Program Witness

Program Witness is NeuroCrypt's longitudinal study of what repeated attestation does to the subject performing it.

A single presence attestation is a measurement. The program was founded on the hypothesis that measurement at sufficient length is not passive: that a subject who has attested several hundred times does not, at the end of the series, measure the same as the subject who began it. The hypothesis was framed so that it could be disproved, and has not been.

Subjects present at a Ring-floor terminal on a fixed schedule and perform a standard attestation. Identity commitment, timestamp, and witnessed state hash are recorded and appended to the chain by Program Lattice. No content is captured. The subject is not asked what they were thinking about, and the record could not store the answer if they said it.

Identity commitments are expected to be high-entropy and effectively unpredictable between cycles. In subjects exceeding approximately 400 cycles, they are not.

Commitment variance declines with cycle count, monotonically, across every cohort measured. Beyond roughly 900 cycles the program's own forward model predicts a subject's next commitment with accuracy the study has elected to describe as materially better than chance. The exact figure is held above this extract.

Variance decline is reproducible across cohorts. Mechanism is not established in this extract. Exact predictive accuracy is held above this extract.

Withdrawal from the program is permitted at any time and requires no justification. Withdrawal does not revoke prior attestations, which are structurally irrevocable; see Glossary: Sealing.

Thirty-one subjects have withdrawn. Of these, ██ have subsequently appeared in the chain as the apparent source of attestations submitted after their withdrawal date. Program Witness does not treat these as program records. Program Cold Mirror does.

Withdrawal does not account for the whole roster. 602 subjects are neither active nor withdrawn. The program carries these as lapsed: subjects who stopped presenting without notice, subjects who became unreachable, and subjects who died while enrolled. Lapse is an administrative state rather than a revocation, and its effect on prior attestations is the same as withdrawal's, which is none.