मानक·Editorial standards
The corpus, the scholar review, the voice spec, the editorial decisions — all logged and visible. You should be able to audit how every reading was written before you trust it.
A—Six commitments
01
The AI cannot generate a verse it did not actually read. If a citation appears in any output, the source verse exists in our retrieval index. Hallucinated citations are not a probability we accept — they are a class of error the architecture rules out.
02
Our scholar is a practising astrologer with formal training in the classical texts. Public-facing material — knowledge base, research, blog — is reviewed before publication. Client-facing material is voice-checked weekly against the published spec.
03
Ten principles, locked at v0.5. Each output across every provider conforms.
04
63 classical texts, listed publicly with verse counts.
05
Every portrait the AI drafts comes with a considered_aspects audit list — what the AI looked at before it wrote. You can disagree with the conclusion; you can also see how it got there.
06
When we are wrong — a misread shastric reference, a mistranslated Sanskrit term, a voice-spec violation that shipped — we log it in the editorial changelog below. Open standards include open mistakes.
B—Editorial changelog