A daily test now checks whether Opus 5.5 is quietly getting worse.
A developer launched LiveNerf on 24 September to retest Claude Opus 5.5 against its launch-week scores. It runs an older model alongside as a control, and calls a change real only past three points. Six of thirty days are collected, and the first honest verdict cannot arrive before day twenty-four.
Why it mattersEveryone claims their AI got worse after launch, but nobody kept a clean opening score. Now one exists, so the next argument can use numbers instead of impressions.
Read in the app — free, in 9 languages
Related stories
OpenAI's new model nearly matches its best at a fifth of the price.2026-09-30OpenAI's new always-on agent comes with its own cloud computer.
2026-09-30A model trained at home in two hours tied a paid decision service.
2026-09-29Meta's agent sold his keyboard cheap and handed over his address.
2026-09-29Microsoft ran 1,024 agents on one job with nobody in charge.
2026-09-29