Nathan Lambert questions Anthropic's AI R&D metrics as too 'soft' on skills

natolambert · x · 2026-09-18

Nathan Lambert critiques Anthropic's new framework for measuring how much AI R&D is done by AI, arguing its appendix definition of R&D skills includes too many 'soft' skills compared to what someone who has actually built many models would count. He adds that the headline 26% figure will likely get widespread airtime despite the questionable methodology.

Original post →

More from Companies & People

Companies & People channel →