Current Capabilities

NTIS has a stabilized deterministic foundation for representing and comparing curated reasoning records. These are engineering capabilities, not validated measures of analytical quality, cognition, or evidentiary strength.

Provenance-preserving source representation

NTIS preserves deterministic source identity, source time, metadata, exact source spans, and contribution or attribution records. It distinguishes a sender, subject, contributor, and proposition holder rather than assuming they are the same person.

Attribution and proposition observation

Source-grounded observations can represent proposition holder, stance, polarity, modality, qualifiers, status, and temporal scope. This makes the conditions around a statement explicit without treating a linguistic guess as a final interpretation.

Explicit adjudication

Candidate artifacts can be reviewed explicitly. Adjudication records conflicts, refusal, supersession, and correction history without erasing earlier decisions. Canonical observations are curated outcomes of that process.

Reasoning-state representation

Curated ReasoningState records contain exact-span claim, premise, evidence, and conclusion elements, plus explicit source-grounded structural relations. Semantic alignment is accepted explicitly; it is not inferred automatically.

Deterministic longitudinal comparison

The Recursive Comparison Engine (RCE) performs eligibility-checked, directed pairwise comparison of accepted reasoning states. It reports structural transitions, including retained, introduced, changed, reversed, missing, not-applicable, and not-comparable.

Experimental candidate perception

Automatic extraction, structural parsing, and learned classification may produce unreviewed candidate artifacts with their own provenance. They may help with bounded tasks, but they do not create canonical state, determine analytical correctness, or turn model confidence into analytical weight.

Current boundaries

NTIS does not currently provide validated stylometric drift measurement, cognitive-condition or burnout detection, reasoning-quality or evidence-strength scoring, automatic semantic alignment or relation extraction, automatic comparison context, historical baselines, prediction, or production deployment.