HowHow is in private alpha•Request early access
HowHow

Inspect the work behind each conclusion.

Episodes keep sources, evidence, runs, failures, claims and reviews together.

Inspect ClaimLedgerField notes

FEATURED EPISODE

ClaimLedger

01DETERMINISTIC FIXTURE

A CPU-only benchmark that tests whether a recorded claim-to-evidence link fails closed after its underlying source is mutated.

The episode records a bounded question, one short local source fixture, one controlled mutation, one repetition and a deliberately preserved failed run.

READY FOR HUMAN REVIEW

Inspect the complete episode →

02WHAT TO LOOK FOR

Three things the record has to show

  • The stale source invalidates the claim instead of keeping VERIFIED
  • The negative control and the failed run remain in strict verification
  • The source archive is extracted, hash-checked and rebuilt in a clean directory

IN PROGRESS

Other episodes

IN DEVELOPMENT · BOUNDED APPLIED-ML

Harth calibration

A bounded applied-ML protocol under development. It is not a published scientific result, and no end-to-end real research run is claimed here.

NOT YET VERIFIED

NEXT

Bring a bounded question.

A bounded research question is the right starting point for a private-alpha fit discussion.

Discuss a research episodeRequest a walkthrough
Episodes — HowHow