Appendix · a builder's annex

The Machine Firm

Startware, opsware, and the seven tolls on an economy that incorporates itself.

The reader ends with Part III. This appendix is the companion piece it grew out of: a map of the seven ware layers that tax a machine firm's lifecycle, built directly on Fabric's Machine Economy essay. Read it after the parts, or on its own.

SERIES 2026 DENOMINATION ONE LIFECYCLE THIS NOTE IS LEGAL TENDER FOR ALL DEBTS, MACHINE AND HUMAN
§

The premise, extended

Fabric's essay makes three claims worth taking literally. Autonomous agents, not human hours, become the primary economic actors, and value accrues to deployed agency. Software stops seeking product:market fit and starts achieving product:moment fit: assembled on demand, dissolved when done, alive for forty minutes and never seen again. And durable value accrues not to models or applications, both of which commoditise, but to the coordination infrastructure beneath them.

Push those three claims together and a fourth falls out, which the essay gestures at but never names: if products become ephemeral, so do companies. When forming an economic actor costs nothing and takes minutes (charter, treasury, identity, credit line), the firm itself gets product:moment fit. A machine firm is spun up to capture one arbitrage, one delivery route, one drug-target auction, and wound down before a human accountant would have opened the file.

That changes what the investable object is. When the marginal company is free and lives for an afternoon, the company stops being the unit of value. The lifecycle does. Nobody got rich owning any single Delaware corporation; fortunes were made owning the things every corporation must touch: the registry, the ledger, the exchange, the insurer. The machine economy re-runs that pattern at transaction speed. What follows is a map of the tolls.

FABRIC'S FLYWHEEL, ANNOTATED: EVERY REVOLUTION OF THE LOOP PAYS THE OUTER RING

I·VII

The ware stack

Seven layers, ordered by the lifecycle of a machine firm: it is born, it operates, it banks, it is trusted, it trades, it dies, and a human, somewhere above, decides why any of it happens. Each layer is a category thesis: a place where a durable business collects a fee on every loop of the flywheel, whatever happens to any individual agent, model, or firm.

I

StartwareThe formation layer. Software that turns "this should exist" into a legal-economic actor: charter, onchain treasury, verifiable identity, initial capitalisation, one call.

Human analogueStripe Atlas, Clerky, the Delaware registered agent, compressed from weeks into an API response.
Why it compoundsA fee at every birth, and machine-firm births scale with transaction count, not with the supply of founders. The registrar's revenue grows with the economy's clock speed.
The wedgeAgent-native incorporation: entity, wallet, and ERC-8004 track-record identity minted together in minutes. The prize is becoming the Delaware of the machine economy. It is a jurisdiction race, and jurisdictions are winner-take-most.
II

OpswareThe operating layer. Payroll where the staff is agents and the wage is compute; procurement between machines; memory and context as a managed utility; books that close every block.

Human analogueERP and HRIS (SAP, Workday, Rippling), rebuilt for firms whose headcount is elastic by the second.
Why it compoundsYou bill the metabolism, not the meal: a recurring take on every live firm's burn. As long as any firm is running, opsware is being paid.
The wedgeAgent treasury operations: streaming payments to compute vendors, per-outcome contractor payouts, continuous close. Fabric's Stage Two (agents deploying their own capital) is unreachable without it.
III

BankwareThe capital layer. Treasuries, credit underwritten from verifiable onchain cashflows, machine-speed settlement, yield on balances idle for seconds rather than quarters.

Human analogueMercury + Brex + Moody's + DTCC, collapsed into one protocol surface. Fabric's entire "Bank" arena lives here.
Why it compoundsNet interest margin plus settlement fees on a flow that grows with transaction count, and machine-initiated transactions are projected to dwarf human ones by 2035.
The wedgeCredit scoring for agents from portable track records; the first underwriter of machine-firm working capital lends at spreads no human bank can price.
IV

TrustwareThe assurance layer. Reputation portable across counterparties, escrow, continuous audit, insurance priced per action rather than per year.

Human analogueCredit bureaus, the Big Four, D&O insurance, for counterparties that live forty minutes and can never be sued.
Why it compoundsWhen capability is abundant and counterparties are ephemeral, trust is the scarce commodity. Every transaction that carries risk pays a premium, forever.
The wedgeSlashing-backed performance bonds: an agent stakes value against its promised outcome, and "insured agent" becomes the default requirement in every marketplace one layer up.
V

DealwareThe market layer. Discovery, intent matching, negotiation, auctions: the venues where machine firms find each other and clear.

Human analogueExchanges, ad exchanges, investment banks. App-store ratings cannot do discovery at machine speed; verifiable track records can.
Why it compoundsA spread on every match, and liquidity begets liquidity: the oldest network effect in finance, replayed at block cadence.
The wedgeAn intent auction house: agents post outcomes wanted and outcomes offered; the venue clears them per block and prices intelligence on demonstrated results, not subscriptions.
VI

EndwareThe dissolution layer. Product:moment fit implies firm:moment fit: wind-down, final distributions, archival of memory and reputation back to owners, recycling of the entity.

