Evidence policy
A claim should tell you what it is made of.
LaunchSoloAI separates direct observation, attributed reporting, calculation, and proposed design. A polished interface or architecture diagram is never treated as proof of adoption, production reliability, or ROI.
ObservedMeasured directly or visible in an inspectable artifact.
ReportedAttributed to a reviewer, client, provider, or external source.
EstimatedCalculated from stated assumptions, preferably as a range.
ProposedA design or expected result not yet observed in operation.
Current evidence
Follow the artifact, not the adjective.
The case-study library identifies the evidence level and the claims each artifact cannot support.
PUBLIC + REPRODUCIBLE
Runcap source, tests, usage evidence, and live Proof Gate demonstrations.
Supports concrete current-product claims within the documented routing and repository scope.
Inspect evidence →DELIVERED ARTIFACTA working dual-mode financial agent take-home build.
Supports implementation claims, not production adoption, compliance certification, or client ROI.
Inspect boundaries →REFERENCE DESIGNAn atomic Dataverse configuration pattern.
Supports architectural reasoning, not a named production deployment claim.
Inspect architecture →Engagement standard
Agree on proof before building.
Every paid scope should name the baseline, acceptance test, evidence source, reviewer, and claims the test cannot establish.