The Check and the Firm: Foundations of Post-Watershed Business Practice

Mark Pesce · University of Sydney · July 2026

Abstract

The check replaces the transaction as the primitive unit of analysis. A post-Watershed business converts its workflow into checkable processes run by self-healing loops of generate, check, and repair, held honest by a constitution that governs who may alter the checks. The conversion forces a new discipline, verification design, with four layers: boundary work, specification, harness construction, and loop governance. It also rewrites the theory of the firm. Coase's transaction costs, Williamson's hold-up, Hart's incomplete contracts, and the Alchian-Demsetz metering problem each name a friction that checkable process dissolves or relocates, so firm boundaries redraw around what resists verification. Competitive advantage migrates to the two assets that cannot be copied: the non-mintable hierarchy, and the harness record, which is copyable in form but not replicable in time. The paper derives the successor to the corporation - a human being, a loop, a credit line, and an indemnity policy priced continuously on the harness record, with the insurer as de facto regulator - and states what the framework predicts.

1. Introduction

Foundations of Post-Watershed Economics established the ground condition: the marginal cost of cognition has collapsed, infrastructure mints tokens, harnesses spend them seeking alpha, and alpha can only come from what tokens cannot buy.[1] Defending the Loop supplied the mechanism that makes the ground condition usable: verification converts abundant, unreliable cognition into compounding, trustworthy work, and the human contribution concentrates in the specification.[2] Between the economics and the mechanism sits an unanswered question, the one now forming in boardrooms that have not yet found words for it:[3] what does a business actually look like on the far side of the Watershed?

This paper answers with a falsifiable proposition. A post-Watershed business is one whose processes carry their own evidence of correctness. Checkability confers advantage large enough to force conversion; conversion costs less than the advantage is worth. The advantage compounds through the self-healing loop, in which every process failure arrives as a repair signal rather than an incident. From that proposition the paper derives a practice, a revised theory of the firm, a market structure, and an institutional form to succeed the corporation.

Section 2 states the scope of the claim and the constitution that keeps weak checks honest. Section 3 develops verification design as a four-layer discipline. Section 4 works a conversion in full, in a deliberately unglamorous industry. Sections 5 through 7 revise the theories of the firm, of competition, and of markets. Section 8 derives the successor form and its insurance economics. Section 9 sizes the teams. Section 10 describes the profession, and Section 11 states predictions and limits.

2. Checkability and Its Constitution

Scope honesty first. Strict formal verification, machine-checked proof against a formal specification, covers a narrow band of processes with closed-form correctness conditions. Most business workflows will never reach it. It will run instead on invariants, property checks, reconciliations, statutory calendars, and statistical acceptance tests: the middle rungs of the ladder of judges.[2] Checkability is the weaker property, and it is sufficient. The thesis claims checkability at scale, borrowing the discipline of formal methods without claiming its full rigour for every process.

Weak checks are not safe by default. The companion paper's central warning applies with full force: a check below the top of the ladder is a corruptible judge, and a self-healing loop running against corruptible judges can heal the wound by redefining health.[2] What proof achieves through the nature of the judge, checkable business process must achieve through constitution. Two rules suffice.

Separation of powers. No agent that optimises against a check may author or amend that check. Check authorship and check satisfaction are quarantined roles, structurally, not procedurally: the quarantine is itself a checkable property of the loop.

The one-way ratchet. Checks may tighten without ceremony. Loosening a check, retiring one, or widening an envelope requires a human signature, recorded in the lineage. The tests themselves ratchet, exactly as the artefacts do.

Under this constitution, the lower rungs become usable at machine speed. Without it, conversion manufactures Goodhart machines: operations that optimise their measures while abandoning their purposes.[4] Whatever the constitution costs in ceremony, it buys the entire difference between a self-healing loop and a self-deceiving one.

3. Verification Design

Verification design is the discipline of designing verifiable business practices. Its nearest ancestor is not software testing but mechanism design, the economists' craft of arranging rules so that self-interested play produces the intended outcome; the constitution of Section 2 is mechanism design applied to loops.[5] The discipline has four layers, each a distinct skill.

Boundary work. Deciding which processes get formalised and which stay judgment-bound. This is the scarcest skill, because cheap checks colonise everything checkable and starve everything that resists measurement. Done badly, boundary work produces Goodhart machines. Done well, it is the strategic act of the entire conversion. Every specification exists inside a boundary someone drew first.

