seed
self-modifying webpage, on-device LLM, site in the URL
Last verified:
What is seed?
Seed is a self-modifying webpage that uses an on-device large language model (LLM) to dynamically rewrite and modify itself in real-time. The tool runs entirely on your device using Chrome's built-in window.LanguageModel API powered by Gemini Nano, requiring no API keys and making no network calls to external servers.
Key features include multiple demo pages that showcase the self-modifying capability: a Hacker News viewer, a Pokédex, a rock paper scissors game, a '96 solar system simulation, and a step sequencer. Users can access these demos through the config page where they can select which demo to generate and view.
Seed is designed for developers, AI enthusiasts, and anyone interested in exploring on-device AI capabilities and self-modifying web applications. It demonstrates the potential of edge AI by running inference locally on Chrome browsers that support Gemini Nano, making it ideal for privacy-conscious users who want to experiment with LLMs without sending data to the cloud.
The project is open-source and available on GitHub under the oxedom repository, allowing developers to study the implementation, contribute to the project, or build upon its concepts for their own self-modifying webpage experiments.
seed pricing
Pricing model: Freemium
Free - Seed is completely free to use with no paid tiers. The tool is open-source and requires no API keys, subscriptions, or payments. All processing happens on-device using Chrome's built-in Gemini Nano, so there are no cloud API costs.
seed pros
- Runs entirely on-device with no network calls
- No API key required for usage
- Uses Chrome's built-in window.LanguageModel API
- Powered by Gemini Nano for local inference
- Completely private - no data sent to servers
- Multiple demo applications included
- Open-source code on GitHub
- Hacker News viewer demo works well
- Interactive Pokédex demo available
- Rock paper scissors game included
- '96 solar system simulation demo
- Step sequencer music demo included
- Low latency due to local inference
- No subscription or payment required
- Works directly in Chrome browser
seed cons
- Requires Chrome browser with Gemini Nano support
- Must enable chrome://flags/#prompt-api-for-gemini-nano
- Only works on devices with Gemini Nano capability
- Limited to Chrome's LanguageModel API
- May not work on older devices
- Gemini Nano has limited model capabilities vs cloud
- No cross-browser support (Chrome only)
- Requires specific Chrome flags enabled
Frequently asked questions about seed
What is Seed?
Seed is a self-modifying webpage that uses an on-device large language model to dynamically rewrite itself. It runs entirely on your device using Chrome's window.LanguageModel API powered by Gemini Nano, with no API keys or network calls required.
How does Seed work?
Seed uses Chrome's built-in window.LanguageModel API which leverages Gemini Nano running entirely on-device. The frontend harness allows the webpage to rewrite itself on the edge using the local model, enabling self-modification without sending data to external servers.
Do I need an API key?
No, Seed requires no API key. It uses Chrome's built-in window.LanguageModel API with Gemini Nano, which runs entirely on your device without any external API calls or authentication.
Does Seed make network calls?
No, Seed makes no network calls. All LLM inference happens entirely on-device using Gemini Nano, ensuring complete privacy with no data sent to external servers.
What browser does Seed require?
Seed requires Chrome browser with Gemini Nano support. You must also enable the chrome://flags/#prompt-api-for-gemini-nano flag to use the window.LanguageModel API.
What demos are available in Seed?
Seed includes multiple demos: a Hacker News viewer, a Pokédex, a rock paper scissors game, a '96 solar system simulation, and a step sequencer. You can select which demo to generate from the config page.
Is Seed open-source?
Yes, Seed is open-source and available on GitHub at github.com/oxedom/seed. The repository contains the source code for the self-modifying webpage framework and all demos.
What models does Seed support?
Seed currently supports Gemini Nano (local/on-device), Anthropic, and OpenAI models. The default and primary experience uses Gemini Nano for on-device inference.
Is Seed free to use?
Yes, Seed is completely free. It's open-source with no paid tiers, no API key requirements, and no subscription fees. All processing uses your device's local Gemini Nano.
Who is Seed for?
Seed is for developers, AI enthusiasts, and anyone interested in exploring on-device AI and self-modifying web applications. It's ideal for privacy-conscious users wanting to experiment with LLMs without cloud dependencies.