The current state of gpt-5

Summary

The transcript discusses the creator's experience and observations with GPT-5, exploring potential performance changes, inconsistencies, and speculative reasons behind variations in the model's capabilities. The speaker references a Less Wrong blog post and mentions investigating the nuances behind perceived performance degradation, while remaining curious about whether special versions or access points might explain the differences. Ultimately, the analysis suggests that the problems may be simpler than initially assumed, though the full details remain somewhat uncertain, with the speaker emphasizing a desire to be transparent about their findings.