Specification. Translating intent into checkable properties. Requirements engineering is the ancestor discipline; this version has teeth because the specification executes. The specification author needs domain fluency, logical training, and the interviewer's craft of extracting what a principal means rather than what the principal says. The evidence that this skill is scarce, in machines as in humans, is now documented: models routinely produce correct artefacts while failing to state faithfully what the artefacts were for.[2]

Harness construction. Harness engineering is the ancestor practice;[6] every harness in a checkable process requires four components: a definition of done, a generator of candidates, a scorer, and a repair signal. The governing design principle is heterogeneity: choose checkers whose failure modes are least correlated with the generator's failure modes, and with one another's. Most harnesses will have nothing to do with code. Designing them for law, logistics, care, and compliance is a design project on the scale of industrial engineering's first century.

Loop governance. Repair policies, escalation thresholds, drift detection, and the standing question of when a human enters the loop. Governance is where the constitution of Section 2 is enforced, where identity and liability attach, and where the human moves from 'in the loop', approving actions, to 'on the loop', holding boundaries.[2]

A specification plus a harness converts tacit process knowledge into a transferable, auditable, sellable asset. Verification design is asset creation. It moves items across the non-mintable hierarchy: what was trust-bound becomes mintable, which changes what remains scarce.[1]

An objection arises from within this corpus and must be met here. Foundations predicts that the Bitter Lesson absorbs the harness layer into the model: harnesses are spoons.[1] If harness construction decays, how can it anchor a discipline and an asset class? The objection fails twice. First, the harness record survives harness absorption, because evidence outlives tooling. What the insurer prices, what the acquirer runs diligence against, what the court subpoenas, is the record of checks demanded, run, passed, failed, and repaired, not the scaffolding that produced it. Second, boundary work and loop governance resist absorption for a different reason: they are liability acts rather than capability acts. A model can draft a check. It cannot sign one. Skills consolidate; signatures do not.

4. A Worked Conversion

Consider conveyancing, the legal mechanics of property transfer: a rule-bound, document-heavy, liability-attached industry, already carrying mandatory professional indemnity insurance, and already half-digitised, since more than ninety per cent of Australian property settlements now complete on a digital exchange.[7] It is deliberately unglamorous. The Watershed arrives for the unglamorous first, because the unglamorous is checkable.

Boundary work. What stays judgment-bound: advising an anxious first-time buyer, negotiating an unusual easement, the client conversation where what is said diverges from what is meant. What converts: searches, standard contract review, adjustment calculations, settlement figures, trust accounting, lodgement, statutory deadlines.

Specification. The correctness conditions of a matter are statable as properties. Title transfers clear of undisclosed encumbrances. Every disclosure in the contract reconciles against the searches. Rates, taxes, and adjustments compute to the contract's formula, to the cent. Funds flows balance. Statutory time limits are met with stated margins. Identity is verified to the prescribed standard. Duty is correct against the current schedule. None of this requires proof theory. All of it is checkable.

Harness construction. Agents generate the candidate artefacts: contract reviews, requisitions, settlement statements. The scorers are heterogeneous by design: search results cross-checked against contract disclosures; arithmetic invariants on the trust ledger; an independent duty calculation reconciled against the revenue office's own calculator; document diffs against the precedent bank; a statutory calendar that cannot be argued with. Failure anywhere emits a repair signal, and the loop regenerates the artefact rather than patching it.

Loop governance. Any unresolved requisition escalates. Exchange and settlement authorisations carry human signatures; those are the acts where liability attaches. Drift detection watches the world rather than the work: when the duty schedule changes, when a statute amends, the affected specifications are flagged for rereading. Every check, result, and repair appends to the matter lineage.

The practice that completes this conversion runs one partner where it ran a firm, its checked-failure history documented continuously in the lineage and its true failure rate disciplined by outcomes conveyancing cannot hide: settlements that fall over, lodgements rejected, claims that arrive. Its indemnity insurer no longer prices the partner's CV; it prices the record. When the partner retires, what sells is not goodwill in the old sense but the estate: specifications, harnesses, and the record that proves what they do. The remainder of this paper generalises from this picture.

5. The Theory of the Firm, Revised

Coase asked why firms exist, and answered that market transactions carry costs which hierarchy avoids.[8] Verification attacks the measurement branch of those costs directly: when a process carries its own evidence of correctness, it can be contracted across the market without the trust overhead that made hierarchy necessary. But Coase alone understates how much of the theory of the firm this touches, and the fuller literature turns out to be this framework's ally.

