Coding Intelligence, Machine Learning & AI Group Lab

A Podkey summary of Coding Intelligence, Machine Learning & AI Group Lab, from WWDC 2026.A lot of Apple’s latest AI story is less about flashy promises and more about the plumbing that makes real apps work. The big themes are portability, privacy, and evaluation: how to swap models without rewriting everything, how to manage tiny versus bigger context windows, and how to test these systems like actual software instead of wishful thinking. There’s also a very clear message here that local AI is graduating from demo status. Between on-device agents, MLX scaling across multiple Macs, and tools for UI testing and coding workflows, this is starting to look like a production stack, not a science fair project.A single doorway to different modelsContext windows are still very real constraintsHow model handoffs handle privacyKeeping within the windowEvaluation before vibesLocal AI is becoming a real product layerAgents that write code and test interfacesGuardrails, refusals, and what they meanThis podcast was created with Podkey. Make your own at https://podkey.fm