The last couple of weeks using Claude has been…interesting to say the least.

Additionally I run a constant hacking contest between GPT and Claude. It’s a toy project and it simulates an attack/defense of a small corporate network.

Claude used to win pretty handily. Suddenly it’s started to lose 90% of the time. I thought GPT had gotten better but no, looking at the logs it seems that Claude is slower and more prone to running in circles. This is still the case when switching to Opus 4.7.

I don’t know what that means but it’s undoubtedly worse.