How the reports get written

The reports on 1n2.org — the early MB era, ATM scams, sanctions evasion, Michael Saylor, Western Union — share a pretty rigid methodology. The reports look long and discursive but they're built from a short checklist.

The checklist

1. Pick a thread. A topic that the shows have actually covered. Not a guess about what they would have said — a thing they did say.
2. Pull the transcripts. Search brain.db for every episode that touches the topic. Pull the transcript chunks plus the episode metadata.
3. Verify against audio. Whisper transcripts are noisy. Anything that's going to be quoted gets listened to, time-stamped, and confirmed.
4. Quote verbatim, attribute clearly. Every quote is from a named person on a dated episode. No paraphrasing into the third person without flagging it as paraphrase.
5. Cite the episode. The episode number, the date, the timestamp. Readers can verify.
6. Show the data. If a count of episodes or a dollar figure or a percentage is in the report, link the underlying table.

It's not investigative journalism in the New York Times sense — it's closer to the way an archivist would write a finding aid. But the discipline of always going back to the tape, always citing the episode, always quoting verbatim is what makes the reports usable as a source.

Why verbatim matters

The temptation when writing about a long-running show is to summarize. "TBG has been calling Bitcoin a Western Union killer for a decade." That sentence is roughly true and completely unverifiable. The reports refuse to do that.

Instead, the Western Union report quotes Andreas Antonopoulos saying, on TBG #34, in 2014, the specific words he used. The Saylor report quotes the panel reacting in real time to the August 2020 MicroStrategy buy. Verbatim quotes are slower to write and easier to fact-check, and that's the trade.

The Whisper-mangle problem

Auto-transcription is mostly miraculous and occasionally a disaster. The systematic failure modes the reports project watches for:

• Proper nouns. Whisper has never heard of half the people who appear on TBG. "Anrdeas Antonopoulos" becomes "Andre Anono-poulous" routinely.
• Crypto-specific words. "Lightning" gets transcribed correctly; "satoshi" sometimes becomes "Sotoshi" or "the toshi"; "Curio" almost always becomes "Cure-EE-oh."
• Numbers spoken as words. "Two point five million" comes back as "to point five million" enough times to be funny.
• Speaker attribution. Whisper doesn't do diarization well; the reports have to attribute speakers by hand.

The brain.db transcripts get used as a search index, not as ground truth. A search hit gets the report writer into the right minute of audio. The audio is the ground truth.

The VERIFY structure

Most of the reports follow a five-part structure that's evolved over the project:

1. Verbatim quotes from the shows, with timestamps.
2. Evidence from external sources — FBI numbers, court filings, blockchain data.
3. Review of the show's track record on the topic.
4. Inventory of every episode that touched it.
5. Framing for what the reader should make of it.
6. Yield — the part the show could not do, or the part the world hasn't done yet.

It's not religious; some reports skip a section. But the recipe holds.

Why the reports exist at all

The shows are good. The archive is huge. But an archive isn't a story. The reports are the part of 1n2.org that translates "we have it on tape" into "here is what the tape says about a thing."

If a future reader wants to know what crypto people actually thought about Western Union in 2014, or about Adam Back in 2018, or about Bitcoin ATM scams in 2019, they shouldn't have to watch 600 episodes. They should be able to read a report, click the episode link, and verify. That's the whole thing.