Accord
Get a verdict that survives disagreement.
Multiple AI passes analyze, challenge, and refine a news claim before returning a result.
Accord runs multiple AI passes over the same claim.
- An initial pass analyzes the claim.
- A critic pass challenges the reasoning and looks for gaps.
- A swapped round tests whether the result stays stable under disagreement.
- A final pass produces the verdict, confidence, and consensus strength.
Confidence reflects the final result after challenge, not just the model’s first impression.
Claim analyzed
Normalized claim
Confidence
Consensus strength
Agreed points
Key disputes
Residual uncertainties
Solver output
Raw model confidence
Critic review
Swapped solver output
Swapped critic review
Swapped main objections
Swapped missing context
Main objections
Missing context
Grounding evidence
Audit trail
Accord is an early verification prototype. Results may be incomplete and should be checked against reliable sources when stakes are high.