News for hackers. Stuff that matters.
Re:$500 GPU outperforms Claude Sonnet on coding benchmarks by nl on Sunday March 29, 2026 @02:42AM (#47560894)
That isn't true. The whole point it to quickly pick up statistically significant variations quickly, and with the volume of tests they are doing there is plenty of data.

If you turn on the 95% CI bands you can see there is plenty of statistical significance.

Re:$500 GPU outperforms Claude Sonnet on coding benchmarks by withinboredom on Sunday March 29, 2026 @05:04AM (#47561517)
Unless you and I are looking at different web pages… it only goes back to February, not December or January.