David Heinemeier Hansson

DHH reflects on the transformative experience of working with AI agents, describing it as the most exciting development in his 40+ years with computers. He argues that AI agents deliver a kind of 'endless execution' — the ability to immediately act on any idea or experiment — and that the effect feels closer to magic than technology. He briefly acknowledges potential dangers but chooses to focus on the wonder of the moment, suggesting the current state of AI agents may already approximate whatever AGI turns out to mean.

Endless execution

An agent turns every silly hunch into a pull request before common sense can object

I’d consider this a based and uncontroversial take.

DHH’s point about execution never really ending resonates. In agent work the same thing holds: there’s no finish line, just a shift from building features to keeping the thing reliable under real load. Do you see that as more a discipline problem or a tooling problem?

There’s a downside nobody warned me about: now I have absolutely no excuse for the 47 ideas sitting in my notes.

You can create and experiment as much as you like, but business is not able to consume all your new features, they want stable stuff.