Behind the reading room
AI transparency.
A transparent account of AI-assisted research and writing, source checks and the limits of human review.
What AI did
This foundation was produced with AI assistance for research discovery, drafting, evidence notes, data modelling, interface design and code. The assistant used the material described in the article source notes to check draft claims. The extent of inspection varies: a note may describe an abstract, indexed excerpt or catalogue record rather than a complete work.
What has been reviewed
Current counts: 1500 written entries; 1500 AI source-checked; 0 human-reviewed; 1500 included in this release. These are separate measures. A page is not described as historian-reviewed or native-speaker verified.
Source checking does not remove the possibility of an incorrect interpretation, an incomplete reference or a missed contradiction. Every article therefore includes its evidence limits and a correction route.
What stays outside release
Unwritten commissions stay in the editorial ledger. Some linked sources may be restricted or become inaccessible; their notes distinguish what was inspected from material that was not. The owner has waived human review; this does not turn AI checks into expert approval. Native-script forms and lesson answers must still be checked against inspected linguistic references.
The reader experience
The deployed reader application has no AI service dependency. Search runs against a build-time index; saved reading and orientation progress are stored locally in your browser. There is no chatbot generating historical answers on demand.