Every long-form article I read has 3–5 words or concepts I only half-know. Not so foreign I skip them — just unfamiliar enough that understanding them would make the paragraph actually land. The classic fix is a new tab, Wikipedia, read, come back, lose your place, repeat. The realistic outcome is I skim past them and finish the article with a shallower read than I wanted.
I built rabbitholes to fix that for myself.
Highlight any text on any page and a shadow-DOM tooltip renders an explanation next to your cursor using Claude Haiku. Shadow DOM so it doesn't pollute the host page's styles — no layout shifts, no font mismatches. The explanation appears in place, you read it, and you're still on the original article.
The part that made the extension actually useful rather than just convenient: every word in the explanation is itself clickable. Drag across a phrase to select it. You can go as many hops deep as you want without opening a new tab. A counter tracks how deep you've gone — hit a chain long enough and you get a shareable trail.
Two other features I use constantly:
- The globe icon re-runs the query enriched with Brave Search results and adds source chips. Useful when the model's explanation alone isn't enough — you want to know what an actual source says.
- The pencil icon opens a free-form follow-up input that carries current context as background. So if you've already gone three hops deep, the follow-up question knows where you are.
Privacy setup that mattered to me: zero analytics, zero telemetry, no intermediary server. API requests go directly from your browser to api.anthropic.com and api.search.brave.com. Your Anthropic key sits in chrome.storage.sync, encrypted, never touches a third-party backend. Manifest V3.
// The extension just needs your Anthropic key in the options page.
// All requests are direct browser → api.anthropic.com.
// No account, no server, no data collection.
Code and install instructions: https://github.com/robertnowell/rabbitholes
Top comments (0)