Decision date: 2026-08-20
Decision: Theory of Freedom is retained as an optional upstream normative
layer, not as the primary industrial or security-capability claim.
Should the stack be reordered as:
Theory of Freedom
→ rights ontology
→ legitimacy / FDK
→ delegation / capability
→ AuthGate execution
→ tools / MCP
…or should freedom-theory remain an independent philosophical contribution while AuthGate is pitched as systems/security engineering?
- Kill-tests on AuthGate authorization (
STATUS.md,WHY_NOT_OPA.md) — authorization is absorbed by Cedar / OPA / Zanzibar. The surviving industrial thesis is purpose/flow governance and non-amplifying composition, not a new moral theory. - Ownership discriminant experiment (
REVIEW_REQUEST_2026-08-10.md§3–4) — the claim that deriving legitimacy from an ownership/consent ontology is a unique contribution was falsified after pre-registration and blind audit. - Pattern across the program — ideas collapsed toward philosophy and survived toward runtime enforcement, distributed systems, and capability management.
| Layer | Role | Public claim |
|---|---|---|
Theory of Freedom (freedom-theory-work, PHILOSOPHY/) |
Optional normative interpretation of why veto-only legitimacy exists | Academic / lineage only; never a product differentiator |
| Legitimacy (ex-FDK; Freedom Formal engine) | Veto-only evaluator (DENY/DEFER); co-equal under lattice meet | Constraint input, not grant authority — see authgate.integrations.legitimacy |
| AuthGate / decision-os-min | Authority, attenuation, PEP, audit | Primary claim — executable AE-1…AE-10 under non-amplification |
| MCP | Standardization surface for tool mediation | Adoption path (not a new RFC) |
Invariant that does not depend on the theory being true:
compose(a, k₁…kₙ) ⊑ a— No Amplification. Constraint inputs (including any legitimacy evaluator derived from any ontology) are veto-only.
- Pitching AuthGate as “the Theory of Freedom made executable” as the main hook.
- Claiming philosophical correctness of A1–A7 as a security result.
- Letting normative vocabulary outrun the ASSUMPTIONS table.
- Keeping
PHILOSOPHY/on thenazariye-azadibranch as lineage documentation. - Using FDK as one veto-only evaluator among others (safety, privacy, budget).
- Academic papers on the theory as a separate track from the conformance / MCP engineering track.
Revisit when: an external reviewer produces evidence that the ontology discriminant holds on a new, pre-registered corpus — or when a standards body requires a normative annex. Until then, this decision stands.