VC: no one who has stared into frontier models stays bullish on open source
abhiadesai · x · 2026-10-12
Investor Abhi Adesai argues that "cheap is good enough" doesn't map onto the physics of superintelligence, citing Karpathy's point that only 5,000 people worldwide have seen frontier models' upward capability curve firsthand.
The quoted Metacritic post adds evidence: Musk himself doesn't believe in open-weight superintelligence—xAI could compete with Fireworks and Baseten in open-weight inference, but instead commits hundreds of thousands of GPUs to frontier training and routes Grok consumer users to Anthropic.
More from AGI Musings
- Meta intern research: IdeaScientist boosts AI research proposal novelty by 25% — JeremyNguyenPhD · 2026-10-12
- AI's lowering of the barrier to applied math in engineering will be a landmark tech achievement — _onionesque · 2026-10-12
- AI's biggest achievement will be making applied math easy in ordinary engineering — _onionesque · 2026-10-12
- Why a million abstract-thinking AIs won't master rocket design: the tacit knowledge argument — binarybits · 2026-10-12
- Hinton argues the brain is analog: LLMs simulate intelligence's output, not intelligence itself — AlexTensor · 2026-10-12
- How Do Junior Engineers Become Seniors Without Coding in the AI Era? — KlausCodes · 2026-10-12