The notes on this site are written with AI. I want to be plain about that, and about why I think it's worth doing anyway.
Source Library is a project to digitize, transcribe, and translate historical texts — tens of thousands of books, most of which have never been available in English. AI does the reading at a scale no person could: it transcribes the pages, translates them, and searches across the whole corpus at once. When something interesting turns up — a first translation, a forgotten diagram, a line of influence between two books three centuries apart — I want to write it down while it's fresh. These notes are those write-ups.
The process
I direct the research: the question, the angle, what's worth pursuing and what isn't. AI models do the digging — searching the library, reading the sources, assembling the evidence — and draft the note. I review and publish. Some notes get heavy editing; some go out close to how the machine drafted them.
The rule that makes this trustworthy is simple: every quotation is checked, by machine, against the scanned page it comes from, and linked to it. You can click any quote in these notes and land on the original page image — the actual ink. That's a standard of verifiability most writing about historical sources doesn't attempt, and it's the discipline that separates a research report from AI slop. Claims without a source under them don't survive review.
Why publish them at all
Partly it's working in public: the library grows every week, and the notes are a record of what we're finding as we go. Partly it's discovery for myself — writing these reports is how I learn what's actually in the collection. And partly it's an experiment in what AI-assisted scholarship can look like when it's done with sources showing. I'd rather run that experiment in the open, labeled honestly, than pretend a human wrote all of this.
So think of these less as blog posts and more as lab reports from an unusual library. They aren't finished essays and they don't claim to be the last word on anything.
Living documents
Because they're research reports, they get revised — when a reader catches an error, when the library acquires a better source, when a claim turns out to need tightening. Every note shows its last-revised date, and the full revision history is public: this whole site is open source, and each note is a file in the repository, with every change on record.
Which also means anyone can propose a correction. If you spot an error — a wrong date, a misattributed quote, a claim you can refute — use the “suggest an edit” link at the bottom of any note, or the feedback link in the site footer. Corrections are the point of publishing this way, not an embarrassment to it.
— Derek Lomas