Take the interactive tour ↗

THE PRODUCT, IN ACTION42 seconds · 6 clickable chapters

Click a chapter to jump into the recording. Explanations follow playback.Actual recording · September 2026 index
Your turn

One query. Useful things resurface.

A clickable replay of the “Claude Code” search in the video. This uses fixed excerpts, not a live search of anyone’s private library.

TWEET INDEX / RECORDED SEARCH
ti> Claude Code

Search by topic, then inspect why a saved post is worth revisiting.

A better second lookYou saved it for a reason.

Run the demo to reveal three matches from the recording.

Behind the search

From a saved post to a useful answer.

The video shows the search experience. The repository explains the five stages that make it possible. Select any stage.

Local data. Model calls when needed.

Captures, storage and embeddings are local. Enrichment and smart ranking send relevant content to model APIs; enrichment also fetches linked pages and images. Basic embedding search can run without model API keys.

A walkthrough, not a hosted search service.

The actual product runs locally with Python and Firefox. This page explains the recorded workflow; it does not connect to X, import your saves or access Nachiketh’s database.

Your bookmarks have more to give.

Read the code, follow the setup guide, and search your own saves.

Get Tweet Index on GitHub ↗