I usually give AI little coding tasks. Go off and fix this one thing, make this small tweak, add this one feature that I will review. Today I had it help with something bigger. I iterated on the plan, then it worked for 30 minutes. Pretty incredible.
@flotisserie @g I’ve been using OpenAI’s Codex. It seems comparable to Claude Code with the latest models. I already pay OpenAI so just can’t justify another subscription. I also think there’s a sort of lock-in as these companies have a suite of tools… Planning to write more about that.
@manton same here. I had a test I wanted to run to compare memory usage of different languages doing the same thing. I had a well structured python program that I got it to port to Go and Rust. Iterated a bit but basically had 2 ports up and running in no time. I was very impressed by it being able to identify external code modules to pull in the replicate some of the python modules I’d used.
