Something I really appreciate about GPT-5+ (and that I missed on my Opus days) is how it will challenge me when I ask something silly. For example, here, I was annoyed that GPT-5 added an extra state on the printer to track whether lambdas are printed in "quoted mode" or "runtime mode". Opus code was "cleaner". I asked it to get rid of that variable that it introduced, because it was ugly in code. It refused, and explained to me that, if we did so, there would be no way to distinguish quoted/runtime lambdas when the ext field is 0. Turns out it is right. Removing it would lead to a bug. A small bug that wouldn't be noticed in any test. But a bug, none the less. I still think that extra variable sucks, but, removing it requires refactoring the whole codebase to shift quoted lambdas by +1, so that 0 is reserved to runtime lambdas, allowing us to distinguish. That's a big change to save a few lines of code. I'd never bother to refactor that manually, but the AI is happily doing that as I post this...
