[AINews] Open Models, Model Labs vs Agent Labs, and What's Untrainable — Sarah Guo

Short summary

Essay comparing open models versus model labs versus agent labs architectures, exploring which aspects of AI systems remain untrainable or architecturally constrained.

• •Contrasts open-source model availability with proprietary model lab approaches
• •Examines agent lab architectures as a distinct pattern vs pure model labs
• •Identifies theoretical limits on what can be trained vs what requires architectural choices

Generated with AI, which can make mistakes.