Pioneer of Browser-Native AI

AI That Runs
Inside Your Browser

ChilledSites pioneered browser-native AI editing with Chrome's Gemini Nano and WebLLM's Llama 3.2. Sub-50ms edits, 100% private, zero API cost — the future of web development.

Try Browser AI How It Works
chilledsites.com/editor — Gemini Nano ready
Local AI — 47ms response
[00:00.012] Gemini Nano initialized via Chrome AI API
[00:00.019] WebGPU context acquired
[00:00.031] Model loaded from browser cache (0 API calls)
[00:00.047] Edit complete — "make headline bolder"
[00:00.051] No data sent to server — 100% private
// Fallback: WebLLM (Llama 3.2) if Nano unavailable
Multi-Backend Architecture

Three AI Backends, One Seamless Experience

Automatically selects the best backend for your browser and use case.

Live

Chrome AI — Gemini Nano

Google's Gemini Nano built directly into Chrome. Zero latency, zero cost, zero data sent off-device. Requires Chrome 127+.

Latency~50ms
Privacy100% on-device
API CostZero
RequirementChrome 127+
Experimental

WebLLM — Llama 3.2

WebGPU-accelerated local inference via WebLLM. Runs Llama 3.2 entirely in-browser with GPU acceleration. No server required.

Latency~200ms
Privacy100% on-device
API CostZero
RequirementWebGPU browser
Standard

Cloud AI — Claude / GPT

Full-power frontier models for complex edits. Falls back to Claude Opus or GPT-4o when browser AI isn't available or the task requires it.

Latency~2-8s
PrivacyEncrypted transit
API CostUses tokens
RequirementAny browser
Comparison

Browser AI vs. Traditional AI

Why running AI in the browser changes everything.

Feature
Browser AI
Cloud AI
Local App
Response Speed
~50ms
2–8 seconds
100–500ms
Data Privacy
100% on-device
Server upload
Local
API Cost Per Edit
Zero
Paid per token
Zero
Internet Required
No
Yes
No
Installation Required
No
No
Yes
Works on All Browsers
Chrome 127+ only
All browsers
Platform-specific
Technical Details

How Browser AI Works

The engineering behind on-device AI editing.

Chrome AI API

Uses the window.ai API introduced in Chrome 127. Gemini Nano is downloaded once (via Chrome's model store) and cached locally. All inference runs in a sandboxed browser process. No network requests during editing.

WebGPU Acceleration

WebLLM uses WebGPU — the modern GPU API for the web — to run Llama 3.2 at near-native speeds. Requires a GPU-capable device and a WebGPU-enabled browser (Chrome, Edge, Safari 18+).

Intelligent Fallback

Browser AI detection runs on page load. If window.ai is unavailable, the editor falls back to WebLLM, then to cloud AI. You always get the fastest available option — automatically.

Optimal Use Cases

Browser AI excels at fast, simple edits: text tweaks, color changes, style adjustments. Complex structural changes automatically upgrade to cloud AI for better results without interrupting your workflow.

What You Can Do

Instant Edits with Zero Latency

The kinds of edits that feel like magic at 50ms.

Text Refinement

Rewrite headlines, adjust tone, simplify copy — instantly, without leaving the editor.

Color Adjustments

"Make the button darker" or "change the background to navy" — done in milliseconds.

Layout Tweaks

Adjust padding, spacing, and alignment with natural language commands.

Style Edits

Font sizes, weights, borders, shadows — all editable via natural language.

Private Brainstorming

Generate copy ideas and content variations without any data leaving your device.

Offline Editing

Continue editing even without internet — browser AI works completely offline.

Try It

Experience AI at Browser Speed

Build a website with ChilledSites and experience on-device AI editing for yourself — 50ms response, zero cost, completely private.

Try Browser AI

Best experience in Chrome 127+ with WebGPU enabled