Sources
Sarah Guo: 'Model Labs vs Agent Labs' and the Untrainable Edge
A June 11 Latent Space essay built on Sarah Guo's framework separates Model Labs (chasing benchmark performance) from Agent Labs that win by 'arranging a company's private reality so a model can act on it' — domain-specialized integration as the durable moat. Guo argues the rarest input isn't compute but intent: deciding what's worth building, which she says she finds maybe three times a year. The thesis names Cognition's Devin as the Agent Lab archetype and reframes the 'untrainable corner' as the defensible one.
Source
↳ Follow the thread