4 ms·

Can you recommend any model that doesn't do this?

• croon 17d ago
  Not GP, but IME it's not fixable by model selection, but being zealous about guiding output and vision, and pushing back on all the bad habits LLM in general has (eg verbose output as a band-aid for emergent intelligence). As soon as something is introduced into your codebase, it will continue being picked up into context until you remove it and any reference to it from any potential context entrypoint. If you don't any model will keep venturing down wrong/bad paths.
• RamblingCTO 14d ago
  sadly not. I just run circles trying to remediate it after the fact