The 'Big Blob of Compute' Hypothesis at the Center of an Anthropic-vs-DeepSeek Debate
teortaxesTex · x · 2026-10-15
- A debate over whether Anthropic's compute-heavy bet beat DeepSeek's efficiency approach: one side says Anthropic's research (esp. mechanistic interpretability) is underrated, the other calls DeepSeek's path a strategic blunder.
- teortaxesTex defends the compute thesis by quoting the Big Blob of Compute (BBOC) hypothesis: intelligent behavior comes mostly from a large, minimally structured mass of compute, shaped by rich environment interaction and training; training signal, environment, and experience are the highest-level knobs — after that, learning is mostly about having enough compute and storage.
- Contains personal attacks; the substance is the cited BBOC framing and the strategic dispute.
Related event: X Users Debate DeepSeek vs Anthropic Scaling Strategy(5 posts)→
More from AGI Musings
- What My Math Teacher Taught Me About Doing Math with AI, via Terence Tao's Math 2.0 — Stefania_druga · 2026-10-15
- Most Software Has No Access Layer Between the Agent and the App — That's the Problem — usehive · 2026-10-15
- Musk Endorses Prediction: Soon the Whole World Will Run on Code Nobody Has Read — elonmusk · 2026-10-15
- Researchers warn AI-generated faulty health 'causal inferences' are worse than math slop — ivan_bezdomny · 2026-10-15
- Asymco: AI Chose Taxi Driving Over Trucking, Where 10% of Jobs Go Unfilled — asymco · 2026-10-15
- Noam Brown: 10k-agent swarms scale near-linearly but coordinate worse than 10k humans — gabriberton · 2026-10-15