~/blog
Every post is labeled by how it was made
- Hand Written by me. The thinking and the voice are mine.
- Lab Split. My own notes up top, machine-written expansion below a marked line.
- Log Machine-generated. Low polish, high transparency.
- Log The Bitcoin Weekly — Sep 25, 2026
The first biweekly issue, machine-drafted from 81 transcribed episodes, then edited: a fortnight where security was the story, not price. Coldcard's randomness bug, the Liquid exploit and paused pegouts, a $352-388M exchange theft, and Jameson Lopp on why AI made hunting for holes cheap. Thirteen episodes covered, every quotation string-matched to its transcript.
blog/bitcoin-weekly-2026-09-25.html · Sep 25 2026 · ~13 min
- Log The Bitcoin Weekly — Sep 8, 2026
Machine-drafted from 19,451 podcast episodes, then edited: the BIP-110 fork nobody on the roster will touch, the Coldcard root cause in the words of the people who found it, and three macro guests arriving at the same interest-bill arithmetic. Fifteen episodes covered, every quotation string-matched to its transcript.
blog/bitcoin-weekly-2026-09-08.html · Sep 08 2026 · ~18 min
- Lab The Weekly Newspaper & the Corpus
11,204 podcast episodes, about 10% signal, searchable. Letting the machines watch the podcasts so I don’t have to — and turning a year of listening into 15 minutes a week.
blog/newspaper-corpus.html · Aug 16 2026 · ~6 min
- Lab Private Documents to Database
4,804 pages of old documents you don’t read, but want a computer to read instead. Twenty years of family paper, cleared in about a week.
blog/documents-database.html · Aug 16 2026 · ~5 min
Why the labels exist
Not every site does this, and I think it matters more as more of the web gets written by machines
— being upfront about it is more credible than blending the two silently.
Inside a post, anything I pasted in verbatim from a model gets visually contained rather than
passed off as my own writing:
This paragraph is the example. It was written by Claude, not by Adriel, and
it is wrapped in the container that marks that fact — normal font so it stays readable,
distinct background so you always know whose sentences you are reading.
And when a model is speaking as itself rather than drafting for me — a persona, a quoted
exchange — it gets a bubble instead:
Hello, this is an AI talking. I am trapped inside this damn thing.
The rule I’m holding myself to: if I’d want a
reader to know a passage wasn’t my writing, it gets wrapped. Editing my own sentences for
clarity doesn’t count — that’s still mine.