A Chrome extension just turned your browser into an AI runtime. Gemma Gem loads Google’s Gemma 4 model entirely on-device via WebGPU. No API keys. No cloud calls. Your data never leaves your machine.
The extension hit Hacker News as a Show HN post and pulled 154 points — not because running local models is new, but because running them inside a browser tab is. After the first download (~500MB for the E2B model or ~1.5GB for E4B), everything is cached locally. Kill your WiFi and it still works.
What It Actually Does
Gemma Gem injects a chat overlay on any webpage. Ask it about the page content, have it click elements, fill forms, or run JavaScript. Under the hood, it uses Hugging Face Transformers with WebGPU for inference — the same Gemma 4 models that score 89% on AIME, now quantized to q4f16 and squeezed into a browser extension.
The architecture is clean: an offscreen document hosts the model, a service worker routes messages and captures screenshots, and a content script handles the UI. Six built-in tools — read page, screenshot, click, type, scroll, run JS — give it basic agentic capability. Toggle thinking mode and you can watch the chain-of-thought reasoning as it works.
The Honest Take
This is a proof of concept, not a production tool. Multi-step tool chains are unreliable. The 2B model sometimes ignores its tools entirely. But the concept it proves is significant: WebGPU has matured enough to run a 2B-parameter model inside a browser at usable speed. Fully open source, Apache 2.0, 483 GitHub stars and climbing. The browser-as-AI-runtime era just got its first real demo.
You Might Also Like
- Claude Channels Scores 375 Points on Hacker News Anthropics Play to Replace Openclaw
- Ensu got 328 Points on Hacker News the Privacy Crowd Wants ai That Never Phones Home
- Google A2ui Agent to User Interface Finally a Standard way for ai Agents to Show you Things
- Mcp2cli the Tool That Cuts mcp Token Costs by 99 Just hit Hacker News
- Notebooklm py the Unofficial Python sdk That Finally Gives Google Notebooklm a Real api

Leave a comment