The web, re-rendered for you.
Enter any URL and this tool fetches the page, strips it down to its content, and asks an LLM to rebuild it from scratch with your preferences applied. Dark mode, minimalist, 2000s-themed; whatever you want.
The web was supposed to be for the reader. Early browser proposals imagined user agents that would present content however you wanted — your fonts, your layout, your rules. HTML described meaning, and your browser decided how to show it.
That vision didn't survive. Designers wanted pixel-perfect control and CSS gave it to them. The browser became a runtime for other people's designs, and the user lost the ability to shape their own experience.
LLMs make it possible again. Instead of trying to restyle arbitrary CSS, we can just ask a model to re-render the content from scratch with different constraints.
This runs on fast, affordable models. Pages would look significantly better with larger models but that would be slow and expensive.
This is a proof of concept: the idea works, and it'll only get better as models get cheaper and faster.
1. Fetch
We scrape the URL and clean the HTML down to semantic content — no scripts, no ads, no tracking.
2. Render
An LLM rebuilds the page as clean HTML+CSS, applying your style and accessibility preferences.
3. Stream
The result streams into the viewport in real-time. Results are cached so revisits are instant.
4. Navigate
Click links to browse between pages. Each page gets re-rendered with your preferences.