LLMs need a way to actually read the web — not screenshots, not raw HTML. I built a Chrome extension that generates a Hybrid DOM Snapshot: readable like Markdown, structured like an accessibility tree. Demo + repo inside.