Trained language models for the AI linter, so a machine pulls what it needs instead of re-reading the documentation itself.
Training version docs-v116-canon-principles-only-agnostic-proven-and-code-languages-only.
A machine reads index.json, checks its Ed25519 signature in index.sig against a trusted key, then fetches only the entries it needs and verifies each file's SHA-256 before loading it. Anything failing verification is refused and the machine reads the documentation itself instead — so a bad or unreachable registry can slow a machine down, but never teach it something false.
| artifact | revision | file |
|---|---|---|
| char.global.bin | rev 14 | brain/char.global.bin |
| char.global.fp | rev 14 | brain/char.global.fp |
| english.global.bin | rev 14 | brain/english.global.bin |
The character brain is the substrate every language module is read through. It is built from a system dictionary, which is why it is served here: a machine without one has no meaning network at all, and no way to build it.