I think there are still some out there holding on to impressions of how AI for programming worked a year ago.
It has changed so drastically since then. Not just in the quality of its work, but in the quality and precision of its investigation, the thoroughness of its testing,…
Commit
372a953
Pretty much exactly how I feel. Furthermore, if you make it write unit tests and make it confirm every bug via a unit test, it reduces the amount of nonsense bugs it might report if you just ask it "analyse the code base".
Source: me, I've asked for "analyse the code base" before and the quality of bugs goes down with each iteration :P
More and more I find myself trusting the _code_, although the *UI* still leaves some to be desired in some cases.
commit 372a953
Yes this! And UI is a great place where your discretion and intuition and your physicality in how you interact with an interface, they all come in to fill that gap that AI is not able to step into.