Expert onchain judgment for frontier agents

Agents read the chain correctly. They still get the meaning wrong.

Frontier models can trace every transfer and still misjudge what the money means: new demand or internal treasury, organic flow or recycled inventory, safe to sign or escalate to a human. As agents get more autonomy and move more money, that judgment has to be tested independently.

Assay is an applied research lab that builds and evaluates AI agents for onchain investigation. Our experts work real onchain cases, document the evidence and the economic ground truth, and turn each case into training data, private holdout evals and executable environments. An agent does not pass because the answer looks fluent. It passes a named expert rubric.