From "the AI says 100%" to "proven"

Three audit rounds of the same code base against the same spec. Green tests are not enough: SoloTrace only counts a requirement once its tests catch deliberate sabotage.

How an audit runs

IBM Bob does the reading, auditing and fixing. SoloTrace's code checks every claim. A human signs off before any code changes.

Traceability matrix

Select a requirement to see its text, acceptance criteria, verified code evidence, tests and the sabotage attempts its tests caught.

ID	Requirement	Baseline	Round 1	Final

The evidence verifier

AI auditors cite code. SoloTrace re-finds every citation in the audited commit and rejects anything that is not really there.

How IBM Bob runs SoloTrace

Every stage below ran inside IBM Bob IDE. Times, Bobcoins and subagent counts come from Bob's own task history.

Task session summaries