Ontology and graphs
Represent actors, claims, commitments and time without losing their sources or uncertainty.
Definitions and record models →Section pages
TACITUS research
How should context work for people and AI?
We study the foundations behind PRAXIS: ontology, portable graphs, expert methods and human review. The test is practical: does structured context improve policy analysis, checking and handover? These papers describe research proposals and open questions.
Research that supports the product
Meaning. Portability. Measured usefulness.
Portable context
Carry graph structure, methods and review state across tasks. Test what survives export and reuse.
Context assembly and AI →Evaluation
Compare answer quality, citation fidelity and review effort against simpler workflows.
Methods and reproducibility →Find your research path
A question.
A method. A test.
Follow the idea from its rationale to a concrete design choice and the evidence needed to evaluate it. Choose the question closest to your work.
Path 01 / 04
What must survive when documents become context for an AI system?
Research papers describe design intent. Evaluation protocols identify what still needs to be demonstrated.
External foundations / reading, not integration claims
Sources → context → task → review.
These external works inform the questions we investigate. Citing them establishes neither implementation in PRAXIS nor certification.
Attribution and precise receipts
PROV-O describes entities, activities, agents and derivation. Web Annotation defines quote and position selectors. These are useful vocabularies for explaining source and review records.
W3C PROV-O ↗Time and retrieval
Zep separates temporal knowledge dimensions. GraphRAG distinguishes local entity-focused search from global synthesis over community reports. A later source does not automatically settle a dispute.
Zep temporal graph paper ↗Bounded context for a task
Context engineering asks what belongs in the working packet. Keep constraints, citations, uncertainty and omissions; a larger packet is not automatically better analysis.
Context engineering guidance ↗Selected TACITUS research note · published 2026-06-10 · research position
What it takes to check an AI-assisted assessment
A citation is a starting point. Useful review also needs the source version, the inference being made, the missing evidence, and a record of what a person actually checked.
Publication is not independent peer review or an analytical-quality result.
Further sources and implementation boundaries
- Web Annotation quote selectors — exact source anchors.
- GraphRAG query documentation — local and global retrieval modes.
- MCP specification — resources, prompts and tools; no TACITUS MCP certification claimed.
- Context Language Models — external adaptive-context research. The referenced implementation has a CC BY-NC 4.0 license; no code is copied or integrated here.
Choose the depth your question needs.
Start with a worked distinction. A cited mechanism is not a deployed integration or analytical-quality result.
Practice, evidence and cases
Representation and engineering
Vision and dated history
Revision responsibility: TACITUS Research. Publication and revision dates appear on individual notes; independent peer review is not claimed. Evidence levels and current results.