Speculation on Permanent Sandbagging by OpenAI and Anthropic

scaling01 · x · 2026-08-31

This post expresses a pessimistic view on the strategies of frontier model companies, suggesting that both OpenAI and Anthropic are now permanently "sandbagging"—releasing models with capabilities significantly below their true potential. The author argues this obscures the actual technical frontier and may lead open-source advocates to mistakenly believe open models are catching up.

Original post →

More from Companies & People

Companies & People channel →