Build agentic app experiences with the Foundation Models framework

A Podkey summary of Build agentic app experiences with the Foundation Models framework, from WWDC 2026.Today’s roundup is really about one big theme: building AI apps that stay useful after the first clever demo. The main ideas here are switching models at the right moment, keeping context under control without losing your place, and using a couple of orchestration patterns that make multi-agent workflows feel a lot less chaotic. And underneath all of it, there’s a very practical reminder: if you change transcript history carelessly, you can pay for it in speed, accuracy, or both.Dynamic profilesComposable instructionsHistory transforms and shared stateTwo orchestration patternsTool-calling modesTranscript changes and performanceWhy evaluation mattersThis podcast was created with Podkey. Make your own at https://podkey.fm