About

About The Fringe Archive

An independent, searchable index of Danny Jones's YouTube channel — every video summarized, quoted, categorized, and fully transcribed.

What this is

The Danny Jones Podcast brings fringe topics to a wide audience through long-form conversations with scientists, intelligence officers, researchers, and people with extraordinary stories. This library makes every episode searchable. Guests' claims are their own: many are contested, and the library reports them without endorsing them. It's also hours and hours of video. The framework you need is usually a few minutes inside one video — and there's no good way to find it.

The Fringe Archive fixes that. Every video gets a summary, key points, verbatim quotes with timestamps, a category, and its full transcript. You search by problem, skim the summary, ask the AI, and jump to the exact moment.

Built for curious listeners who want to find what was said, and who said it, without scrubbing through hours of audio.

By the numbers

516

Videos curated

5.8K

Quotes extracted

376

Guests indexed

How it's built

Every video goes through the same five-step pipeline:

1 · Enumerate

Pull the channel's full video list; filter out Shorts, clips, and re-uploads.

2 · Transcribe

Fetch the complete transcript from captions, with timing preserved, and store it in full alongside the video.

3 · Read

Process the full transcript — not the description, not the title — as source material.

4 · Write

Draft the summary, key points, and quote selections against a fixed editorial rubric.

5 · Validate

Check quotes against the transcript, verify timestamps, and normalize categories before publishing.

Where AI is used

AI drafts summaries, key points, and quote selections from full transcripts.
Scripts — not AI — handle episode enumeration, transcript fetching, timestamp math, and formatting.
Human judgment sets the editorial rubric, defines the categories, and decides what gets excluded.
Ask the Archive, on the Ask page, uses AI at question time, searching the full transcript of every indexed video and always citing sources with timestamps.

Transcripts come from video captions, which are imperfect — names get misspelled and crosstalk gets garbled. Validation catches most of it; the original video is the final authority.

What's left out

Shorts and clips — the same material exists in the full-length videos.
Sponsor reads and ad segments — cut from summaries entirely.
Small talk and housekeeping — trimmed, not summarized.
Episodes that fail transcription or validation — held back, not published half-done.

Partial exclusions (a trimmed quote, a cut segment) follow the rules in the editorial standards.

Tech stack

Platform

Base44 — hosting, database, and auth.

Frontend

React with Tailwind CSS.

Build tooling

Vite.

Curation tooling

Python ingest scripts for enumeration, batching, and QA.

Transcripts

YouTube caption extraction with timing preserved; every full transcript is stored and searchable.

Pipeline

LLM summarization with structured output, validated before publish.

Source control

Git.

Ownership & disclaimer

The Fringe Archive is an independent index built by fans of Danny Jones's work, named for his upcoming book. It is not affiliated with, endorsed by, or operated by Danny Jones or his guests.

All episode content, names, and imagery remain the property of their respective owners. Summaries and quotes exist to help you find and watch the original material.

Nothing on this site is financial, legal, or medical advice. Claims made on the show are reported as claims.