⛏

Video Mining

Transcripts in, ranked ideas out.

Pulls the transcript (captions first, local Whisper if there are none), writes an overview, then extracts and scores ideas against your business profile.

A briefing, not a reading. There's no legitimate full-text source for in-copyright books, so this is written from the model's own knowledge of the title — strong for well-known business books, and it will tell you outright when it doesn't reliably know one rather than inventing it.

Own the file? Drop a DRM-free EPUB or PDF here — or — and it gets read properly instead of briefed.

DRM'd files (Kindle AZW/KFX, Adobe-encrypted EPUB) are rejected with an explanation. The text is analysed and then dropped; it never lands in the downloadable .md.

Recently mined

Account connections

OAuth

Playlists, liked videos and subscriptions, through the official YouTube Data API. Read-only and revocable at myaccount.google.com/permissions.

Cookies

Watch Later and Watch History — Google removed API access to both in 2016, so a logged-in session is the only way to read them. Export cookies.txt from youtube.com with a Netscape-format cookie extension and paste it below. It's stored chmod 600 on the server and expires every few weeks.

Drop a CSV here or

Any export with a URL or video-ID column works — column order doesn't matter and headers are matched loosely. Views, Age and Topic columns become filters. A plain list of URLs, one per line, works too.

Download IDEA_VAULT.md

Business profile

This text is handed to the model with every transcript. It is what relevance is scored against — the sharper it is, the better the ranking. Markdown.

Tags

One per line. The model may only choose from these.