Would Anthropic Researchers Trade 90% of Their Team for Opus 4? A Test of the 12x Speedup Claim
tobyordoxford · x · 2026-11-22
Toby Ord presses further on the credibility of AI research speedup claims: would the Anthropic researchers who reported a 12x uplift from Model 4 (versus 1.3x from Opus 4) be willing to lose 90% of their team in exchange for continuing to use Opus 4?
It's a thought-experiment test — if the claimed efficiency gains were real, the trade-off between an old model with a small team and a new model with a large team should hold; if not, the speedup numbers are inflated.
More from AGI Musings
- V.E. Schwab: Writing is an art, publishing is a business — david_perell · 2026-11-22
- Asking basic questions in the AI era: don't let gaps feel like verdicts — brandongalang · 2026-11-22
- Could a human+Stockfish combo beat 10x compute Stockfish? — mgostIH · 2026-11-22
- Claude predicts AI will crack math proofs, battery materials and abandoned drug candidates within 5 years — alex_verem · 2026-11-22
- Optogenetics could have been invented in the 1970s — how common are biology's 'delayed inventions'? — NikoMcCarty · 2026-11-22
- Most math will never be proved: shortest proofs are astronomically long, says dev — LLMs don't change that — jfischoff · 2026-11-22