Run-Off, Not Absorption

Four memos have, between them, built the clone, and none of them said how it ends. The Operating Inheritance is what crosses from the incumbent: decisions, not organisation. The Counter-Positioned Clone is where it runs: a separate business outside the perimeter, because inside the perimeter the incumbent's collateral-damage calculation closes it. The structure removes every channel that calculation travels along. The Stewards are formed, not appointed, and the mandate they inherit carries its provenance. This memo is about how it ends, and the first thing to say is what the ending is not.

The intuitive end state is absorption. The clone grows, the parent exercises its purchase right, and the clone becomes the parent — or, in the more ambitious telling, the parent is folded into the clone, the legacy organisation dissolved into the autonomous one. That second version is the one that gets a board excited and the one the corpus has to refuse, because it reverses everything The Incumbent Wrote the Mandate and Never Encoded It excluded. The clone was built without the stack, the contracts and the organisation. Absorbing the parent delivers all three: a workforce the clone has no process for, contracts with service obligations the clone's architecture was never designed to meet, and a regulatory and client perimeter that travels with the legal entity regardless of who owns it. The clone that absorbs its parent has acquired a Legacy Liability. It is the hybrid middle the corpus called the most expensive model available, arrived at from the other side.

Run-off

The end state this memo proposes is run-off, borrowed from how insurers and lenders retire a book they will no longer write. The parent stops originating in the lines the clone serves and services what it has — existing contracts to their term, existing relationships to their natural end — with the organisation it already has, shrinking as the book shrinks. The clone takes new business from the first day, from the parent's competitors first and, as Counter-Positioned by Construction said must be allowed, from the parent's own clients as their contracts come up. Neither entity is transformed. One grows on the new cost base; the other declines on the old one, and the decline is funded by the book it is servicing rather than by a reconstruction it cannot complete.

Run-off is an end state, and an end state has conditions under which it is reachable. The memo names five and claims no more than that they can be met. Existing contracts are served to their term by the parent, so the clone inherits no obligation it did not price. Service continuity for customers who move is the clone's to guarantee at the moment of the move, not the parent's to wind down. Customer consent and regulatory notice, where the line requires them, travel with the customer and not with the inheritance. The legacy book's declining cost is absorbed by the book itself, which means run-off is viable only in lines where the book funds its own servicing to the end. And the parent's incentive to stop originating has to be written into the purchase right, so that a parent that keeps competing for new business in the clone's lines is diluting the asset it holds a right to buy — a proposed mechanism whose scope and permissibility are for transaction-specific competition-law and contractual review, since a restriction on a parent's origination is a different thing from a fence on its clients and still has to be lawful where it is written. Where those conditions hold, convergence by attrition has three properties the argument needs. It is gradual, so the parent's board never faces the vote the corpus says it will not take — nobody is asked to dismantle the organisation, because the organisation dismantles itself as the book ends. It is fundable where the legacy book covers its own declining servicing cost and the parent's incentive to stop originating is written into the purchase right — stopping is the one change an incumbent's structure permits, and it is still a change that strands fixed cost and forgoes revenue, which is why the book has to carry it. And it is honest about what the parent's purchase right buys. The parent that exercises acquires the clone as a going concern on the new cost base, and it holds the clone alongside its run-off book rather than merging the two. What it has bought is its own future, as a separate company, on the day its past has mostly finished.

The asset

Engineering for Liquidity argued that autonomous businesses are the ultimate acquisition targets because they lack the three things that make most acquisitions fail: Key-Man Risk, black-box uncertainty, and cultural integration cost. The corpus named the condition a Liquidity Lock — the convergence of operational excellence and governance transparency that makes a business fully acquirable — and the shape of the thing acquired a Turnkey Margin: predictable cash flow, a known low-overhead profile, no integration requirement at close.

The clone is a Liquidity Lock with a property the corpus's earlier specification of the asset did not have. Its buyer is named in advance. The parent's purchase right, written the way The Parent Contributes Knowledge, Not Capital requires, is an exit path that exists from the first round. The investors who funded the clone did so knowing who is likely to buy it and at what floor; the parent that contributed the inheritance holds a route back into a business it could not have built. And because the clone has no organisation to integrate, the acquisition is the one Engineering for Liquidity described: a transfer of ownership, not a merger of firms. The parent takes the keys. Nothing has to be reconciled, because there is nothing on the clone's side to reconcile it with.

