Show HN: Gemma Gem – AI model embedded in a browser – no API keys, no cloud
Google's Gemma 4 now runs entirely in your browser. No API keys. No cloud. No limits.
Why it matters
A significant shift in AI deployment: moving inference from cloud to client-side browsers via WebGPU. This demonstrates practical edge AI implementation and challenges the API-dependent model that dominates today's AI economy.
The key facts
14 to knowGoogle Gemma 4 (2B model) running via WebGPU in Chrome extension
Zero external dependencies - fully client-side inference
Tool-use capabilities: webpage interaction, screenshots, clicks, JavaScript execution
Chain-of-thought reasoning mode enabled
Limitations: unreliable multi-step chains, occasional tool-use failures
Posted to Hacker News with 39 points and 5 comments
Open-source implementation available on GitHub
Gemma 4 (2B) model runs via WebGPU in Chrome extension
Zero external dependencies for agent loop
Tool calling capabilities: read content, screenshots, click elements, type, scroll, JavaScript execution
Includes chain-of-thought reasoning/thinking mode
Limitations noted: multi-step tool chains unreliable, agent sometimes ignores tools
Standalone library potential for experimentation
Published Apr 6 2026 on Hacker News (39 points, 5 comments)
Go to the source
Hacker Newsgithub.com
Publisher excerpt: Gemma Gem is a Chrome extension that loads Google's Gemma 4 (2B) through WebGPU in an offscreen document and gives it tools to interact with any webpage: read content, take screenshots, click elements, type text, scroll, and run JavaScript. You get a small chat overlay on every page. Ask it about…