By @jonfortt
Publication Date: 2026-05-15 16:41:00
Charles Lamanna, Microsoft EVP of Copilot, Agents and Platform, joins Jon Fortt on multi-model AI orchestration: layering models yields a 15-point research-accuracy jump at lower cost than a single bigger one. They debate ‘token maxing’ as vanity, why human warmth becomes the moat when everyone has the same models, the risk of preserving employee ‘digital doppelgangers’ after they leave, and Microsoft’s ‘auto’ routing to the best model per job.