Machine Learning & AI Group Lab

A Podkey summary of Machine Learning & AI Group Lab, from WWDC 2026.A lot of Apple AI development right now comes down to one practical question: what belongs on the device, and what belongs in the cloud. The big themes here are trade-offs, measurement, and graceful fallback. Faster isn't always simpler, cloud isn't always slower, and the teams that do this well are the ones actually testing their assumptions instead of guessing.On-device or private cloudThe evaluation framework matters more than the debateLatency is more nuanced than people thinkModel updates and version realityBackground execution and the limits of local AICustom models, fine-tuning, and size constraintsMigration paths and PyTorch experimentationGraceful user experience when AI isn't availableThis podcast was created with Podkey. Make your own at https://podkey.fm