Yeah, I guess I was under that impression, since if Claude is similar to Codex at coding, while being a larger, more expensive model (which seems likely based on API costs and Nesov’s analysis of their training hardware), then Anthropic has no clear lead (or would be behind if not for Mythos). So I thought your claim of their lead was partly based on an impression (similar to Nesov’s) that Claude is significantly better than Codex at coding.
(And I think a lot of people were probably under this misimpression at some point, including me, due to seeing a lot of talk about Claude Code on X around December, much more than about Codex, which in retrospect I have to attribute to a successful Anthropic marketing campaign.)
Yeah, I guess I was under that impression, since if Claude is similar to Codex at coding, while being a larger, more expensive model (which seems likely based on API costs and Nesov’s analysis of their training hardware), then Anthropic has no clear lead (or would be behind if not for Mythos). So I thought your claim of their lead was partly based on an impression (similar to Nesov’s) that Claude is significantly better than Codex at coding.
(And I think a lot of people were probably under this misimpression at some point, including me, due to seeing a lot of talk about Claude Code on X around December, much more than about Codex, which in retrospect I have to attribute to a successful Anthropic marketing campaign.)