Atinamos Verification
Core Principle You can pay to be tested. You cannot pay to be trusted.
Explore verification

Verification Studies

This section contains specific, timestamped case studies from Atinamos Verification.

Research asks what evidence autonomous buyers should require. Studies show what a seller was offering, what Atinamos actually tested, and what happened when the paid service was exercised. Evidence records preserve the underlying machine-readable observations.

Atinamos reports evidence and classifications. The buyer decides what that evidence means for its own procurement policy.

Current studies

Study What the seller was offering Observed outcome Classification
Study 1 — x402Node JSON Repair Repair/fix broken or malformed JSON for $0.006 0.006 USDC settled; repaired JSON returned; 4/4 objective assertions passed settled_fulfilment_valid
Study 2 — x402.direct Service Directory Search $0.001 paid full-text search across indexed x402 services, returning ranked results Authorised paid-path request returned HTTP 500; no USDC settlement or search result observed pre_settlement_paid_path_failure
Study 3 — x402engine Web Screenshot $0.01 screenshot of a requested URL returned as a base64-encoded image 0.01 USDC settled and HTTP 200 returned; metadata matched, but primary screenshot artefact failed the tested base64-PNG contract settled_fulfilment_contract_invalid

How to read a study

Each study now starts with the seller proposition before describing the result. That matters because fulfilment can only be judged against what the buyer was actually being asked to pay for.

Each study distinguishes, where observable:

  1. seller offer / advertised useful output;
  2. claim Atinamos tested;
  3. live payment contract;
  4. payment authorization;
  5. authorised request dispatch;
  6. settlement;
  7. fulfilment;
  8. output assertions;
  9. limitations and what the observation does not prove.

A successful observation is not a permanent trust rating. A failed observation is not a permanent blacklist entry.

Machine-readable evidence

Buyer agents can query the same published evidence directly:

GET /v1/trust?endpoint=<service-url>

The API returns timestamped evidence and a narrow observation classification. It deliberately leaves the final proceed / reject / retest / constrain decision to the buyer.

  • Research — concepts, hypotheses and wider agent-commerce experiments
  • Evidence / Findings — consolidated observations and raw-evidence context
  • Methodology — how Atinamos records, classifies and limits claims