AI is kinda everywhere!

Everyone is shipping AI features or Agents, but few of those systems can reason about the complex reality of performance data.

This session is a deep dive story into the journey of building a Performance AI assistant and agentic workflows on top of a fork of Chrome DevTools

PerfPlanetCal is a curated calendar of web performance events, conferences, and meetups around the world. Through event listings, PerfPlanetCal offers insights into web performance topics, including performance optimization techniques, browser performance tools, and user experience metrics. Developers and performance enthusiasts can discover opportunities to network with peers, learn from industry experts, and stay updated with the latest trends in web performance.

Web Performance Calendar

Building a performance-specialized AI agent requires teaching LLMs to interpret complex telemetry data from Chrome DevTools trace files. The journey involves segmenting trace data into focused problem spaces, using specialized sub-agents with narrow objectives to overcome context window limitations, and implementing workflow-based architectures inspired by Mixture of Experts. The Model Context Protocol enables orchestrating agents to delegate performance analysis to DevTools' internal agents, creating automated workflows for profiling and course correction. Success depends on careful context engineering, keeping sessions focused, and maintaining human verification of AI-generated insights.

Teaching Agents about Performance insights