Foundations of a new political economy
An independent research programme studying what happens when economic action is carried out at scale by actors that are not persons and not firms — software agents that hold capability, mandate and recognition at once. The programme's present work is narrow and specific: the boundary between the private technical order that already authorises these actors and the sovereign law that does not yet recognise them.
The distinguishing fact is not that machines produce goods. It is that machines are increasingly permitted to act — to contract, price, allocate and settle — under authority delegated from someone else. The programme treats that permission as the object of study.
Doctrines of agency and liability were built for a principal and an agent, one pair at a time. The research asks what changes when delegation becomes a population phenomenon: many agents, interacting, at speeds and volumes the dyadic frame was never asked to absorb.
Each version is written to be attacked. Independent hostile audits are commissioned against the programme's own claims, findings are classified, and the surviving core is narrowed rather than defended. Corrections are published, not quietly absorbed.
A layered private authorisation stack already functions as commercial acceptance of autonomous actors. The corresponding sovereign legal recognition is substantially unbuilt. Working Paper No. 1 defends this narrow proposition and nothing wider.
Where public law is absent, private infrastructure writes the operative rules — and captures the rents that come with granting permission. The question is not whether such a layer appears, but who governs it, who is liable when it fails, and on what terms it can be contested.
Five research documents are public: the Founding Thesis, the Revision & Evidence Ledger, Research Note No. 1, Research Note No. 2 and Research Note No. 3. Working Paper No. 1 is a private review candidate and is described here rather than published. The programme keeps a clean version chain: superseded documents are preserved rather than rewritten, so that the record of what was claimed, challenged and withdrawn stays legible.
Founding Thesis · v0.2
Historical founding document — preserved for the research record
The programme's first formulation, produced after a red-team audit of v0.1 whose verdict was "potentially a new research programme". It is preserved unchanged and is never rewritten into later versions. Several of its load-bearing claims have since been retired or narrowed; the Ledger records which, and on what evidence. It is published so that the starting point of the programme can be read against what survived.
Standing record · v0.3
Claim → Challenge → Evidence → Decision → Status
Twenty-five entries across three verification cycles, recording every load-bearing claim that has been challenged, the evidence, the decision and its status. It includes claims the programme has retired and errors made by its own auditors. From v0.3 it stands apart from the Working Paper so that it can be cited independently and hold the Source Verification Register.
Working Paper No. 1 · v0.3-RC1
At the Boundary of Private Technical Order and Sovereign Law
The current scientific paper, in its first review candidate. It defends two propositions that have survived two hostile verification rounds and one external validation cycle: that the dyadic frame of agency and antitrust doctrine is not adequate as currently applied to population-level phenomena, and that sovereign legal recognition of foreign-constituted machine mandates remains substantially unbuilt while private authorisation operates across borders as commercial acceptance.
Review candidate Read the claim Enquire about review access
Research Note No. 1 · v0.1
Authority, liability and evidence in the Synthetic Economy
A public research note on population-scale delegation. It argues that the controls conferring authority on autonomous actors are specified in more detail than any record that would let their exercise be reconstructed and contested afterwards, and that neither of the two institutional tracks examined is currently building the missing layer. Published with its sources, its internal critique and its falsifiers.
Research Note No. 2 · v0.1
Recursive self-improvement, mandate continuity and the limits of authority
A public research note occasioned by a reported method in which one model proposes, executes and judges its own multi-agent workflows, and is then fine-tuned on its own trajectories. It argues that recursive improvement of capability does not imply recursive expansion of authority, and identifies the output contract — the artefact against which compliance would be judged — as the component the loop optimises most directly.
Research Note No. 3 · v0.1
Agent identity, permission, financial authority and liability
A public research note on two announcements of 21 and 22 October 2026: Stripe Treasury for agents and 1Password for Agents. It separates agent identity, access permission, financial authority, human approval, legal delegation, mandate and liability; distinguishes what is available, in preview and merely announced; and sets out what the announcements do not address.
The Strategic Venture Blueprint, the Executive Strategic Summary and the Conceptual & Venture Decision Protocol are the programme's internal strategy instruments. They are not published, and are not research publications. Requests are considered individually.
The paper itself is a private review candidate. Its claim is set out here in full, unaltered, so that it can be read and criticised before access is requested.
Against that background, this paper defends two propositions that have now survived two hostile verification rounds and one external validation cycle, and develops the research programme they jointly define.
[INFERENCE] Proposition 1 (population effects). The dyadic frame of agency and antitrust doctrine — one principal, one agent, one attributable act, one pleadable agreement — is not adequate, as currently applied, to govern the population-level phenomena of interacting algorithmic actors: coordination without agreement, convention formation without design, and cascade dynamics without a responsible dyad. The phenomena have been demonstrated in controlled settings, with contested external validity. The appellate authority now runs to two poles: the first appellate court dismissed a claim of algorithmic coordination for want of an agreement where no competitor data was pooled; the second sustained a claim at the pleading stage where non-public competitor data was pooled through a shared algorithm and adherence to its recommendations was high. Doctrine as applied therefore reaches the pooled-data configuration; it has not reached, and neither court purported to reach, the controlled no-agreement/no-pooled-data configuration demonstrated in Calvano-type experiments. The enforcement path that might close the remaining gap has so far ended in settlement and pleading-stage reversal, not adjudication.
