Methodology
Tension OS v2.1
A decision-and-governance methodology: the engine and process by which a conforming implementation produces a Decision Plan Brief that satisfies a structural specification: what must be done, in what order, under what thresholds and gates, to produce an auditable decision record. Version 2.1 (2026-04-24) is an errata pass after an independent benchmark of the previous version. It was drafted to a ceiling-quality decision set (an academic methodological posture; a linear-additive multi-attribute value core with explicit value functions) against a stated bar: defensible to a Big-4 partner and to a regulator.
The specification set is being consolidated into a single presentation-grade document so that it can be read in one sitting and tested through the demonstrator below. Until that document is published here, the v2.1 specification set is available on request through LinkedIn.
Conformance cross-walks
The methodology's decision artifacts are cross-walked to the EU AI Act, ISO/IEC 42001 and the NIST AI Risk Management Framework inside Keel's decision architecture. The propagation of v2.1 into the demonstrator was specified in June 2026 and has not yet been applied to it; the demonstrator is labeled accordingly.
Keel: the methodology in practice
Keel is an AI-governability decision system that productizes the methodology: a decision frame coupled to a conformance scope, producing an auditable artifact. The hosted demonstrator implements the MVP specification and predates the Tension OS v2.1 propagation and the ceiling-quality build target, both specified in June 2026. It runs in the browser and makes no network calls to model providers; what it loads is stated on the disclosure page.
AIR v0.5.5: structural modeling before the specification
AIR (AI Research Infrastructure Design System) is a deterministic structural-modeling engine. It takes a research infrastructure's choices, such as intent, governance model, oversight level, coordination, observability, data locality, access and collaboration scope, and derives the constraints, properties and tensions those choices imply: an ungoverned-autonomy configuration, a compliance-observability gap, a rigid cross-boundary control. It is the modeling that preceded the specification of the multi-institution system, and its structural layer was later folded into SyncOrch's strategy stack alongside Keel. It is hosted here as built, with its own key disclosure.
Documentation discipline
This site's build practice has run under a written protocol since before the evaluation instrument was built: corrections by supersession rather than silent edit, a session-close package for every working session, and a defect-and-change register kept across build sessions. The three-consecutive-identical-green-runs gate was adopted after Keel v3 shipped and has governed every build since, including the instrument's. The register on this site is the same practice applied to the instrument; the decision map is the same practice at the specification stage, in March 2026.