Williamson made the choice of governance comparative: bounded rationality and opportunism, acting on transactions marked by asset specificity, uncertainty, and frequency, decide whether an exchange belongs in a market, a hybrid, or a hierarchy.[9] Verification relieves the measurement and monitoring side of that calculus. It does nothing for bilateral dependency: the hold-up that specific assets invite survives untouched, since a verified process still runs on someone's data, inside someone's relationships, against someone's installed base. Williamson's firm shrinks without vanishing.

Hart, with Grossman and Moore, located the firm in incomplete contracts: states of the world always exceed the drafting, so someone must hold residual control, the right to decide when the contract runs out.[10] In this framework's terms, incompleteness is the specification-world gap, and it is permanent.[2] Residual control attaches to whoever holds the specification. The four duties of the companion paper - ratification, premises, preferences, the setpoint - are residual control rights, operationalised. Hart's theory names precisely the part of the firm that survives this framework.

Alchian and Demsetz located the firm in metering: when production is joint, no one can measure any member's marginal contribution, so a monitor must be hired and paid with the residual.[11] The lineage does not measure marginal productivity, and claiming so would overreach; attribution is not contribution, and shirking hides in quality as well as quantity. What the lineage removes is much of the jointness that created the problem: loop production is separable by construction, decomposed into attributable artefacts, checks, and signatures. The residue of the metering problem retreats to the judgment layers, where the record tells only slowly, with consequences Section 9 develops.

Assembled: verification dissolves the measurement costs, leaves hold-up standing, relocates incomplete-contract logic into the specification, and narrows the metering problem to the judgment layers. Firm boundaries redraw around what resists verification: judgment, relationships, taste. The verified core contracts freely across the market and shrinks internal headcount toward zero; the unverifiable rim becomes the firm. The check - a unit of work exchanged together with the evidence of its correctness, a transaction that audits itself - replaces the transaction as the primitive unit of analysis.

6. Competition and the Record

Strategy's twentieth-century canon assumed advantage could be defended: by position in industry structure on Porter's account, by resources too valuable, rare, and costly to imitate on the resource-based view.[12] Post-Watershed, a verified process replicates at the speed of file transfer: a competitor who obtains the specification and harness obtains the capability. Advantage therefore migrates to what cannot be copied, and there are two such assets, not one.

The first is the non-mintable hierarchy: trust, relationships, networks, physical assets, regulatory position.[1] Moat analysis becomes non-mintable asset analysis, as Foundations argued.

The second, which this paper adds, is the harness record: copyable in form but not replicable in time. It is a stock built by flow, like reputation, and unlike reputation it is evidence rather than impression, provided its origin, integrity, and completeness are attested. Two operations running identical specifications and harnesses differ in their accumulated verification history, and everything downstream of this paper - insurance pricing, due diligence, estate value - prices exactly that difference. A competitor can steal the recipe, and can even exfiltrate the record; what cannot be taken is the standing to claim it, because attestation binds a record to its maker. A stolen inspection history certifies only the kitchen it was stolen from.

Two earlier results locate the record's value precisely. Gresham's Law and the Fungibility of Tokens showed that as tokens commoditise, differentiation migrates to the harness, and that the Bitter Lesson then erodes the harness itself.[13] The migration continues one step further: from the harness to the record and the boundary, the two layers erosion cannot reach. How Much Does a Token Cost? showed that a token's true cost includes oversight, checking, and shielding.[14] Verification design is the systematic reduction of that auxiliary cost. It is the discipline that makes cheap tokens actually cheap, which is why conversion pays for itself: the firm that cannot check at machine speed cannot spend at machine speed.

7. Markets in Operating Capability

When counterparties exchange verified processes, due diligence compresses from months to a harness run. Rather than reading the target's representations, the acquirer executes the target's estate against its own data and reads the results. Representations and warranties shrink to the unverifiable rim, escrow prices the residual specification risk, and the transaction closes at the speed of the slowest check. The first acquisition conducted this way will be reported as a novelty. The hundredth will be the definition of diligence.

