About
A two-hour All-In episode turns into a five-minute read: the structure of the conversation, the key takeaways, and timecodes that jump the player straight to the right second.
What this is
A reference site for the All-In podcast, built for a Russian-speaking audience. For every episode we publish a digest in Russian and English: the conversation split into chapters with timecodes, the key takeaways, the companies and people mentioned, and the hosts' predictions.
An episode page is both a table of contents and a navigation tool: clicking a timecode seeks the embedded player, and the chapter list shows what the episode is made of and what each thread covered.
How the digests are made
Once an hour automation checks the show's channel. For a new episode we pull metadata and subtitles from YouTube, normalise the subtitles into coherent timecoded segments, and hand them to a language model with the instruction to split the conversation into chapters and extract takeaways, quotes and predictions.
The result is checked automatically: chapters must be in ascending order, must not overlap, and must start where there is actual speech in the recording. Only then is the episode published. Machine-made digests can still be wrong — the video itself remains the source of truth, and every page links to it.
Legal position
This project is unofficial and has no connection to the All-In podcast, its hosts, or its rights holders. Names and trademarks belong to their owners and are used only to identify the show being described.
What we publish is our own text: digests, chapters and takeaways are written from scratch, not copied. Direct quotes are used sparingly — no more than three per episode, each a short excerpt linking to the timecode in the original.
Why there is no full transcript
We deliberately do not publish the full transcript of an episode. That is someone else's material in full, and it belongs on YouTube, where it is already available and indexed. Inside the project the transcript is used only as raw material for generating the digest.
Feedback
Found a mistake in a digest or an off-by-a-minute timecode? Let us know and we will fix it. Takedown requests from rights holders are honoured without argument.