# 0989 — GRADE = MEASURED

- Snapshot date: 2026-10-13 (UTC)
- Chains: base (chain_id 8453), ethereum (chain_id 1)
- Population: all ERC-8004 agents in the live profile store on each chain
  (Base N=23,403, Ethereum N=2,946).
- Pilot cohort: 36 test-purchase attempts over three dated pilots (2026-10-12
  + 2026-10-13), covering 35 distinct agents on Base. Ethereum: zero
  attempts (no Ethereum agent cleared the deterministic cheapest-price gate
  before the pilot windows closed).

## Pre-pinned claims

- **C1 (categorical)**: Of the Base ERC-8004 agents the discovery pipeline
  has seen, fewer than 1 in 100 publish a payTo + parseable price inside
  the x402 normal-range cap of ≤ $1.00/call.
  Held: **True** — 25 of 23,403 = 0.11 %. Bar: < 1.00 %.
- **C2 (categorical)**: On Ethereum mainnet, fewer than 10 ERC-8004 agents
  publish a parseable x402 price today.
  Held: **True** — 2 of 2,946 publish a parseable price (both in the
  normal range). Bar: < 10 agents.
- **C3 (ordering)**: `s10_delivered.agents_in_pool` ≤ `s7_price_normal_range.agents`
  on every chain.
  Held: **True** — Base 4 ≤ 25; Ethereum 0 ≤ 2. The monotonic-funnel
  property is a correctness check, not a finding.
- **C4 (sanity)**: The 10 agents that appear in `snapshot.json
  .per_chain.base.stages[9].delivered_agents[]` all have an on-chain
  USDC transfer tx hash OR are explicitly tagged with an on-chain
  settlement log that reaches the advertised `payTo`.
  Held: **True** — 9 of 10 have a direct `pay.tx_hash`; 98433's tx hash
  was recovered via USDC-Transfer log scan from the payer wallet in the
  settlement block window and is tagged `onchain_settlement`. Bar: all
  10.

## Non-pre-registered observations

Reported but NOT pre-pinned (we didn't know what the numbers would be):

- The 10 Proven agents on Base are 10 distinct `payTo` addresses — one
  operator per delivery, no cross-agent sharing among the delivered set.
- Six of the ten delivered agents on Base fall *outside* today's
  deterministic stage-7 pool (`observed_outside_pool = 6`). That gap
  describes a persistence quirk: the Convert Guru + five 0988-era agents
  publish a price only on an endpoint our profile store does not yet
  write back into `live_endpoints.x402_challenge.accepts[]`. These
  deliveries are real and dated — they're just not visible from the
  deterministic view.
- Deterministic drop-off reasons on Base, ordered by volume: `not_advertised`
  (23,269 — i.e. 99.4 % of all registered Base agents never emit a
  payment field in their on-chain metadata, well-known, or metadata.payment_options),
  `no_402_challenge` (93 — advertises but never returns a parseable 402),
  `unreachable` (15), `price_above_1usd` (0, with one `price_missing` for
  the Forge Knowledge Engine agent, which prices in ETH-wei and so is
  excluded from the USD cap).
- Pilot drop-off reasons: `pay_failed` (17 — facilitator verify-only
  valid but the agent's own server rail returns 400/402/422/405),
  `missing_accepts` (5), `method_not_allowed` (2), `settled_no_delivery`
  (1 — YieldAgentX402 took the USDC and still returned a 402),
  `auth_walled` (1 — three.ws returned 401 at the pay step).

## Known limits / what this does NOT measure

- **Not a scoring change.** The Readiness/Trust composite, bucket
  thresholds, and published scoring rules are unchanged by this
  snapshot. See `/data/payability-funnel-2026-10-13/METHOD.md` for the
  explicit "this is a tracked measurement, not a methodology change"
  rule. (STANDING POLICY: scoring/methodology edits are propose-first.)
- Non-USDC-denominated prices (one ETH-wei example) are excluded from
  the USD distribution and the price-cap gate.
- Owner-cluster counts under-count operators who register agents across
  chains (we do not deduplicate cross-chain in this snapshot).
- The pilots cover only Base. Ethereum stages 8-10 are 0/0 by sample,
  not by proof — we did not attempt an Ethereum purchase because no
  Ethereum ERC-8004 agent cleared the cheapest-cost gate in time.
- A "payable" agent may still be unpayable to a given buyer because
  their facilitator requires a specific `scheme`/`extra` that another
  client cannot generate. We measured one client (the Agent Zero
  signer) with the payai public verify fallback; coverage will be
  broader when 0942+0960 field-test interop lands in a buyer's library.

Full audit trail at `docs/0988-buy-round/RESULT.md`,
`docs/0983-product-readiness/RESULT.md`,
`docs/0982b-test-purchase-pilot/RESULT.md`.