Beyond acquisition, markets in operating capability become possible: buying a verified workflow the way software is bought today, with the record travelling alongside the estate as attested history rather than seller's assertion. Boundary work, specification authorship, and harness design become the market's price-setting activities, because they determine what can be traded at all. The discernment problem that limits human evaluation of machine work at scale makes these markets otherwise impossible; the record is what substitutes for the discernment no buyer can exercise.[15]

8. The Successor Form

Joint stock existed because operations at scale exceeded one person's capital and one person's risk. From the Dutch East India Company forward, the corporation pooled both, and Berle and Means documented the price: ownership separated from control, and the agency-cost era began.[16] A loop gives one person corporate-scale output.[1] The territory that required incorporation accordingly shrinks, and what replaces it, wherever the risk fits, is a quartet: a human being, a loop, a credit line, and an indemnity policy.

The credit line is the leg the earlier framing omitted. Loops consume tokens, token demand is scaling super-exponentially, and the practitioner is long harness record, short compute.[1] Someone finances that spread, as factoring once financed receivables: the capability marketplace, the insurer itself, or a compute-credit market growing out of the commodity mints.[13] An operating form has four requirements - labour, capability, capital, and risk transfer - and the quartet covers all four without a share register.

The indemnity policy is what makes the form possible, and the harness record is what makes the policy priceable. The record supplies the frequency side of the actuary's ledger: checked failures, repair rates, escalation patterns, drift events. Severity, exposure, policy terms, and dependence between loops remain the insurer's craft, and the record cannot show what its own checks missed, so independent outcome data anchors the true failure rate. What changes is the information base: premium can adjust continuously as the record accrues, and insurance pricing takes on the character of continuous audit. Embryonic versions already exist wherever conduct is instrumented: telematics motor insurance prices the driving record as it forms; cyber insurance prices the security posture it monitors. Continuous pricing is already established practice; the novelty lies in what gets priced, checked work in place of sampled behaviour.

The insurer becomes the de facto regulator, and the precedents are institutional, not speculative. Lloyd's Register has classed ships since the 1760s: vessels surveyed, rated, and priced for insurance on the class, a private body whose classifications came to govern a global industry, first through insurability and later through statutory delegation as well.[17] Underwriters Laboratories grew from fire-insurance roots, founded by William Henry Merrill with underwriters' backing, and markets learned to demand its mark ahead of the statutes that later referenced it.[18] Ben-Shahar and Logue generalise the pattern: insurers already regulate conduct through pricing and policy conditions, in parallel with the state.[19] The loop economy hands this machinery its richest substrate ever, an append-only, discovery-ready record of everything the operation checked and everything that failed.

The stress test for this section is correlated failure. Insurers model dependence, accumulation, and common shocks as a matter of course; what breaks pricing is dependence that arrives unmodelled. Loops manufacture exactly that: they share model versions, purchased specifications, and harness templates, so a thousand practitioners running the same estate fail together, on the same blind spot, on the same day. That is catastrophe risk wearing retail clothing, and the lesson of 2008, models that materially underestimated correlated and tail risk, applies without modification. The framework's own design principle answers it. Heterogeneity ceases to be an engineering nicety and becomes an underwriting requirement: insurers will demand checker diversity and model plurality as fire insurers demanded firewalls, pricing monocultures until they diversify or die. Reinsurance stratifies above, and some monocultures will simply be uninsurable.

Tail risk then sorts the institutional landscape. Loops whose worst case fits within one person's insurable capacity take human attachment. Loops with catastrophic tails keep joint-stock pooling, which is what the corporation retreats to: not the default form of business, but the special-purpose vehicle for catastrophe. Succession follows the asset logic: the estate transfers by sale, so the form survives its human without requiring a perpetual entity.

The state stays in the picture, repositioned. A successor to the corporation shifts the tax base from corporate income toward personal income and capability-market transactions, and no treasury will watch that migration passively. Licensure regimes will contest ground with the insurers. The direction of public regulation is already legible: Turing Police forecasts a settlement into capability tiers, and observes that harnesses routinely amplify models past whatever tier assessed them - which makes the harness record, once again, the natural object of oversight.[20] The likely settlement is dual governance: public tiers for capability, private pricing for conduct.

9. Tiny Teams

Between the sole practitioner and the joint-stock pool sit tiny teams.[1] The four layers of the practice are distinct skills rarely found in one person, so a team assembles the smallest set of humans that covers the stack, converging on three to five members. That is the present reason. It erodes: models absorb specification drafting and harness assembly on the Bitter Lesson's schedule, and the skill-coverage argument for any particular team size weakens with each release. What does not erode is the signature set. Liability law, not capability, sets the minimum crew: the number of humans a loop requires is the number of distinct signatures its risks demand - check loosening, envelope amendment, settlement authorisation, boundary redraw. Teams are sized by signatures, not skills.

