If AI writes the code in 2 days and you spend 10 debugging it, where do the 10 actually go?
A dev I was talking to last week put it like this: the build took 2 days with AI, the debugging took 10. He didn't say it as a complaint — more like it's just the shape of the work now.
What I can't work out is whether that's the same debugging we've always done, just more of it, or whether it's a different activity wearing the same name.
So, concretely — think of the last time you shipped something substantially AI-generated and it took much longer than expected to get working.
- Roughly what was the build : debug ratio? (Even a rough guess — 1:2, 1:6, whatever.)
- On the worst day of that debugging, what were you physically doing? Reading code? Adding print statements? Re-running the same test? Talking to the AI about its own output?
- When you finally found the problem — what had you been doing in the ten minutes right before you found it?
- Was the bug in code the AI wrote, code you wrote to glue it together, or in the space between them?
And the one I'm most curious about: did it feel different from debugging code you'd written yourself? If yes, what specifically was different. If it felt exactly the same and the ratio thing is overblown, I'd genuinely rather hear that — I'm trying to find out if this is real, not collect agreement.