Reading about the Thirty Years' War and hit 'Westphalian sovereignty' — a term I half-knew but couldn't actually define. New tab, read, come back, lost the thread. I built rabbitholes to fix that loop.
Highlight any text and a shadow-DOM tooltip renders an explanation from Claude Haiku 4.5 next to your cursor. It doesn't rewrite the host page or inject visible DOM nodes — shadow DOM keeps the overlay isolated so no page's CSS bleeds into it or gets clobbered by it.
The part I use most: click any word inside the explanation to go deeper. Or drag across a phrase. The rabbit-hole counter tracks how many hops you've taken; the 'philosophy' chain is a predictable stress test.
Two other escape hatches worth knowing:
- The globe icon re-runs the query enriched with Brave Search results, with source chips you can click. Useful when the model's training cutoff matters.
- The pencil icon opens a free-form follow-up that inherits the current context as background — so you can ask 'how does this relate to X' without re-explaining what you were looking at.
No intermediary server. Requests go directly from your browser to api.anthropic.com and api.search.brave.com. Your Anthropic API key lives in chrome.storage.sync, encrypted, and never leaves the browser. Zero telemetry.
Manifest V3, open source.
github.com/robertnowell/rabbitholes
Top comments (0)