43% of AI-generated code changes need debugging in production, survey finds
venturebeat.com> 43% of AI-generated code changes require manual debugging in production environments even after passing quality assurance and staging tests.
As compared to what for non AI written code? 53%? 33%
Furthermore, there’s no indication of the complexity of the problem AI was asked to solve?
It’s important to stay informed and open minded about AI, but BS articles like this aren’t part of that education process.
I'd guess... 6% for non-AI written? I'm thinking "debugging" in production means there's already a bug in production, so a very gentle term for a SEV/incident.
This is what probably keeps happening at GitHub, AWS, and Anthropic.
Apparently, software quality doesn't matter to them given they are certainly using AI agents and even with the outages, it seems to be okay for them.
Seeing an increase of engineers debugging AI gen'd code in production (which is what juniors do) is a result of this 'vibe-coding' nonsense in the first place. Looks like AI coding has de-skilled these "seniors" into "juniors" without them knowing it.
So when the next time Claude goes down, they'll be debugging it in production and all the "engineers" will complain about it here.