Comma for either/or — dharma, courage. Spelling forgiving — corage finds courage.

    Falsafa by the numbers

    Every claim in the launch writeup deep-links to its source. Audit anything.

    Corpus

    3,087,298

    total words across all variants

    view raw manifest

    Eras & languages

    6

    French, German, Kawi, Old English, Sanskrit, Urdu

    view languages

    5

    eras spanned

    view era index

    Eval

    146

    cases run, paper-grade post-patch

    view raw eval JSON

    94%

    mechanical pass rate

    view raw eval JSON

    917

    paragraph citations resolved against the corpus

    audit each one

    4

    verse-marker hallucinations (target: 0)

    why this matters

    Sample size is small (post-anti-cheat-patch). Scoring is deterministic — citation match against expected_works, no LLM judge.

    Benchmark vs hybrid RAG

    Citation-validity rate

    Falsafa forthcoming
    Hybrid forthcoming

    p95 latency

    Falsafa forthcoming
    Hybrid forthcoming

    One-time embedding cost

    Falsafa $0
    Hybrid forthcoming

    Hybrid RAG baseline (apps/baseline/) is in development. Falsafa's approach reads markdown directly through MCP tools, so there are no embeddings to compute. Hybrid figures land when the baseline runs against the same 1,000-question pool.

    Build

    818

    logical chapters

    view manifest

    2,053

    variant entries (translation, transliteration, original)

    view manifest

    76,210

    cited paragraphs with stable hash IDs

    view corpus tree