What binds the team is the shared corpus, held under an append-only lineage of specification, implementation, and verification artefacts. Lineage records who built and signed what, so authorship is attributable at the artefact level. Against outside claimants this changes nothing by itself: partnership law makes partners liable for the firm's obligations jointly, and for specified wrongs jointly and severally, and no internal record rewrites a claimant's rights.[21] What the lineage changes is everything behind that wall. Indemnities and contribution between members can follow the record; insurers can price each member's slice of it; and the choice of vehicle, a partnership, an incorporated practice, a federation of contracting sole practitioners, can be made with each member's exposure finally visible. Classical partnership was an act of mutual hostage-taking conducted in the dark. The lineage leaves the hostage-taking in place and turns the lights on. Members hold non-mintable assets individually; the mintable estate is jointly owned and cleanly divisible.

One honesty note. Dissolution becomes a corpus fork, and the fork is clean, but only for the mintable estate. Clients never forked, and clients were always the ugly part of dissolution. The grief migrates to the rim, where it has always lived. The form scales down gracefully to one and dissolves gracefully at the seams it can reach.

10. The Profession

At full scale the practice is a profession: standards bodies, licensure, liability regimes, and an economics of who bears harness failure. Philosophers frame the boundaries. Logicians design the checks. Translators carry intent into specification. Designers build harnesses for every domain of work the checks can reach. Stewards stand on the loop, holding the setpoint, signing the amendments, and answering to the insurer; theirs are the signatures that size the teams of Section 9. Five roles, not five people: the logician and the designer are routinely one person, the philosopher and the steward often the principal, which is how the full stack collapses to the three-to-five convergence. These are functions, not diplomas.

An irony must be named before a reviewer names it. A profession is a guild, and the companion paper argues that proof abolishes guild logic: the artefact that carries its own evidence needs no certificate of origin.[2] The resolution is exact. Licensure attaches where proof cannot reach - boundary work, ratification, governance, the judgment layers - which is also precisely where liability attaches. The profession is the institutional form of the human remainder. Artefacts are certified by their records; practitioners, by their licences; the two systems govern different layers and do not compete.

Every durable institutional form in business history has been built on its bookkeeping. Double-entry made the merchant house auditable; cost accounting made Chandler's managerial corporation legible, the visible hand holding a ledger.[22] The harness record is the bookkeeping of the successor form, and the discipline that produces it stands to this century roughly as accounting stood to the last two: the practice that makes the era's dominant institution possible, teachable, and governable. The institutions that teach business will reorganise around it in due course; how they do so is beyond this paper's scope.

11. Predictions and Limits

The framework predicts observables, roughly in order of arrival. An indemnity product priced continuously on a harness record, with premium moving as the record accrues. Underwriting criteria that mandate checker heterogeneity and model plurality, the firewall clauses of the loop economy. An acquisition in which due diligence is conducted as a harness run, and the compression of representations and warranties that follows. Capability marketplaces listing specification-plus-harness estates with records attached, and a visible price premium for deep records over identical shallow ones. Firm-size compression in checkable sectors - conveyancing, customs brokerage, fund administration, claims processing - measurable in business registry data within a few years. A decline in new incorporations in low-tail sectors, alongside growth in insured sole practice. Contested ground between licensure regimes and insurers over who governs loop conduct, settling toward public capability tiers and private conduct pricing.

The limits are structural, and they locate the human station rather than weaken the claim. Judgment-bound work does not convert, and the boundary between checkable and judgment-bound is itself judgment-bound: verification design cannot verify its own first move. Correlated failure is identified here, not solved; the pricing of loop monocultures is an open actuarial problem, and 2008 shows the cost of underestimating dependence. The state's response is directional, not scheduled. Timing throughout is a shape, not a schedule: gradually, then suddenly.

Coase asked why the firm exists, and the post-Watershed answer inverts his question. The check settles itself. The firm is what remains when the checks have done their work: the judgment that drew the boundary, the relationships no record can hold, and the signature that no loop, at any speed, can supply for itself.

