DN Research / Reality Index · Inaugural evidence edition · 19 November 2026
What matters
The economic reality of agentic finance is narrower than its promotional transaction counts. A payment-required response is not a settled transfer. A settled transfer is not proof that an AI model decided to pay. An agent payment is not proof of merchant fulfilment. The inaugural DN Reality Index preserves those distinctions and publishes dated evidence rather than an invented headline score.
The November 2026 evidence ledger
| Observation | Reported value | Denominator / period | Attribution | DN interpretation |
|---|---|---|---|---|
| x402 rail settlement | $52.7M | 198.9M transactions | TRM Labs, report 9 Sep 2026 | Observed across known facilitators since May 2025; not necessarily agents |
| Agent-like share of screened commerce | 0.6%–7.5% | Base $25.62M screened | TRM Labs, report 9 Sep 2026 | Model bounds, not verified agent identity |
| Payer concentration | 87.9% | Two payers; 20.4M transactions | MachineEconomy.ai, 30 days to 31 Aug 2026 | A count-concentration statistic, not a market share |
| 30-day settled value | ~$977,000 | Rolling 30 days to 30 Sep 2026 | MachineEconomy.ai, published Nov 2026 | Different dataset/period; not comparable with TRM cumulative total |
| ChatGPT Wallet | Launched Nov 13 | Initial eligible US Pro dots rollout | OpenAI, 13 Nov 2026 | Availability announcement; no confirmed merchant success or settled totals |
Index status: UNSCORED. These are separate research snapshots with different windows and collection methods. Do not add them, plot them as a continuous quarterly series or call them total agent spend.
What the largest numbers actually measure
TRM Labs’ September 9 investigation identified roughly $52.7 million across 198.9 million known-facilitator x402 settlement transactions on Base, Solana and Polygon since May 2025. After screening obvious self-payments, bulk flows and thin-buyer merchants, it analysed $25.62 million of likely commerce. Its strict and permissive classification methods estimated that just 0.6% to 7.5% of that screened amount appeared agentic. Those bounds are modelling assumptions, not independently established agent identity.
Separately, MachineEconomy.ai reconstructed a 30-day August transaction window in which two payer addresses generated 87.9% of 20.4 million x402 transfers. Its report illustrates why counting settlement events without inspecting payer concentration can exaggerate broad adoption. That analysis is not directly interchangeable with TRM’s cumulative multi-chain estimates.
Decision record: five verdicts
- Protocol settled value: observed within a specified analytical perimeter, but not a global agent-commerce market size.
- Agent attribution: plausible-model range; requires direct user, developer or merchant instrumentation to verify.
- Merchant acceptance: announcements demonstrate availability, not order completion or fulfillment ratios.
- Activity concentration: critical to judging whether a network represents distributed demand.
- Quarterly growth: withheld until compatible periods and reproducible measurements are collected.
DN Claim Audit: classify a headline
What does the headline actually prove?
Machine-readable data and licence
The source-attributed observation records are embedded on this page as application/json with the stable element ID dn100-evidence-ledger. A versioned CSV and JSON research export can accompany releases; the data records cite third-party sources and their own restrictions. DN-authored ledger annotations are available under CC BY 4.0, but third-party datasets retain their original rights.
How the quarterly score will be calculated
We will not choose a score denominator after seeing the results. A future scored release requires consistent quarterly windows for settled value, unique economically distinct payers, merchant fulfilment, refunds/chargebacks, permission controls, repeat usage and attributable human-versus-machine initiation. Each observation must have a source trace, currency, definition, exclusions, licensing status and revision record. Metrics without sufficient coverage will remain unscored, rather than becoming zero.
Sources and methodology
- TRM Labs — Who’s Actually Paying? (9 September 2026): methodology and strict/permissive agent attribution ranges.
- MachineEconomy.ai — Third Reading (7 November 2026): payer concentration, September settlement and indexer caveats.
- OpenAI — Introducing ChatGPT Wallet (13 November 2026): availability and limits; not adoption totals.
- DN — Ten indicators and falsifiers for 2027: the monitoring framework used for future updates.
Disclosure and editorial integrity: These are attributed external observations, not original DN on-chain extractions. Their sources may differ in method and may revise their figures. No vendor paid for inclusion, and no inferred market-size total or unsupported agentic share is claimed. Research does not constitute investment advice.
Frequently asked questions
Is every x402 transaction an AI-agent purchase?
No. Protocol-level settlement can also be caused by scripts, repeated payers, tests and non-commerce activity.
What is the DN Reality Index score this quarter?
The inaugural evidence ledger is explicitly unscored because independently comparable quarterly agent-attributed settlement and merchant-outcome series are not yet available.
Are these observations DN measurements?
No. The first edition transparently reproduces and critiques dated external research. Source-specific methods and periods are not pooled into a false single market total.




