A WWDC26 session walkthrough on using Apple's Instruments tool to debug and profile agentic app experiences built with the Foundation Models framework. Covers the key challenges of LLM app development (probabilistic output, model-to-model communication, observability), demonstrates a real debugging scenario in a crafting app where a missing tool in DynamicInstructions caused a silent failure, and explains how to use the Foundation Models Instrument's timeline, tree view, and performance metrics (Time to First Token, Tokens per Second, Total Latency) to identify and fix issues. Requires Xcode 27 and latest OS releases.
•14m watch time
131 Impressions