Anthropic IPO chart reportedly shows faster algorithmic progress
AI summary
Former Anthropic researcher Jacob Coxon said an IPO roadshow chart showing model-training compute requirements was one reason he left the company. According to IT之家, the compressed-axis chart indicated accelerating algorithmic efficiency, while Coxon and a Reddit post said Mythos could now be retrained with a fraction of its original compute; 36氪 also described the curve as rising sharply on a logarithmic scale.
In-depth insight
The reported disclosure offers a company-level view of rapidly improving training efficiency, a central issue in debates about AI scaling and automated self-improvement, although the compute claims have not been independently verified.