"It will build exactly what you asked for, even when what you asked for was wrong" is the most important sentence here. It is also why sales turns out to be good training. A discovery call is mostly finding out what the person actually needs before you promise anything.

The numbers are striking, 287 merged PRs in a month, but I would be curious about the other side of them. At that pace the hard question is how you know each change did what you meant, and what you check before a real user sees it.

At Ascendra Ventures we see the same pattern with founders who build with agents: the code gets cheap, and the value moves to writing down what done looks like before the agent starts. People who can explain clearly have an advantage there.

What does your check look like before you merge: do you run the app yourself, or do you have tests the agent has to pass first?