William James Sidis wrote the Book of Vendergood in 1905 at age seven. It documented a constructed language built from Latin, Greek, German, and French roots, with a base-12 numeral system and an eight-mood verb system that included two moods he invented himself — one for desire distinct from uncertainty, one for action performed under external compulsion rather than choice.
The book was never widely circulated. The language never found speakers.
We've been reconstructing it. Two volumes cover the historical record, grammatical reconstruction, and a four-dimensional article system. The reconstruction extends his eight moods to twelve — adding evidential (reported vs. observed knowledge), counterfactual, intentive, and habitual — bringing the mood count into alignment with his base-12 numeral system.
The AI application is why we started: a mood system that grammatically distinguishes direct observation from hearsay, desire from commitment, compulsion from choice maps cleanly onto agent epistemic state. The evidential mood in particular addresses something most agent architectures handle poorly — an agent that can mark "I was told this" separately from "I observed this" can reason about the reliability of its own knowledge base.
The reconstruction is at github.com/MMAI-LLC/libro-vendergood. The Dev.to post linked above has working examples and the full mood table.
Sidis also predicted black holes in 1925 in a book discovered in an attic in 1979. He was early to most things.
William James Sidis wrote the Book of Vendergood in 1905 at age seven. It documented a constructed language built from Latin, Greek, German, and French roots, with a base-12 numeral system and an eight-mood verb system that included two moods he invented himself — one for desire distinct from uncertainty, one for action performed under external compulsion rather than choice. The book was never widely circulated. The language never found speakers. We've been reconstructing it. Two volumes cover the historical record, grammatical reconstruction, and a four-dimensional article system. The reconstruction extends his eight moods to twelve — adding evidential (reported vs. observed knowledge), counterfactual, intentive, and habitual — bringing the mood count into alignment with his base-12 numeral system. The AI application is why we started: a mood system that grammatically distinguishes direct observation from hearsay, desire from commitment, compulsion from choice maps cleanly onto agent epistemic state. The evidential mood in particular addresses something most agent architectures handle poorly — an agent that can mark "I was told this" separately from "I observed this" can reason about the reliability of its own knowledge base. The reconstruction is at github.com/MMAI-LLC/libro-vendergood. The Dev.to post linked above has working examples and the full mood table. Sidis also predicted black holes in 1925 in a book discovered in an attic in 1979. He was early to most things.