§11. Service-Instrument Design Space

Jason St George. "§11. Service-Instrument Design Space" in Next Generation Stores of Value: Privacy, Proofs, Compute. Version v3.1. /v/3.1/read/part-ii/11-monetary-design-space/

Service-Instrument Design Space

Here is the question this chapter answers: once work can be verified, what should be built on top of it? The triad produces three capacities, and Work Credits turn capacity into claims. But a claim is a design decision, not an inevitability, and the decision has consequences that outlive whoever made it. Pick the wrong instrument for a real buyer, and the buyer quietly routes around you: a treasury that needs a two-year hedge will not buy a perpetual index, and an AI lab hedging inference cost will not buy privacy credits it cannot redeem. The failure mode is not loud. Nobody announces they are leaving; the service just never gets the volume, and the instruments sit on the shelf with impeccable specifications and no counterparty.

That is why this chapter walks the design space rather than picking from it. Monetary candidacy remains confined to the base asset and subject to the Triad Coherence Test (§6: The Triad and the Monetary Candidate); what follows is the inventory of everything else you can build, organized by which triad leg the claim draws on, and by who — concretely — would ever hold it.

The legacy labels are kept because they name real demand clusters, but read them as lenses, not products. “ZK Money” is not a coin; it is the observation that privacy and proofs are usually bought together. “AI Money” is not a token; it is the observation that compute is bought forward, in size, by parties who fear shortage. Each lens describes a buyer, and each buyer implies a different instrument.

The “ZK Money” Analytical Lens

The legacy “ZK Money” label groups service instruments that primarily reference Privacy + Proofs. In practice this means claims on shielded settlement capacity — corridor LP positions, privacy-rail credits — and claims on zero-knowledge proof capacity such as SNARK/STARK service tiers.

What these share is a particular pairing: bearer-like, censorship-resistant settlement rights, bundled with attestations of compliant behavior. Neither half is worth much alone. Settlement rights without attestations are an offshore account with better cryptography; attestations without settlement rights are an audit trail nobody can act on. The value is in the bundle, which is why regulated actors — who need privacy from counterparties and provability toward supervisors simultaneously — are the natural buyer.

One transaction to fix the idea. A corporate treasurer operating under an emergent capital-control regime needs to move payroll to overseas staff without exposing vendor rates to competitors or the transaction graph to the local banking authority, and must still satisfy the parent company’s auditors at quarter-end. She buys corridor credits that settle shielded, and each settlement emits a PIDL receipt her auditors can verify without a viewing key. That single purchase — privacy toward the state, disclosure toward the auditor, both on purpose — is what “ZK Money” was always pointing at. Treasuries under repression, regulated actors needing simultaneous privacy and compliance, and individuals wanting cash-like digital assets are variations on the same shape.

The “AI Money” Analytical Lens

The legacy “AI Money” label groups service instruments that primarily reference Compute + Proofs: claims on verified FLOPs, futures on model-specific inference capacity with proof guarantees, and equity-like positions in proof factories whose output is standardized, verified compute.

The common thread is that compute is the one triad leg bought forward. Nobody pre-pays for privacy next year; plenty of parties pre-pay for inference next year. That asymmetry is the entire commercial logic of the lens: it exists because shortage risk, not usage, is what generates a term structure.

One transaction. An AI lab training a frontier model has a contractual delivery window for a customer deployment and has watched spot GPU prices move 40% in a quarter. It buys futures on verified inference for the specific model hash it will serve, with proofs of correct execution attached, locking both price and correctness for the delivery window. The hedge is only as good as the verification — an unverified forward on compute is just a promise from whoever sold it — which is why the claims are typed to workload and proof tier rather than to “compute” in the abstract. AI labs and application builders hedging compute shortage, and investors who believe AI demand outlives any particular model cycle, are the same buyer at different horizons.

Hybrid Instruments

Hybrids reference all three legs at once, and they exist because real obligations rarely arrive one leg at a time.

A triad basket is an index-like instrument backed by diversified Work Credits across privacy, proof, and compute workloads — for an allocator who wants exposure to the stack’s aggregate throughput without forming a view on which leg tightens first. A corridor-linked compute claim couples private settlement rights with AI inference capacity: the holder can pay for and receive verified inference without revealing who is asking, which is the shape an agent acting for a pseudonymous principal actually needs. An agent service treasury takes the same idea and makes the holder the agent — an on-chain system holding combinations of typed proof, settlement, and compute claims to fund its own operation, buying its own compute the way a factory buys its own electricity.

One transaction, and the one worth watching. A software agent under contract to serve a customer’s inference workload for a quarter holds its own treasury of corridor-linked compute claims, draws them down as it serves, and settles with its customer over privacy rails — no human in the loop, and every drawdown receipted. If that transaction ever runs at scale, the instruments holding it up are hybrids by construction.

The key design constraint is always the same, and it is the chapter’s answer to the question it opened with: claims must stay tightly coupled to verifiable work and capacity, not to governance mood or marketing narrative. The moment a claim’s value can be moved by a vote or a slogan, it has stopped being a service instrument and started being a governance lottery — and the buyer who needed the hedge is already gone.

New here? Start with the one-minute version.

Tip: hover a heading to reveal its permalink symbol for copying.