The purchase right can lapse, and the structure is built so that the clone is just as acquirable when it does. A clone the parent declines is a Turnkey Margin for a competitor, a financial buyer, or a second incumbent that wants the architecture without having contributed the inheritance. The drag-along The Parent Contributes Knowledge, Not Capital specified is what makes that real.

The Flywheel with a corporate counterparty

The corpus's studio thesis rests on the Arco Flywheel: each business built generates operational proof, resolved failure patterns and reusable infrastructure that reduce the cost and time of the next. The Case for the Studio set a target of a 60 per cent reduction in time-to-market for each successive build. The clone changes two inputs to that cycle and leaves the mechanism alone.

The first input is the content. A studio building from a clean sheet in a new market has the architecture and has to discover the states by operating. A studio building a clone has the architecture and the inheritance, and the discovery is largely done before the first transaction. That is Infrastructure Drag reduced on the one axis the Agentic Core could not reduce on its own. The second input is the asset at the end. A clone has a buyer with a right and a floor from day one, so the Flywheel's output is not only infrastructure for the next build but a liquidity event with a known counterparty for this one.

And the platform carries. The builder retains the architecture and licenses it; the clone runs on the licence indefinitely; the parent that exercises acquires the company and the licence, not the platform. The second parent in the same industry starts from the first clone's architecture with its own inheritance to encode, and the builder's position improves with each: more resolved failure patterns, more encoded exception classes, a formation path for Stewards that has been run before. The Flywheel was described for ventures Arco originates. It turns the same way for ventures an incumbent's knowledge seeds, and it turns faster, because the hardest part of any build — knowing what the business has to decide — arrives as an input.

What Is Claimed, and What Is Not

One claim is made. Where an incumbent will contribute and the five conditions hold, the design gives a shorter build, because the content arrives as an input; a more liquid asset, because the buyer is named; and a defence the incumbent's own barrier supplies. These are expected consequences of the design, not results. Each has a measurement that would test it: extraction and replay time against a clean-sheet discovery, the financing terms a round in fact accepted, customer migration out of the parent's lines, and realised exit value against the floor.

The claim stops at four boundaries. Some markets have no incumbent willing to contribute. Some inheritances, once replayed, resolve to a handful of opinions and no mandate. Some legacy books cannot fund their own run-off. And the Flywheel's first turn in any industry is still built from the architecture alone.

Run-off is borrowed from insurance and lending, where it is established practice. Applying it to a parent and a clone is a projection from that practice and from the derivations in the earlier memos. There is no external exhibit here. Arco has built no clone, extracted no inheritance and run off no book, and no parent has exercised a right.

The Operator's Verdict

An operator inside an incumbent, or a studio approached by one, should take one test from this memo. Ask what the parent wants to do with the clone at the end. A parent that intends to merge the clone back into itself intends to transform, and the corpus's answer to transformation has not changed. A parent that intends to hold the clone beside a book it is running off has understood the argument: the clone is not the parent's next version. It is the business the parent could not become, held by the parent, as the parent's old book finishes on its own terms.

The structure, the Stewards and the end state are derived; the checks run before any of them exist:

1. Name what crosses from the incumbent and what stays. If the stack, the contracts or the organisation are on the crossing list, it is a transformation, and the corpus has said what happens to those.
2. Name every channel by which the parent's accounts, budget, brand or board reach the clone's decision to continue. Each one is a Finn.
3. Confirm the parent is a minority holder with a floored purchase right on a revenue trigger, with an expiry and a drag-along, and no client fence. Then map every dependency — data licence, distribution, a single expert — through which the parent can stop the clone without a vote, and close each one.
4. Confirm the founding Steward is from outside the parent by default — an insider only by passing the same test on evidence — and that every domain Steward from the parent is added on Escalation Rate, not on convention.
5. Confirm every rule in the mandate carries the exception history that produced it, so that a Steward who did not write it can be shown why.
6. Ask the parent what happens at the end. If the answer is "merge," the parent has described absorption, and the clone will inherit the Legacy Liability it was built to escape.

Technology changes what is possible. Attrition determines how the old book ends.