TACITUS NOTES
Published essays from the team on ontology design, grounded generation, policy intelligence, mediation, benchmarks, and what it means to build experimental tools for high-stakes reasoning.
PUBLISHED · 9
newest first
A chatbot is a query interface. A knowledge layer is the institution’s typed reasoning, persisted, contestable, and inheritable. Confusing the two is how policy and political teams end up paying for fluent prose with no provenance.
Institutional reasoning fails at handoff. The analyst leaves; the file becomes opaque. Generic AI stores text. The knowledge layer stores typed reasoning that survives the analyst — because the kernel is fixed, the extensions are logged, and the provenance is mandatory.
The 2024 framing was: hand-design eight primitives and 41 classes. The 2026 framing is: hand-design the kernel of eight; let extensions grow per case. Here is why both moves matter, and why the second one is the harder bet.
Medical ontologies assume an authoritative voice. Legal ontologies eventually do. Telecom ontologies do by design. Policy and political work has none — the file is the disagreement. The system's job is to shape it, not to dissolve it.
We label our products experimental because they are, not as a legal shield. Here is what we mean by it and what we ask of early users.
Three structural failures: time, causality, provenance. Each one a property of transformer architecture, not a tuning problem.
Fisher and Ury split positions from interests in 1981. Forty years later, almost no software implements the distinction. Here is what happens when you do.
Claims are what a party says. Commitments are what they are on the hook for. Conflating these is how careful negotiations become careless ones.
A schema tells you how to store data. An ontology tells you how to argue about it. Conflating the two is how you end up with yet another JSON standard and no shared grammar.
Seen something that should be a note? Written one yourself?
hello@tacitus.me