Ben Todd pushes back: Epoch ECI is math-driven, so was Anthropic really behind on agentic coding?
ben_j_todd · x · 2026-11-22
Responding to herbiebradley's claim that Anthropic's acceleration starting in 2025 may exist precisely because it lagged OpenAI in 2025 and had to catch up, Ben Todd pushes back: were they actually behind on agentic coding?
He concedes OpenAI was better at math in 2025 and agrees the catch-up acceleration story is plausible, but argues the Epoch ECI benchmark is heavily math-driven at the top end, so it may not reflect the real agentic coding gap between the two labs.
More from AGI Musings
- Researcher slams journalist's 'AI is just good at Number' take as deeply confused — ShakeelHashim · 2026-11-22
- Today's AI solves our posed problems; can it invent new conceptual frameworks? — AlexTensor · 2026-11-22
- If an AI lab proves P=NP first, society will read it as an attack, not a breakthrough — ryunuck · 2026-11-22
- Open-weight models can no longer be suppressed: Pirate Face turns model weights into immortal torrents — esrtweet · 2026-11-22
- Kalshi traders now put 54% odds on P vs NP being resolved before 2030 — ricklamers · 2026-11-22
- David Friedberg says Google has hit RSI loop and Anthropic likely has too — Key_Stand_1731 · 2026-11-22