When the Market Decides What Gets Made

A market has always answered one question: who gets what, at what price. Market Computation is the condition in which market allocation — what gets produced, not merely what gets sold — is performed by continuous computation across agents rather than by periodic, human-mediated price-setting and planning. Continuous here describes the allocation process, not continuous compute: allocation is recomputed as relevant market state changes, not run as a permanently active process. Memo #121: Executable Economy established the loop this requires: demand, price, and allocation feeding production and logistics directly, once a signal clears the Physical Intervention Threshold — the planning cycle removed, the exception gate not. Market Computation is a claim about where that loop leads once it runs at the scale of a market rather than a single business — not a claim about a market that already operates this way.

What market selection already assumed, and what changes

Memo #05: Markets That Work argued Arco selects Breakable Markets: markets with a structurally high Human-to-Logic Ratio, no Systemic Resistance, and fragmented competition. That selection logic assumed the unit being reconstructed was a single business inside an otherwise unchanged market. Market Computation asks what happens to that logic once the market itself, not just the business operating inside it, runs on computed allocation. A market with a high Human-to-Logic Ratio — a large share of gross margin consumed by human labour costs — is attractive today because a business can capture the resulting cost and output delta through Operational Arbitrage. Market Computation asks what happens when the opportunity moves beyond substituting that labour inside one business and into the allocation mechanism connecting many businesses: Agentic Market makes supply-side capacity addressable and responsive; Executable Economy makes the response executable; Market Computation asks what happens when those executable loops collectively become the allocation mechanism of the market itself.

That shift extends a move Arco already makes at the level of one business under Operational Arbitrage: replacing a coordination-heavy team with a small agentic stack captures the value inside a market. Owning the logic an entire market runs its allocation through is a different kind of position — not a business operating inside the market, but the infrastructure the market runs on. Arco has no term for that distinction yet; it is the economic implication Market Computation raises, not a claim this memo is coining. Market Computation creates the possibility of a new economic position: ownership of the allocation logic itself. Whether that position necessarily concentrates power is a separate hypothesis, tested on its own terms below.

The obvious objection is that markets already do this — prices have always signalled scarcity, and production has always responded, eventually. This is not a claim that markets discover a new function. It is a claim about the mediating step between the signal and the response, the same distinction Memo #121 draws for a single loop, applied at market scale. One company routing its own supply chain faster is an Executable Economy. The allocation mechanism of the market itself running as continuous computation, rather than as commerce mediated by periodic human negotiation between many parties, is a different order of claim — and a considerably less proven one.

From who gets what to what gets made

Consider an unnamed market for a perishable industrial input, sold by several producers to many buyers through negotiated contracts renewed on a fixed cycle. Today, a shortage shows up as a price spike a human notices, followed by a renegotiation that takes weeks. Under Market Computation, the same shortage is a signal every connected supply-side agent sees simultaneously, and the response — reprice, reallocate existing stock, adjust the next production run — happens on each agent's own decision cycle rather than the contract's renewal cycle. The market does not stop being a market. What changes is the unit of time in which it clears, and the fact that the clearing price now carries information about production capacity, not just current stock.

The allocation rule becomes the new object of stewardship

None of this collapses the need for judgment; it adds a level above where judgment already sits. In a single Executable Economy loop, the Physical Intervention Threshold still governs the physical commitment: the Execution Layer applies the rule to one action at a time, and an action that does not clear the threshold escalates to the Judgment Layer — the Steward. What changes under Market Computation is not that threshold. It is that the Steward now also has to govern the rule generating the commitments the threshold gates — a higher-order design decision that becomes explicit at market scale.

An individual allocation may be routine — reprice a batch, confirm a delivery window — and cheap to reverse if wrong. The allocation rule that produces many such allocations is a different object: it is not one decision, it is the mechanism generating all of them. A wrong individual allocation carries its own Rollback Cost. A wrong allocation rule can propagate that Rollback Cost across every participant and every claim it has already resolved — a rule governing one warehouse's reprice logic is one problem; the same class of rule governing a regional logistics network is a different order of problem, even if no single allocation it produced was individually irreversible. The Stewardship Model's intervention discipline therefore moves up a level. Market Computation suggests a second stewardship surface alongside the first: the Judgment Layer would not only resolve exceptional executions, but also govern the rules that determine how repeated allocations are made.

And a market whose allocation rule is computed rather than negotiated concentrates a different kind of power: whoever owns the rule holds a position no single participant in a negotiated market ever held, because a negotiated market distributes that judgment across many independent parties by construction. The Rollback Cost logic above sharpens why: the more consequential the rule, the more valuable and the more dangerous it becomes to centralise. Market Computation does not make that question go away. It makes it the central design question, in place of the labour-replacement question that has governed market selection until now.

The Operator's Verdict

A market that computes its own allocation would not be a bigger version of a business that automated its own workflow. It is a different kind of asset: the logic itself, sitting above every participant who still transacts inside the market it governs. And once allocation is encoded, the rule that allocates can become an economic asset in its own right — and, under the Stewardship Model, a point of stewardship, not just a point of computation. That is a larger claim than anything in this body of work so far, and it is proportionally less proven.

This is a hypothesis about how a market could evolve, not a plan Arco or any operator is executing. If it holds, the implication would be to treat Market Computation as the outer edge of the projection, not the starting point. The starting point stays one addressable market, one calibrated Physical Intervention Threshold, and one loop that survives a real forecast being wrong. Owning allocation logic at scale is what that starting point could compound into, if this plays out — not what it begins as.

Technology changes what is possible. Allocation determines who owns the market.