A website that's generated by LLM for each request on the fly
In 2021 there was a famous site this-person-does-not-exist.com, which showcased how generative artificial intelligence could be used to create lifelike images of people that weren't real. This spun out many clones, from cats and cars to resumes and even questions.
This website aims to be a tribute to the original, with a twist that now we have entered an era where LLM output can be generated in obnoxious speeds (1000+ tokes/sec). This allows for e.g. generating entire websites while the website is loading.
To see this website live, go to:
https://thiswebsitedoesnotexist.net
The website is hosted on a free Render.com account, so if the dyno is sleeping it might take a few seconds to load.
You can also run this website locally and hack it as you like. Grab a token from
cerebras.ai and add it to CEREBRAS_API_KEY environment variable (or to .env
file). Then run:
npm install
npm startThis installs all necessary dependencies and starts serving the website at http://localhost:3000/
This website is deployed automagically on push to main via render.com CI.
This project is a website that:
- Generates the whole website from scratch every request
- Has a simple throttling and rotating for occasional spikes to avoid exhausting cerebras.ai quota
- Just a proof of content created to inspire others to build something based on it
- Inspired heavily by https://www.val.town/x/stevekrouse/cerebras_coder
This repo is meant as a proof of concept, but any help is of course welcome. If you'd like to contribute, please fork the repository and use a feature branch. Pull requests are warmly welcome.
The code in this project is licensed under MIT license.
