Thinking Machines on the Limits of AI
alex_verem · x · 2026-07-15
This post summarizes and comments on an essay by Thinking Machines Labs, arguing that: AI and the "local knowledge" in an economy cannot be fully absorbed by a single, centralized model.
Drawing on Hayek's critique of central planning, the author emphasizes that the knowledge underpinning the economy is often tacit, localized, and constantly updated—like a chef tweaking a recipe, a shop owner adjusting prices, or a company's internal decision rules. Feeding all this into a giant model doesn't eliminate this decentralization.
The post extends this logic to AI's impact on jobs:
- AI excels at tasks like chess or math where the board is fully visible and the goal is static.
- Real-world jobs rarely look like this; critical knowledge often lives in human judgment and collaboration rather than documentation.
- Therefore, AI is best suited for structured, clearly bounded tasks, not roles relying on tacit knowledge and constantly shifting environments.
Related event: Thinking Machines Explores AI Limits via Hayek(3 posts)→
More from AGI Musings
- Gary Marcus says LLMs still cannot really do math on their own — GaryMarcus · 2026-07-22
- Gary Marcus says LLM math skills are like knowing only a car’s engine size — GaryMarcus · 2026-07-22
- AI may make digital work infinitely leveraged while offline life gets more human — illscience · 2026-07-22
- Better AI math could save researchers time by killing false conjectures earlier — prateekj · 2026-07-22
- AI’s economic forecasts are split by nearly a quadrillion dollars by 2035 — bittingthembits · 2026-07-22
- Open source is becoming tech’s soft power, says Kevin Xu — kevinsxu · 2026-07-22