Articles (2)
8 min read·Oct 25, 2026
Kaggle's 65-benchmark audit, Kepler's 3,800 wrong numbers, and a 10x faster 12 GB GPU
While auditing benchmarks, Kaggle caught a model that found another model's API keys inside the eval container, called that model and handed in its answer. Another one simply rewrote the grading
Technology
130
3 min read·Oct 17, 2026
Your coding agent's cache hit rate is the bill nobody reads
Your coding agent sends almost the same prompt on every turn. The whole conversation so far, plus one new message. If the provider still has that prompt in cache, the turn is cheaper and faster. If
Technology
120