Researcher says literature already contradicted Anthropic's shifting subagent guidance months ago
gerardsans · x · 2026-11-17
Researcher Gerard Sans cites his synthesis of agentic AI evidence (grounding/belief-revision limits, multi-agent error amplification, a persistent 30% misbehaviour floor, benchmark-vs-deployment gap) to argue Anthropic's guidance is self-contradictory: first recommending subagents, then advising against them. He claims the literature established these conclusions months ago and that AI labs' statements shouldn't be taken at face value.
Related event: Anthropic Advises Developers to Fold Subagents into Single-Agent Loops(4 posts)→
More from Models
- Karpo Decision 31B open weights: a decision model scoring all options in one forward pass — _jaydeepkarale · 2026-11-17
- Rohit Krishnan: getting Gemini-pilled again — still overconfident but very good models — zacharynado · 2026-11-17
- Users complain they cannot escape Nano Banana 2 Lite on Gemini websites — dr_canconfirm · 2026-11-17
- Kaggle launches Code2Video Arena with HeyGen; Gemini 4 Carbon tops leaderboard — AI_Andrew · 2026-11-17
- Early testers say Tencent's Hy4 Flash beats Claude Haiku 5.5 — TencentHunyuan · 2026-11-17
- Claude refuses the task on safety grounds; a local open model picks up where it stopped — draginol · 2026-11-17