Agent Economics Lab

agent-economics-lab

Can a production AI decision survive missing evidence, weak graders, and the full cost of the work it creates?

This open-source, standard-library-only Python project turns versioned traces, outcomes, costs, and counterfactuals into an evidence-bound decision to scale a system, supervise it, or stop it. Its feedback contract spans predictive models, generative components, and tool-using agents without flattening their distinct failure modes. It publishes corrections when a result does not survive its own checks.

10

public agent datasets audited under one protocol

31,389

runs carrying both outcome and proxy signals

588

controlled required-gate perturbations

These are dataset audits, controlled software experiments, and one small paired public case. They are not model rankings, production prevalence estimates, or deployment recommendations.

• Code
• Findings
• Dataset audits
• Limitations

reproducible artifacts

• Decision record template closed evidence schema
• Filled decision record synthetic worked example
• Generated decision review emitted only after verification
• Decision verifier replays the decision
• Feedback contract template predictive, generative, agent
• Filled feedback record synthetic worked example
• Feedback validator closed schema and policy digest
• Instrument scorecard measurements and limits