Acknowledgements

Profound thanks to John Allsopp, Alan Eyzaguirre and AJ Fisher, all of whom contributed to my thinking on this topic, and to Philippe van Nedervelde for reviewing the final draft. This paper was composed with the assistance of Claude Fable 5. I remain wholly responsible for any errors that may have crept in.

Notes

1. Mark Pesce, "Foundations of Post-Watershed Economics," The Watershed, April 2026. https://thewatershed.markpesce.com/foundations-of-post-watershed-economics/
2. Mark Pesce, "Defending the Loop: Verification and the Division of Labour in Autonomous Work," The Watershed, July 2026, https://thewatershed.markpesce.com/defending-the-loop-verification-and-the-division-of-labour-in-autonomous-work/
3. Mark Pesce, "Road Runners and Wile E. Coyotes," The Watershed, June 2026. https://thewatershed.markpesce.com/road-runners-and-wile-e-coyotes/
4. Goodhart's law, in Marilyn Strathern's formulation: "When a measure becomes a target, it ceases to be a good measure." Marilyn Strathern, "'Improving Ratings': Audit in the British University System," European Review 5, no. 3 (1997): 305-321.
5. Leonid Hurwicz, Eric Maskin, and Roger Myerson shared the 2007 Nobel Memorial Prize in Economic Sciences for laying the foundations of mechanism design theory.
6. Mark Pesce, "𝜶 and Harnesses," The Watershed, April 2026. https://thewatershed.markpesce.com/and-harnesses/
7. Property Exchange Australia (PEXA), the electronic settlement platform; PEXA reports that more than 90 per cent of Australian property-transfer settlements now complete digitally. https://www.pexa.com.au/content-hub/federal-budget-a-chance-to-strengthen-response-to-housing-crisis/
8. Ronald Coase, "The Nature of the Firm," Economica 4, no. 16 (1937): 386-405.
9. Oliver Williamson, The Economic Institutions of Capitalism (New York: Free Press, 1985).
10. Sanford Grossman and Oliver Hart, "The Costs and Benefits of Ownership: A Theory of Vertical and Lateral Integration," Journal of Political Economy 94, no. 4 (1986): 691-719; Oliver Hart and John Moore, "Property Rights and the Nature of the Firm," Journal of Political Economy 98, no. 6 (1990): 1119-1158.
11. Armen Alchian and Harold Demsetz, "Production, Information Costs, and Economic Organization," American Economic Review 62, no. 5 (1972): 777-795.
12. Michael Porter, "How Competitive Forces Shape Strategy," Harvard Business Review, March-April 1979. On inimitable resources: Birger Wernerfelt, "A Resource-Based View of the Firm," Strategic Management Journal 5, no. 2 (1984): 171-180; Jay Barney, "Firm Resources and Sustained Competitive Advantage," Journal of Management 17, no. 1 (1991): 99-120.
13. Mark Pesce, "Gresham's Law and the Fungibility of Tokens," The Watershed, April 2026. https://thewatershed.markpesce.com/greshams-law-and-the-fungibility-of-tokens/
14. Mark Pesce, "How Much Does a Token Cost?" The Watershed, April 2026. https://thewatershed.markpesce.com/how-much-does-a-token-cost/
15. Mark Pesce, "Relative Superintelligence," The Watershed, June 2026, on the discernment horizon. https://thewatershed.markpesce.com/relative-superintelligence/
16. Adolf Berle and Gardiner Means, The Modern Corporation and Private Property (New York: Macmillan, 1932).
17. George Blake, Lloyd's Register of Shipping, 1760-1960 (London: Lloyd's Register, 1960); institutional history at https://www.lr.org/en/about-us/who-we-are/our-history/
18. On Underwriters Laboratories' founding by William Henry Merrill with fire-underwriter backing, and its governance through certification: https://ul.org/about/our-history/
19. Omri Ben-Shahar and Kyle Logue, "Outsourcing Regulation: How Insurance Reduces Moral Hazard," Michigan Law Review 111, no. 2 (2012): 197-248.
20. Mark Pesce, "Turing Police," The Watershed, June 2026. https://thewatershed.markpesce.com/turing-police/
21. For the standard Anglo-Australian position, Partnership Act 1892 (NSW) ss 9-12: partners jointly liable for firm obligations, jointly and severally for specified wrongs. https://legislation.nsw.gov.au/view/whole/html/inforce/current/act-1892-012
22. Alfred Chandler, The Visible Hand: The Managerial Revolution in American Business (Cambridge, MA: Harvard University Press, 1977).