The Institute

Standards, credentials, and measurements for professional judgment.

The Open Institute for
Professional Judgment

A repo ships features. An institute holds standards. Every instrument below is open, free, and verifiable — credentials you can check, rankings you can reproduce, a census you can rerun.

📐 SkillSpec

The conformance standard for agent skills — three levels, a standalone validator, and a live badge any repo can earn.

Versioned · RFC process

🎓 The Operator's Exam

The open credential for professional judgment — performance-based, cryptographically verifiable, impossible to talk your way into.

Always open

🏆 SkillBench Rankings

The professional-work benchmark for models — frozen tasks, blind judging, full transcripts published.

Versioned task sets

🌍 The Census

The measurement of the public skill ecosystem — counts, conformance pyramid, security findings, methodology in the open.

Weekly · flagship annually

🏟 The Season

The global competition: one scenario a month, identical hidden state for everyone, attested scores, permanent Hall of Seasons.

Monthly

📕 The Annual Report

The State of Professional AI — census, rankings, Season results, and the vitals ledger, synthesized into the citable document.

Yearly
The principles that make the marks worth something:
1. Verifiability over authority — methods before results, hashes over trust.
2. Money never touches judgment — sponsors buy naming, never placement.
3. Open corpus, open instruments — MIT, forkable, auditable.
4. Honest limits, stated in the artifact itself.

Charter and governance: docs/INSTITUTE.md · the standard's independence trajectory: FOUNDATION.md