The Settlement: Capability-Centric Enterprise Design — Research Corpus and Framework v4.0

Most enterprise design practice treats the capability map as its central artefact, and most capability maps become wallpaper. This corpus develops a different foundation: the capability as a promise-keeping socio-technical system whose design is governed by a Settlement. The settlement records the bargain that fixed its rights (mandate, capital, change, custody, continuity), each held on explicit terms, settled at identifiable constitution moments, and carrying a priced risk of obsolescence. The research arc runs from a literature review drawing upon six areas (strategy theory, capability-modelling practice, modularity and interfaces, business models and funding, organisation design, and social systems) through four framework generations. Each generation concluded with a stress test: five public-domain organisations (cohort one), then eight edge forms — joint ventures, project alliances, SPV concessions, DAOs, SPACs, federations — chosen to break the framework's assumptions (cohort two). Findings include: outcome-referent capability frameworks are structurally diagnostic, not generative; design authority exists in exactly two zones (open constitution moments in time, interfaces in space); no no-agreement default is neutral — every settlement covertly allocates power to whoever suffers least from disagreement; and capability persistence is usually held outside the organisations that use it. The corpus includes a practitioner translation and a government-domain application covering policy, machinery of government, legislation, PPP and the third sector. The accompanying v4 toolkit — templates, a machine-readable settlement schema, engagement designs and agent skills — is specified in the corpus but is not included in this deposit. All externally verifiable citations were audited against primary sources in August 2026, with the audit trail included. Remaining flags are declared caveats and open research debts. Produced by a human–AI research method: framing and judgment by the author; retrieval, assembly and drafting compute by Claude (Anthropic).