[kikimora]

> You don’t give detailed prompts. You give objectives like “Refactor the API layer for scalability and update all related documentation.” The AI agent then plans, executes, tests, and reports the results.

lol, I wonder if the author (if there is a human author behind this) ever tried this in production. I did multiple times, never worked for me. All it did is produced a tone of spaghetti code and lots of overlapping tests with medium coverage. I’ll be fine with it if I could use AI to evolve this code without ever reading and understanding it. But we are far from it as of today.