Loading request...
Users experience frustration when AI generates incorrect code and then, when asked to fix it, points out errors in its own previously generated lines, often requiring a full context reset. They need AI to be better at self-correction and debugging its own output.
I am pretty sure AI code generation will be pretty amazing and important in the future. But for the moment I worry we are just building amazing efficient tech debt generation tools. We can now generate badly specified code faster than ever before. Presumably orgs are hoping that there will be better coding AIs in future that can fix all the problems current AI tools will cause. We live in strange times. Maybe I have called this wrong. Who knows.