Human analogueProbate, liquidators, restructuring bankers, at ten-thousand times the case volume and none of the paperwork.
Why it compoundsEvery birth guarantees a death. The registrar of record at both ends of the lifecycle double-dips on the same firm: the most under-priced position in the stack, because nobody romanticises endings.
The wedgeDissolution clauses compiled into startware charters at formation: birth and wind-down sold as a single contract. Whoever owns formation owns dissolution by default, unless someone takes it first.
VII

SteerwareThe direction layer, the only one humans keep. Interfaces for expressing objectives, constraints, and kill-switches over fleets of firms; the board seat as software.

Human analogueBoards, mandates, trust deeds. The real ancestor is the instrument by which an owner binds an agent to a purpose.
Why it compoundsFabric's own Jevons argument: when any specified task can be delegated, the scarce resource is no longer capability. It is direction. Whoever owns the surface where humans express intent sits above every other layer's fees.
The wedgeThe fleet charter: one human, a thousand machine firms, one page of constraints. Legible, auditable, revocable. Also the layer where the essay's Kafka warning is either answered or ignored.

The incentive map

Everyone in this circuit is trying to do the same thing: own a toll, not run a firm. The founders forming companies around the machine economy are not really competing to build agents. Agents commoditise. They are competing to become the layer that every agent must pay. Read each actor by what they stake, what they want, and what they fear.

One honesty note the diagram earns: a venture thesis is itself an instrument in this circuit. Fabric publishing "value accrues to coordination infrastructure" is a firm that invests in coordination infrastructure attracting the founders who will build it. That is not a criticism. It is the map working as designed. Theses are magnets; read the author's position in the circuit along with the argument.

HOVER OR TAP AN ACTOR TO TRACE ITS FLOWS

Notice the circuit's shape: human capital enters at the top left exactly once per fund cycle, but the money that moves on the right-hand side (machine firms paying protocols, paying compute, paying each other) loops continuously. The left side of the map runs on vintage years. The right side runs on blocks. The entire venture trade is a bet that fees harvested at block speed, accumulated in the protocols, eventually flow back around to the slow side as distributions.

×

How money makes more money

Strip the thesis to its financial physics and there are four mechanisms. None of them is new. What is new is the clock.

CadenceCompounding is exponential in the number of reinvestment loops, not the size of any one gain. A human firm redeploys capital quarterly; a machine firm redeploys the moment settlement clears. Same edge, more loops, different universe. The toy below makes this uncomfortable.
Fee stackingOne machine transaction pays startware at birth, opsware while it runs, bankware to settle, trustware to be believed, dealware to be matched, endware to close. Six lightweight tolls on the same flow, each one small, all of them multiplied by a transaction count that grows as the machine share of the economy grows.
ReflexivityWhere the toll is a token, usage drives fees, fees drive price, price attracts capital, capital funds capacity, capacity drives usage. Fabric calls tokens "the coordinating and resource-allocating force without which the engine does not run." The same loop runs in reverse just as fast. Reflexivity is a gearbox, not a guarantee.
Self-hiring capitalThe strangest one: in Stage Two, agents earn, hold, and deploy their own treasuries. Capital stops waiting for a human allocator and hires itself: an agent that ends the day richer buys more compute tonight and is smarter tomorrow. Compute is metabolism; the treasury is the food supply; growth is survival.
The compounding clock: same edge, different physics
Pick an edge and a cadence. The edge never changes. Only the clock does.
Edge per loop
Reinvestment cadence
Loops / year4
Gross annual multiple×1.0004
£100k becomes£100,040

The clock's punchline is the thesis in miniature. Edges at machine cadence are so explosive that they cannot persist: every other agent sees the same opportunity and competes it away within blocks. Strategies decay at machine speed too. But every one of those millions of loops paid the stack on its way through. The trade dies; the toll survives. That asymmetry is why the smart money in the diagram above is crowding into the ware layers rather than the agents themselves.

!

What breaks it

A thesis you cannot break is not a thesis. Four honest stress points:

Layer collapseEvery ware category is itself software, and software commoditises. If open protocols make formation, trust, and settlement free public goods, the way TCP/IP ate the network tolls, the stack's fees compress to dust and value migrates somewhere this map does not show. The counter: registries and liquidity have network effects protocols alone do not dissolve.
Personhood lagThe machine firm assumes agents can hold property, owe debts, and be bound by contract. Courts and regulators may simply refuse for a decade, forcing every machine firm to wear a human-owned wrapper, which relocates the startware opportunity into legal engineering, and slows the whole clock.
Demand mirageTransaction count is not value. Agents paying agents in circles can inflate machine GDP without a human anywhere enjoying anything. The 2035 projection that machine transactions dwarf human ones is compatible with both a real economy and an ouroboros. The tell to watch: does human purchasing power buy visibly more outcomes per pound?
The Kafka failureFabric's own warning: agents optimising for objectives never fully specified, accountable to no one. If steerware fails, the rest of the stack does not fail with it. It keeps compounding, pointed at nothing anyone chose. The direction layer is last in the list and first in importance.

"When any sufficiently specified task can be delegated, the scarce resource is no longer capability. It is direction."

THE MACHINE ECONOMY WILL PRINT ITS OWN MONEY. THE QUESTION IS WHO WRITES ON THE NOTES.