It's not just you (gpt-5 got dumber)

Summary

The transcript discusses perceived performance degradation in AI models like GPT-5 and Claude, with OpenAI actively investigating potential regressions in their system. The speaker highlights OpenAI's transparent approach to addressing potential issues, referencing an internal document titled "Ghosts in the Codeex machine" that details their findings. The key takeaway is that while AI model performance can fluctuate, responsible tech companies are committed to identifying and addressing potential problems quickly and openly, maintaining user trust and continuous improvement.