Millions of Small AIs Could Beat One Giant Model | Andrew Trask

Ensembling-and-routing theory and early evidence suggest many small routed models may economically beat one giant frontier model.

• Ensembles beat monoliths: Trask cites ML theory that a weighted ensemble of best-in-class models is always at least as capable as the single most accurate model, undercutting scaling-law orthodoxy.
• Router economics evidence: When Fable was turned off, Open Router could reportedly match Fable-level quality at half the price by ensembling other models, suggesting a PC/internet-like future of 'millions of models' over one mainframe-style frontier model.

Themes, sentiment and reasons are AI-written paraphrases, checked so they never quote an episode. They may contain errors. Not investment advice. Episodes remain the property of their creators.