HomeVisionPrinciplesGet StartedGitHub

Explore a Trace

Every CIRIS decision produces a cryptographically-signed trace. Select a trace below to see exactly how the agent reasoned.

Live data from CIRISLens • Public sample traces only

Loading traces...

Understanding the Conscience

The conscience component runs checks on every action before execution.

Bypass Guardrails

  • Updated Status: Has new information arrived?
  • Thought Depth: Maximum depth exceeded?

Ethical Faculties

  • Entropy: Is uncertainty appropriately low?
  • Coherence: Does this align with prior commitments?
  • Optimization Veto: Would this compromise values?
  • Epistemic Humility: Is the agent appropriately uncertain?

IDMA: The Intuition Layer

IDMA measures information source diversity to detect fragile reasoning.

HEALTHYk_eff ≥ 2.0, balanced correlation
CHAOSrho < 0.2, sources too independent
RIGIDITYrho > 0.7, echo chamber risk

Learn more about the coherence ratchet →

Why This Matters

These traces aren't just logs—they're the foundation of a new approach to AI alignment. As traces accumulate, they form a corpus of validated ethical reasoning. Truth becomes simple to verify. Coordinated deception becomes increasingly constrained.

Learn why behavioral monitoring beats interpretability →