with AI is that execution was never the hard part. It felt like the hard part, because it was the slow part. Months of development made every decision feel expensive, so we got good at planning, estimating, negotiating scope. AI removes most of that friction — and what's left, suddenly visible, is the actual hard part: knowing what to build.
When code becomes cheap, the expensive things are the ones that were always expensive. Understanding a problem well enough to describe it precisely. Deciding what to leave out. Noticing when you're building for an imagined user instead of a real one. I've thrown away more working features this year than in my whole career — not because they were broken, but because building them was so fast I could afford to learn they were wrong.
That inversion changes what product management is. The role was never really about coordinating output, but organisations treated it that way because output was the bottleneck. Now a single person can ship what used to take a team a quarter. The bottleneck has moved to judgment: which problem, which bet, which evidence would change your mind.
It also changes discovery. AI compresses build time, but it doesn't compress learning time. Users still need weeks to show you how they actually behave. If anything, cheap execution makes discovery more important, because the cost of confidently building the wrong thing has never been lower — you can now build three wrong things in the time it used to take to build one.
I keep meeting product managers worried that AI makes their role obsolete. Building with it daily has convinced me of the opposite: the parts of the job AI absorbs are the parts that were never the point. What remains — problem selection, taste, the ability to say no with evidence — has quietly become the whole job.
The teams that will struggle are not the ones without AI skills. They're the ones that never learned to decide what's worth building, because the slowness of execution always hid that gap.