Related Reads

Paste an article URL. We'll extract the text, embed it in your browser, and find 5 semantically-similar reads from a curated corpus we pre-embed nightly. Every match links straight to the source — no proxy, no redirect.

How this works

  • Our server fetches the pasted URL (SSRF-safe) and extracts the article body text — nothing else.
  • Your browser downloads a 23 MB embedding model once (cached forever after).
  • The model embeds your article into a 384-dim vector, right on your device.
  • A pre-computed corpus (~100 KB) is ranked against your article via cosine similarity — the top 5 show up above.
  • Nothing about your article goes to a third-party LLM or inference API. Your text never leaves the browser after we hand it to you.

Model: Xenova/all-MiniLM-L6-v2 (int8 quantized). Corpus refreshes nightly from a curated list of ~50 feeds across tech, news, and culture.