evaluation

7 posts tagged “evaluation”.

1. Our Baseline Was the Three Worst Samples in the Log (2026-09-02)
2. Every Failure Silently Scored 0.5 (2026-08-19)
3. Every Surprising Eval Result Was Our Infrastructure (2026-08-15)
4. You Can't Prompt Your Way to Both Commitment and Abstention (2026-08-11)
5. The Benchmark Said 18%. Production Said 0.07%. (2026-08-10)
6. Zero of Twenty-Five Skills Were Loading (2026-08-06)
7. Our Model Stopped Saying No (2026-08-05)