12. Minimum Viable Experimental Roadmap

Jason St George. "12. Minimum Viable Experimental Roadmap" in AfterFiat: The Load-Bearing Thesis. Version v2.0. /v/2.0/brief/read/minimum-viable-experimental-roadmap/

Minimum Viable Experimental Roadmap

The roadmap is ordered by dependency and designed to produce disconfirming evidence early.

1. Receipts and verification harness

Define a compact, chain-agnostic receipt schema binding claim hash, workload ID, circuit or model hash, proof commitment, service tier, timestamps, resource use, hardware profile, and provider signatures. Ship deterministic test vectors, reference verifiers on ordinary hardware, signed benchmark results, and archived raw data. Publish VerifyPrice distributions, failure rates, and adversarial-input behavior. If independent parties cannot reproduce the dashboard, stop.

2. One privacy corridor

Ship one non-custodial corridor before announcing many. Specify atomic completion, bounded timeouts, unilateral refund, transaction-scoped viewing keys, and clear failure UX. Publish VerifySettle success, time to finality, refund outcomes, anonymity-set health, route concentration, and incident reports. Any repeated refund-safety failure excludes the route.

3. One proof factory and canonical starter set

Stand up an open-admission proof factory for a small workload set: a generic proof, matrix multiplication, model inference with honestly graded assurance, media provenance, and atomic settlement. Support multiple proof backends so the experiment does not encode monoculture at launch. Measure queue depth, entry latency, verifier cost, prover concentration, and service tiers.

4. DVC and physical evidence

Create Facility Energy and Capacity Receipts for participating facilities. Build the energy-to-settlement graph, name dependency groups, measure substitution latency, publish benign and stressed DVC, and identify each active minimum cut. Bind Work Credit issuance to the conservative stressed result with a declared haircut. Test transformer loss, hardware restriction, network partition, curtailment, common firmware failure, and liquidity disruption.

5. Developer surfaces and neutral routing

Only after the primitives work, expose simple verbs to declare a claim, request a proof, await a receipt, verify a receipt, and pay conditionally for proof or compute. Add neutral routing once there is enough activity to measure fairness. Publish time to first work for a new provider, house share, top-provider share, geography, ASN diversity, and rejection reasons.

6. Monetary and collateral experiments

Test the three Triad architectures rather than assuming one token. For any base asset, publish native fee share, burns, issuance, collateral, bypass prices, the capturable wedge, and the Native Monetary Buyer Map. Stress pure-native, mixed, and native-plus-insurance collateral against the same drawdown and DVC scenarios. Keep Work Credits typed and DVC-bounded; keep project finance in explicit credit instruments.

7. Useful-work pilots last

Only after receipts, corridors, proof markets, telemetry, and routing are stable should useful-work consensus be piloted. Candidate laboratories include matrix-multiplication work, verified-inference service, and a proof-receipt ledger. Their purpose is to test verification asymmetry, open entry, decentralization, and cheap checking at scale—not to announce a final monetary base.

Decision rule

Advance by telemetry gates, not calendar promises. Before verification, settlement, reachability, DVC, value capture, and holder quality are public and healthy, the base asset is research exposure. A working service stack is infrastructure. Only repeated performance under real or simulated pressure, with a material native holder anchor and no red-line breach, permits store-of-value language. Even then, the premium remains unsized.

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