The OpenAuditor blog

What changes the verdict.

Experiments in audit evidence, decision risk and the assumptions hidden inside a passing check.

Field notes

Synthetic inputs, recorded engine outputs, downloadable evidence.

The batch boundary changed the audit verdict

The same payments expose a $100 excess together and pass separately when comparison history is omitted. A controlled experiment in audit scope.

One $1,000 invoice. Payments of $600 + $500.

Reviewed together$100 excess found
Reviewed separatelyBoth pass

Same records; payment history omitted.

Read the investigation

99.5% of which decisions?

Worked counterexamples in verification error budgets, abstention and selection: why a configured certainty target needs a denominator.

Configured error budget: 0.5%.

Three holds at 0.2% eachUncertain
One qualified fail at 0.2%False

Gate behavior with synthetic bounds.

Read the investigation

Practical guides

Procedures for the work around the finding.

Follow new articles in your feed reader.

Subscribe via RSS ↗