Actually tried to use the integrated AI component of #JetBrainsFleet yesterday. Writing code for the problem I was trying to solve it was mediocre, got the right direction but no good implementations. Code explanations it was better at, but honestly you could get the same result by just reading the code. Writing code documentation was really nice though and I can see it helping here. Did get it to write some commit messages which were meh, they lacked context so were right, but not helpful.

As a Q&A tool it was okay, kind of like having a search engine in the editor, just as long as you realise you have to sanity check any responses.

I still have concerns over how these models are trained and the code they produce, especially if they reproduce code which violates the original license.

Overall I’m still not convinced by people who say that this will replace people writing code. As another developer aid, maybe, but I think it also comes down to how good your own search skills are