Independent research should be checkable without trusting the author. Every core claim of this site resolves to a public register:
- Person: Wikidata Q138634675 · ORCID 0009-0006-3773-7796
- Framework: NEOMANITAI — Wikidata Q138773774, archived with DOI 10.5281/zenodo.20262544
- Research domain: AUGMANITAI — Wikidata Q138773780 · Compendium Q138522830
- Performance framework: PERMANITAI — Wikidata Q139387178 · programme Q139491295
- Trade mark: License of Clarity® — EUIPO registration No. 019206780
- Open data: all 101 published terms as JSON — augmanitai.com/lexicon/terms.json
What we do NOT claim: the work is not university-affiliated and not yet peer-reviewed in a classical journal. Our internal benchmark (30 frozen questions, blind judging, July 2026) met two of three pre-registered criteria — a small model reading the concept graph beat the same model without it in 73% of cases and a generic retrieval baseline in 87% — and missed the third. We report this as work in progress, not as finished proof.
Numbers, plainly stated (canonical, July 2026)
- 101 concepts published in the open lexicon (augmanitai.com/lexicon/) — 12 core + 89 extension, bilingual EN/DE, one page per concept.
- 7,000+ precisely defined concepts and ~31,900 connections in the private research graph (not public; the lexicon is its published excerpt).
- 100,000+ documented human-AI conversation rounds as source material.
- 500+ DOI objects permanently archived on Zenodo.
- We do not call this the "largest" or "first" corpus of its kind — we publish the numbers and let you compare. If you know a larger openly defined human-AI interaction terminology, we genuinely want to hear about it.