Anthropic chief architect Sam McCandlish's first post hints at internal-acceleration report, possibly hitting 2x RSP threshold
inductionheads · x · 2026-11-06
Sam McCandlish, Anthropic co-founder and Chief Architect who leads pretraining and RL, posted for the first time, previewing upcoming disclosures about his safety work and commenting on METR's latest independent evaluation report.
- He calls the pace of model capability progress one of the most important questions in AI safety, praising METR for giving the public an independent external view into Anthropic's internal acceleration
- Amplifier Andrew Curran speculates the forthcoming internal-acceleration report will be the announcement that Anthropic has hit the 2x Responsible Scaling Policy threshold—implying doubled capabilities with correspondingly tighter safety requirements
- More safety details are promised in the coming months
More from Models
- Tinker team benchmarks model finetunability: Inkling 2 Small learns faster for less — ziqiao_ma · 2026-11-06
- Inkling 2 Small posts top vision benchmark scores, endorsed by PyTorch creator — tessybarton · 2026-11-06
- 28,849-trial Harness Leaderboard: bash-only mini-swe-agent beats openclaw 51.4% vs 27.3% — bansalg_ · 2026-11-06
- Qwen 27B + Python REPL harness hits 78.9% on Harvey benchmark, beating larger flagships — bansalg_ · 2026-11-06
- Running quantized DeepSeek V4.1 on a single RTX 3090 at 20 tokens/s — GodComplecs · 2026-11-06
- Unsloth's danielhanchen to release quantized versions of Thinking Machines' new model — danielhanchen · 2026-11-06