METHODOLOGY · DECISIONS
Every decision, on the record.
We publish every decision we make about the methodology here. Each decision starts from a written account of the facts, is checked against the methodology and our earlier decisions, and is tested against objections before we adopt it. A decision that changes what the index measures takes effect only through a new methodology version, published with its reasons and its effect. Reports published before this page call these decisions rulings: Ruling #19 there is Methodology Decision 19 here.
- Methodology Decision 26Adopted
Readings and recorded concentrations for x402
Sets what x402 readings record: every recipient with at least 1,000 one-time payers, and every payer with at least 1,000 recipients that no one else pays. These recorded concentrations are counts that change no value; the decision also fixes the reading days before version 1.4's switch.
No new version: no value changed
- Methodology Decision 25Adopted
One data view for x402
Confirms that every x402 metric reads x402scan's sender-level 30-day view, after a check showed it computes the same totals over the same window as the other view. On the switch day the count and the volume each have one stored value from that view, and the version note gives both views' totals as facts.
No new version: the view changes with version 1.4
- Methodology Decision 24Adopted
When one token rule makes a stablecoin framework
One binding provision can make a stablecoin framework if it sets real conditions for fiat-referenced tokens, such as who may issue or deal in them, their backing and redemption, or their use in payments. A rule that only caps, delays, monitors, taxes or bans them does not, so Türkiye's withdrawal cap scores 0.
No new version: no value changed
- Methodology Decision 23Adopted
National rules, local admission
A test regime is national when a national law sets its admission rules, test term, legal effect and formal end, even if local authorities admit the participants; if a local rule sets any of these, the regime is excluded. A local list of admitted participants counts when the admitting authority publishes it on its own government site.
No new version: no value changed
- Methodology Decision 22Adopted
A test or a type approval
A permit is a test only if it states a test purpose, a test term and a formal end; otherwise it is a standing authorization, whatever it is called, and scores 0. China's 2023 pilot for connected vehicles is therefore a type approval, not a test.
No new version: no value changed
- Methodology Decision 21Adopted
What a formal end of a test is
A sandbox test has a formal end only if a record is required when it ends, by a set deadline: an end-of-test report, the authority's conclusions or exit decision, or proof of the exit steps. A test term, interim reports or a renewal review are not enough, so such a permit scores 0.
No new version: no value changed
- Methodology Decision 20Adopted
When we say a score rose or fell
We say that the index, a component, a category or the LRRS rose or fell only when that direction passes the direction test; otherwise we show both values and claim no direction. A single metric's value is a measurement and is shown as measured.
No new version: it governs what we say, not the index
- Methodology Decision 19Adopted
What the x402 metrics measure
Agent participation on x402 becomes a count of distinct paying addresses instead of agent roles, and the transaction count gets a cap per payer; value settled stays uncapped. The two changes ship as separate versions, and each new series is joined to the old one on a single day.
Version 1.4, not yet live, and a later version
- Methodology Decision 18Adopted
Every EU member is surveyed
The list of jurisdictions we search for new laws now includes all 27 EU members: 48 in all, covering 89.1% of world GDP. No score changes by design; the most the LRRS could be too low falls from 14.4 to 10.9 points.
- Methodology Decision 17Adopted
Automated vehicles and drones
A permit for a bounded test of an automated vehicle or drone is judged as a sandbox, under the same four conditions; a standing licence to operate one fits no category and scores 0.
- Methodology Decision 16Adopted
Stablecoin rules for issuers or for service providers
A stablecoin framework counts if its binding rules address fiat-referenced tokens as such, whether they govern issuing them or providing services in them; generic virtual-asset or payment rules do not. Brazil's central-bank rules count at 0.5 from 2 February 2026.
- Methodology Decision 15Adopted
A law counts from the day it starts to apply
An enacted law scores 0 until its operative provisions commence, 0.5 from then, and 1.0 on current operating evidence; a power to set up a sandbox is not a regime. The US GENIUS Act therefore scores 0 until it takes effect on 18 January 2027.
- Methodology Decision 14Adopted
One AI participant qualifies a general sandbox
A financial or sectoral sandbox counts once the authority's own record names an admitted participant whose system under test uses AI; no AI-specific track is needed.
- Methodology Decision 13Adopted
Operating evidence must be current
A sandbox scores 1.0 only while a participant's test is live or its newest operating record is less than 12 months old; after that it falls to 0.5 while the regime stands. France's CNIL sandbox is 0.5 on this rule.
- Methodology Decision 12Adopted
Two dates for every finding
Each status carries its legal date, when the fact took effect, and its update date, when our table containing it went live; the live score counts it from the update date, and history is recomputed on legal dates at every version that changes the LRRS. Every status is checked again at least every 35 days, and every jurisdiction and category is searched for new laws at least every 95 days.
- Methodology Decision 11Adopted
ERC-8004: a cap by content source too
Each ERC-8004 registration counts only while both its address and its content source stay under 100 in the 30-day window, because manufactured registrations reuse one or the other. Recomputed over its history, the metric showed that much of the registry's past activity was manufactured.
Later changed by: Methodology Decision 19
- Methodology Decision 10Adopted
ERC-8004 stays in the index, with safeguards
The capped ERC-8004 count stays in the index as a measure of agent identity, admitted below the usual launch range by a documented exception for very young metrics. The exception ends once at least 2,000 distinct addresses register for 90 days, and a registry that can be verified better would bring a new decision.
Later changed by: Methodology Decision 19
- Methodology Decision 9Adopted
ERC-8004: a cap per address
Each address counts at most 100 ERC-8004 registrations in a 30-day window, so one actor can never look like a busy registry. The cap reads only public on-chain structure, never intent.
Later changed by: Methodology Decision 19
- Methodology Decision 8Adopted
ADGM's DLT Foundations do not count
The Abu Dhabi Global Market's DLT Foundations do not count as a machine-economy free zone: the law governs foundations that people must run, not autonomous systems. Laws that fail on their own terms are recorded as exclusions, with the reason.
- Methodology Decision 7Adopted
Free zones and laws below the national level
A US state, province or Land is not a free zone: the category counts only zones that a national government created, such as DIFC and ADGM. Laws below the national level are tracked, not scored.
- Methodology Decision 6Adopted
What counts as a sandbox
A sandbox counts only if all four hold: a supervising authority runs it, taking part has a legal effect, AI is demonstrably in scope, and it has a formal admission and end; it scores 0.5 once in force and 1.0 once a participant is admitted. An EU duty on member states to set one up, such as the AI Act's Article 57, confers no status by itself.
- Methodology Decisions 1–5Adopted June–July 2026
Reference values of five metrics, before launch
Before the first published score, these decisions set the reference values that place five metrics on the 1–100 scale at launch.