LiveNerf: daily check for Opus 5.5 degradation
Month-long benchmark re-running 78 calibrated questions daily against launch-week results, with Opus 5 as control, to statistically detect any quiet post-release quality drop.

- Launch
- Claude Opus 5.5
- Use case
- Research
- Format
- Dev tool
- Source
- github.com
ninjahawk
Physics student building and studying AI systems

