Notes on the method

Each note describes one part of this instrument: a term it uses, a state it distinguishes, or a thing the standard underneath it does not settle. They document what is measured and what cannot be, and none of them advises anybody on what to do about it.

There are 3, and at most 7 may stand. It rises to 15 when Bing Webmaster reports a search impression of this site — query_stats — alongside a control property that proves the same call is able to report it at all. Read on 2026-09-05: the call answers for both sites, it returned 225 rows for the control property and 0 for this one. Nothing here is published at a rate: the number is the last reading of whether these pages reached anyone, not a schedule.

• AI Agent Autonomy Levels: What Each Rung Requires as Evidence

  This method note defines ai agent autonomy levels L0 through L5, what evidence each rung requires, and which thresholds are assigned rather than measured.

  2026-08-31

• Four states of absence, and the one that hides as a finding

  Distinguishes nothing_qualified, check_did_not_run and unreadable, plus the state where a blind instrument's empty reading reads as a finding.

  2026-08-24

• What ERC-8004 standardizes, and what it leaves to a validator

  This note describes what ERC-8004 fixes on-chain across its identity, reputation and validation registries, and what it leaves to a validator's method.

  2026-08-24