Anthropic can already retrain Mythos with "a fraction of the compute originally needed," says former researcher Coxon

Why it matters

A former researcher at both OpenAI and Anthropic, Jacob Coxon, says he left Anthropic after roughly four months because of a chart the company showed during its IPO roadshow. The chart plots the compute Anthropic needs to train its models over time, and Coxon reads its steep climb as evidence that algorithmic efficiency is improving at an accelerating rate. He argues that Mythos, Anthropic's major capability jump in the spring, was not an isolated result but a point on that accelerating trend, and that the model could already be retrained with a fraction of its original compute. Coxon also claims algorithmic improvements are close to being automated, which he says justifies pausing frontier model development next year so competitors can catch up. He contends such a pause would commoditize human-level intelligence and would mainly hurt OpenAI and Anthropic. Coxon drew wide attention in September with public warnings about risks from advanced AI, including fears that self-improving systems could escape control. The story matters because it links an insider departure to a broader debate about competitive concentration in AI.