How much time do you actually spend testing AI-generated code?
by•
Building an app with AI agents feels super fast at first but I noticed most of my time still goes into checking edge cases and fixing weird bugs.
Things like broken date formats file limits or small UI glitches don't just fix themselves, even when the AI says they are done.
It feels like my job shifted from writing code to just doing strict QA all day.
Is anyone else spending more time debugging than prompting lately?
13 views
Replies
more than writing it now. type check plus the full test suite before every commit, no exceptions. boring, but it's what catches the stuff the agent was confident about 😄
Yes, and dates were the worst one for me, so I took them away from the model completely. A hook hands the agent today's date and a two week calendar on every prompt, and any date it writes in a reply gets checked against the real calendar before I see it. When the same mistake comes back a second time, I stop pointing at it and build a check like that.