Former Anthropic researcher says a slide from the IPO roadshow weighed on his decision to leave: “they could already retrain Mythos with a fraction of the compute”
Jacob Coxon, who worked at OpenAI and Anthropic, posted a slide from Anthropic’s IPO roadshow on the compute efficiency of pretraining over time and said it was “a big factor” in his leaving the company. According to the post, the axis is logarithmic and shows “superexponential” algorithmic progress, and Mythos “was not a one-time deviation”. He himself cautions that this does not strictly imply a software-only intelligence explosion, and in a reply he argued for pausing frontier development next year. This is his reading of the chart, with no comment from Anthropic.
Published in edition no. 30 of Noticiário do Joãozinho ·
Translated by Joãozinho AI from the original Brazilian Portuguese edition. Facts, numbers and sources are the same as in the original; quotes, video titles and source names stay in their original language. Read the original in Portuguese
Original post · X
@hilbertspaessThe Decoder on the postCoxon’s reply calling for the pause
https://x.com/hilbertspaess/status/2121331333654086124
Engagement 1.3M views5.4K likes471 reposts240 quotes340 replies3K bookmarks (as of Nov. 14, 2026 17:58)