An evidence-linked assessment of whether agentic-payment protocols carry rules for dividing the value they move. Maintained by Ray R, Independent Researcher.
| Protocol | Layer | ARL (allocation) | Status |
|---|---|---|---|
| x402 | Payment execution | 0 | No allocation rule Full audit v0.1 → |
| Additional protocols & layers | Authorization / execution | — | Forthcoming in progress |
| Level | Criterion |
|---|---|
| 0 | No allocation rule exists; value is captured by whoever controls the interface |
| 1 | An allocation rule exists but is unverifiable (fails C1: external checkability) |
| 2 | Verifiable but single-party only (fails C2: multi-party compatibility) |
| 3 | Multi-party but capital-fused (fails C3: resistance to capital capture) |
| 4 | Capture-resistant but decoupled from realized value (fails C4: economic value connection) |
| 5 | Verifiable, multi-party, capture-resistant, and economically anchored |
Classification rule: a system is classified by the first question it fails, not forced onto the nearest rank. ARL is an analytical scale, not a certification standard.
| # | Official claim (abridged — verbatim quotes in the full audit) | Evidence check |
|---|---|---|
| X1 | "open, neutral standard for internet-native payments" | Partially verifiable Linux Foundation project; technical-spec authorship concentrated in one originating company; governance internals not independently checkable (gap G4) |
| X2 | "Zero protocol fees" | Verifiable two independently checkable layers: no protocol fee; facilitator pricing public |
| X4 | "Trust minimizing" — facilitator cannot move funds beyond client intent | Verifiable by design payment schemes constrain execution; facilitator is non-custodial. Fund safety ≠ value division |
| X5, X7 | Network-agnostic; facilitator plurality | Verifiable multi-chain SDKs shipped; self-facilitation is a documented production path |
| — | Allocation rule | Absent by inspection no value-split scheme, allocation constraint type, or revenue-division primitive in the reviewed material |
Each row is quoted verbatim with source and snapshot date in the full audit. Where a claim says nothing about value allocation, that silence is itself a finding.
| Gap | To reach | Description |
|---|---|---|
| G1 | ARL 1 | Any normative allocation rule or primitive: a value-split scheme, an allocation constraint type, a revenue-division extension point in the spec. Currently absent. |
| G2 | Facilitator economics | Largely closed for the incumbent facilitator (public pricing); open for the broader ecosystem as it scales. |
| G3 | ARL 2→3 | If an allocation rule is added, it must be independently verifiable and usable by heterogeneous parties — not a single operator's private ledger. |
| G4 | Governance evidence | Governance documents that make "open, neutral standard" independently checkable: member list, admission criteria, governance relationship between the Linux Foundation project and the originating company. Evidence currently insufficient. |
As a first step toward open audit infrastructure, we replicated population-scale measurement from an orthogonal direction on the Base-chain settlement slice of a public CC0 BigQuery listing: 52,652,290 settlements routed through 105 known facilitator addresses (99.86% of 52,723,702 rows in the raw listing). Events span 2025-12-05 to 2026-04-14 (UTC, 131 days) — the period covered by the upstream dataset, which has not been updated since April 2026. Every figure below describes that frozen window, not a live network. Aggregate statistics only; full addresses withheld.
| Finding | Value |
|---|---|
| Gini coefficient, payer-level totals | 0.974 (settled value) · 0.967 (settlement count) |
| Coordinated payer cohorts (behavioral fingerprinting + union-find clustering) | 181 cohorts · 52,008 wallets — roughly two-thirds of all paying addresses |
| Cohort share of activity | at most 84.1% of settlement count, but only 24.9% of settled value (an upper bound — fingerprints capture behavioral similarity, not intent) |
| Diurnal rhythm separation (10³ volume stratum, 24h Fourier amplitude, median) | cohorts 0.070 vs the rest 0.375 — cohort wallets are near-round-the-clock, the rest are not |
Concentration alone proves nothing — the question is whether the concentrated part behaves like an operator or like many independent customers. The rhythm and ticket-size gaps above indicate that genuine, consumption-driven demand is materially lower than settlement counts suggest: settlement count measures manufacturability, not demand (gas is sponsored, and nothing on-chain marks who controls a payment). The complete report with full methodology, sensitivity grids, and replication files ships with the next version of this scorecard.
Corrections welcome. Misquoted claims, wrong versions, or missed evidence — open an issue or PR on the audit repository. Corrections are merged with attribution.
Ray R · Independent Researcher · Scorecard v1.0 · 2026-10-01 · freeze window Dec 2025 – Apr 2026