Revision & Evidence Ledger
Claim → Challenge → Evidence → Decision → Status. Every load-bearing claim the programme has made, including the ones it has withdrawn, and including errors made by its own auditors.
A challenge is resolved into one of five outcomes. It confirms the claim; it qualifies it, meaning the claim survives in narrower form; it displaces it, meaning the substance belongs to work that already exists; it falsifies it; or it opens new territory. Entries are written when the decision is taken, not when a document is published, and are not removed afterwards.
Entries
Covering the review cycle from the Founding Thesis v0.1 through Working Paper No. 1 v0.2, August 2026.
| Claim | Challenge | Evidence | Decision | Status |
|---|---|---|---|---|
| Authority can be written into a production function alongside capital, capability and information. | Independent hostile audit: the formulation does not do the work a production function is required to do. | Existing production-function literature; no specification survived scrutiny. | Withdrawn. Authority is retained as an object of economic analysis, not as a factor in a stated function. | Falsified |
| Recursive delegation between autonomous agents is a novel phenomenon. | Audit: the phenomenon is described in prior work. | Prior literature identified by the auditor. | Novelty claim dropped. Chains of delegation are retained only as evidentiary residue — what a chain leaves behind that can later be proved. | Displaced |
| “Lex Synthetica” names a newly identified body of emergent machine-made order. | Audit: the lineage is older than the claim allows. | Existing scholarship on emergent private ordering. | Term retained but redefined as a question of legal cognisability — when machine-generated conventions become usage of trade under UCC §1-303(c). | Displaced |
| Dyadic principal–agent doctrine cannot handle populations of delegated agents. | Audit: overstated; the doctrine has not been shown to fail at judgment. | The algorithmic-pricing line of cases, including the RealPage litigation, which the Department of Justice settled on 24 November 2025 without the theory being tested at judgment. | Softened to: not adequate as currently applied. The litigation line is identified as the plausible path by which the question closes. | Survives, qualified |
| The recognition layer for autonomous actors has been built privately while its sovereign counterpart has not. | Audit: the framing conflates commercial acceptance with legal recognition. | The layered private authorisation stack in commercial use; absence of a corresponding sovereign instrument. | Narrowed to exactly that distinction: a private stack operating as commercial acceptance, with sovereign legal recognition substantially unbuilt. This is the programme's sharpest surviving finding. | Survives, narrowed |
| Audit characterisation: the claim in Gibson v. Cendyn was allowed to proceed. | Source check during drafting: the characterisation did not match the record. | The case record — the dismissal was affirmed on 15 August 2025. | Auditor's characterisation corrected against the record and the correction logged. Confirmed independently in the second audit round. | Auditor error |
| The EU digital euro and the US GENIUS Act stand in constitutional opposition. | Audit: the opposition is asserted rather than demonstrated. | The two instruments compared layer by layer — instrument, transaction, issuer. | Reformulated as layered legal divergence at distinct layers, not a constitutional conflict. | Qualified |
| A “right to unprogrammed value” exists or is emerging. | Audit: no legal basis offered. | No jurisdiction has legislated it. | Restated as an open normative research question and removed from the paper's load-bearing argument. | Reclassified |
| Composition of synthetic money and credit constitutes a novel contribution. | Audit: the components are established. | Existing monetary and payments scholarship. | Novelty claim withdrawn. Retained as a branch of the programme's map, marked contested rather than open. | Falsified |
| 2026 preprints may support load-bearing propositions. | Second audit round: the first audit had itself over-relied on unverified preprints. | The preprints in question remained unrefereed. | Three-tier source discipline adopted and applied to both author and auditors. Pending preprints moved to a non-load-bearing appendix. | Method change |
| The programme's coverage of the field is complete enough to publish. | Audit: two significant works were not engaged. | UNCITRAL Model Law on Automated Contracting (2024); Kolt, Governing AI Agents (2026). | Both entered as required reading before the next version. Publication route left undecided until they are engaged and the paper is re-audited. | Open |
Rating across both audit rounds was unchanged at the boundary of C+ and B–. Four of six load-bearing claims survived the second round. The programme treats the unchanged rating as a finding rather than a setback: the perimeter was wrong, and narrowing it is the work.
Unresolved problems
- Insurance. Who underwrites loss caused by an authorised autonomous actor, on what actuarial basis, and what happens to the market when that basis is absent.
- Capture of the recognition layer. If permission is granted privately, the permission itself becomes an asset. Nothing currently prevents its capture.
- Enforcement of last resort. Coercion still requires a sovereign. What is enforceable against an actor with no body, no domicile and no assets in the enforcing jurisdiction is unsettled.
A further set of questions is deferred to a separate document rather than answered here. Corrections and objections are welcome and are logged.