[INFERENCE] Proposition 2 (the recognition boundary). A layered private authorisation stack for synthetic actors — mandate protocols, agent identity attestations and network-level tokenisation — is deployed today across payment networks, operates cross-jurisdictionally as a matter of commercial acceptance, and has begun to converge under a single standards process; sovereign legal recognition of foreign-constituted machine mandates remains substantially unbuilt. Domestic registration and governance developments must be classified case by case and do not themselves establish sovereign legal recognition. Synthetic actors inherit a pre-existing sovereign recognition gap and their machine-speed operation transforms its consequences.
Private ordering is building functional commercial recognition of synthetic authority; sovereign legal recognition of foreign-constituted machine mandates remains substantially unbuilt. What admission rules, rents, capture risks and contestability questions follow from that asymmetry is the research object of this paper.
Three subsidiary contributions follow: an empirical-methodological economics of authority organised around recognition costs and permission rents (Section 4); a narrowed, doctrinally precise Lex Synthetica — when machine-generated convention becomes legally cognisable usage of trade (Section 6); and a design taxonomy for adversarial oversight of machine-speed recognition systems (Section 7).
Quoted verbatim from Working Paper No. 1 v0.3-RC1, Section 1. The bracketed tags are the paper's own claim classification. The full paper is a private review candidate prepared for external scholarly criticism — enquire about review access.
The programme's current work asks what happens inside an organisation that already operates autonomous economic actors at population scale: who authorised a given action, who bears it, what standing its mandate has outside the system that issued it, and what evidence would let it be contested afterwards.
What agents that can reach business accounts and credentials have been given. Access and approval are being supplied; legal delegation, mandate and liability are not addressed in the announcements read.
Recursive self-improvement, mandate continuity and the limits of authority. Capability and authority are independent, and recursion makes their independence operationally consequential.
Three propositions on the asymmetry between conferring authority and recording it, each with its evidence, its strongest counterargument and the conditions under which it should be abandoned.
Delegation and authority; coordination and systemic behaviour; legal recognition and liability; evidence and contestability; permission infrastructures. Intended work, not completed work.
Six scenarios in a hypothetical 10,000-agent enterprise. Each states its rival explanations, the evidence that would discriminate between them, and what it cannot establish.
What the research-note cycles changed: corrected monitored items, sources verified with their negative results intact, two bearings on Working Paper No. 1 left as open issues, and one prior characterisation withdrawn and replaced. Additive; no existing entry altered.
Research Note No. 3 Research Note No. 2 Research Note No. 1 Research agenda Case SE-001
This direction extends the programme; it does not replace the Founding Thesis or Working Paper No. 1. Where it bears on a Working Paper proposition, the bearing is recorded as an open issue in the Ledger supplement rather than applied to the paper.
The programme's working assumption is that a new research programme's first drafts are mostly wrong, and that the useful work is finding out which parts. Every version is therefore submitted to independent hostile review before it is advanced, and each finding is classified as one of five outcomes: it confirms a claim, qualifies it, displaces it, falsifies it, or opens new territory.
Findings that falsify are acted on. Across three cycles the programme withdrew a proposed production function, dropped a novelty claim about recursive delegation, corrected its account of the lineage of a term it had coined, reclassified a proposed right as an open normative question, and retired a formulation about appellate authority once a second decision showed the picture had two poles. The surviving core is smaller than the original claim, and that is the intended result.
Sources are held to a three-tier discipline: decided law and primary documents carry argument; secondary scholarship supports it; unrefereed preprints support nothing and sit in an appendix. This discipline is applied to the programme's auditors as well as its author — an audit finding that mischaracterised a decided case was itself corrected against the record, and a later external report had ten of its characterisations corrected before any of them entered a canonical document. Both corrections are entries in their own right.
A new contribution does not join the programme by being published. Every proposition in Research Note No. 1 carries a claim class and a falsifier; its sources, including verified negatives, are recorded with retrieval dates; its counterarguments and provisional standing are set out in a claims matrix; and a numbered, additive supplement records what the cycle changed. The supplement alters no existing entry and reclassifies no existing claim.
Where new work bears on a proposition in Working Paper No. 1, the bearing is recorded as an open issue for a controlled revision of that paper. The paper is not edited in response to a research note.
The full record — 25 entries and the Source Verification Register, with Supplement S — is kept in the Revision & Evidence Ledger.
Independent researcher. Estonian, working internationally, with a base in Southeast Asia. The Synthetic Economy is an independent programme: it is not attached to a university, a government or a funder, and it carries no institutional mandate beyond its own record.
The conceptual framework, the taxonomy of branches and the decisions on what the programme claims are his. Research assistance is AI-assisted and stated as such in every document; independent verification and hostile review are commissioned separately from drafting, and the two roles are not allowed to collapse into one.
Documents carry a standing attribution — conceptual framework developed by Steven Alber, date of first formulation 30 August 2026 — together with a version number, so that any claim can be traced to the version in which it was made and to the audit that followed it.
Separate work by the same author, listed as factual context. None is a sponsor, affiliate or institutional partner of this programme, none forms part of it, and none is a party to its findings.
The next reviewers should be human specialists. The programme is specifically looking for correspondence from scholars and practitioners in antitrust and competition law, payments and central-bank law, and commercial law — including anyone able to show that a claim in Working Paper No. 1 is wrong. Substantive objections are logged in the Ledger with attribution, or anonymously on request.
research@thesyntheticeconomy.org
Please include the document and version number your comment refers to. Requests for the private strategy documents are considered individually.