Coh Controlled Access Preview

What Coh does

Coh places a measurement-and-decision layer between a generative model and you. Every response is measured by CMCI — an operational engine that quantifies structural coherence, internal tension and drift — and the Watchdog decides whether to show it, caution it, or hold it for human review.

How to use this preview

1. Ask a clear question in the box above.
2. Read the response and its coherence verdict (the colored panel).
3. Open “What do these mean?” to see the structural reading.
4. Try a sensitive or ambiguous topic and watch the verdict change.

Example prompts

Reading the verdict

• Green — shown. The response holds together structurally.
• Amber — shown, read with caution. It holds together but shows internal tension.
• Red — held for review. It did not meet the coherence threshold and was routed to human review.

What you are evaluating

• Coherence is a structural signal, not a guarantee of factual accuracy — read verdicts as structural assessments.
• Coh complements, and does not replace, your policy guardrails and human judgment.
• The engine is in active, supervised calibration; verdicts reflect the current calibration.
• Held responses are routed to human review; reviewed cases may support a supervised, auditable calibration process, separate from the live user loop.