State of client side AI
summary
Claims that web AI runs entirely locally on consumer GPUs (demo on a 5‑generation‑old Nvidia 1070) and that Web AI usage grew from 1M to 1.2B yearly downloads in four years are backed by concrete demos: a Gemma + MediaPipe web LLM agent that parses messy voice input, asks followups, calls site tools to perform flight searches, and adapts queries; a one‑day Web AI DJ using coorro via transformers.js and Spotify; mentions a caching proxy library and emerging WebMCP standards for an agentic internet.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 0:48Welcome to the summit
- 2:20Summit overview and history
- 4:04The future of the web
- 6:12Acknowledgments
- 7:01Defining web AI
- 9:11Agentic behaviors
- 10:48Tool usage in agents
- 12:37Agent runtime architecture
- 14:17Flight agent demo
- 17:12Model caching library
- 18:01DJ agent demo
- 18:29Future of the internet
- 23:18Conclusion
related talks
22mClient side Web AI Agents: Building smarter user experiences for future agentic ...
29mClient side Web AI Agents Building smarter user experiences for a future age
21mUnlocking Client-Side AI in Your Web App
53mAI Didn’t Kill the Web, It Moved in!
24mWeb Apps of the Future With Web AI
33mLocal Models: How to Run AI in the Browser
23mWebAI: Building the next user experience in the browser
39mThe AI PC and the Era of the Modern Web