Also had a few misses. But in general it is ok. Still prefer ai assistant, because i can then direct the result into a certain direction. It also feels faster, it probably is not because of the manual stuff involved.

It has showed promise, enough to quell my FOMO about other IDEs, since I am extremely happy with the Jetbrains suite otherwise.

I believe it’ll get much better as LLMs start editing code by invoking refactoring tools (rename, change signature, etc.) rather than rewriting the code line by line, as this will let them perform large-scale changes reliably in a way that's similar to how software engineers do them now using IDE tools

It's also missing quite a few features still (like checkpoints, stopping the agent mid-task to redirect, etc). But the core feature set is mostly there.