I’ll spare you a few minutes by not writing some dumb intro: we all know what AI is, and what it is doing to our industry.
For the longest time, the bottleneck for engineers was writing code. Assuming they had proper specs and requirements of course. Now, with something like Claude, every IC and their manager is literally pumping out thousands of lines of code. In a day. So that bottleneck no longer applies.
What about reviews though? There’s just no way things are getting reviewed properly anymore. Because you can’t review so much as a human being. It’s physically (mentally?) impossible I think. Unless you have the perfect TDD setup or whatever.
In practice, this (in my opinion) has led to tons of broken products, absolutely garbage codebases, horrendous performance issues, etc. Because everybody uses Claude, but nobody can review this level of sheer slopocalypse that’s being generated.