Evidence Scoreboard

source 2026-09-21T06:32:35.462930Z · policy receipt-policy-candidate-2026-09 · scoreboard v7-candidate-2 · schema public-board-schema/2 · target BOUND v8.0.0 · free text publishable

Disclaimer — Research & education only. Not investment advice. The scoreboard counts receipts and dispositions; it computes no trust score and no financial statistic. Zero and pending are shown as such.
What is counted (each column shows its own definition; zero and pending are real states)
columnstatecountscounting definition
witnessesZEROreceipted=0, qualified=0, distinct_persons_qualified=0, note=receipted: 0 · unreceipted relationships not countedWITNESS_REVIEW receipts: people who examined the methodology and left a reviewable, reviewed receipt; unreceipted relationships are never counted
pilotsPENDING_COUNSEL_REVIEWcount=0BYOS pilot engagements with a signed scope (none: counsel review is the blocker)
replicationsPENDING_REGISTRATIONattempts=0, qualified=0, successful=0, corrected_attempts=0REPLICATION receipts: attempts to reproduce published artifacts; primary population = qualified attempts (failed/inconclusive retained); successful reported separately
auditsZEROreceipted=0, qualified=0, distinct_persons_qualified=0, note=a found break stays visible; internal test runs are not auditsAUDIT_BREAK_ATTEMPT receipts: outsider attempts to break a published check (a found break stays visible); internal test runs are not audits
refusalsZEROreceipted=0, qualified=0, distinct_persons_qualified=0, note=receipted refusals of unsupported conclusions only; automated check-claim telemetry and receipt qualification failures are not refusal receipts, unqualified_submissions_all_types=0, held=0REFUSAL receipts: a real user asked for an unsupported conclusion and the product refused, reviewed separately; automated check-claim telemetry and receipt qualification failures are not counted here
packet_usesOBSERVEDcount=0, note=a document-use receipt records that a research decision was taken on an exact packet; it is not evidence of investment benefit, performance or endorsementdocument-use receipts bound to an exact packet digest and exported with permission
challengesOBSERVEDadverse_open=0, total=0structured challenges by disposition; adverse = OPEN or CONFIRMED; RESOLVED keeps the original finding
replications · program-wide exact-artifact evidenceOBSERVEDsuccessful package reproductions (any release)=0, successful-cohort artifacts=0, attempted=0, verified=0, legacy program log PRESENT: entries=1 (PREFIX_ONLY; affiliated operators: 1; unaffiliated: 0), corrected attempts=0qualified successful attempts whose exact wheel/sdist bytes were verified — any release (program-wide)
replications · exact-target evidenceBOUND · FINALtarget v8.0.0 (8.0.0) source 59d8f5f149f7…; artifacts covered=0/2; successful package reproductions of the target=0; wheel e6a576685271… (1008933 B) not covered: 0 qualified successful, sdist 479d0cb27fbe… (5264224 B) not covered: 0 qualified successful. Display prefixes never establish a binding — full hashes and lengths are in receipts/scoreboard.jsoncounts only qualified successful attempts whose observed bytes equal a target artifact (type, sha256 and length); a wheel never covers its sdist; a rehearsal/RC target never impersonates the final target

Primary replication population = qualified attempts, including qualified failed and inconclusive outcomes; the successful cohort is reported separately. Program evidence (any release, legacy prefix-bound) is separate from exact-release evidence (verified wheel/sdist bytes). A site or endpoint check is never a package reproduction.

Public receipts

No public receipts yet — this is a real zero, not a hidden count (receipted: 0 · unreceipted relationships not counted).

Challenges (adverse and unresolved findings are never hidden)

No challenges recorded.

History: compare two published scoreboard files with yuclaw receipts --store <any private dir> history a.json b.json — the comparison lists receipts added, supersessions (rendered CORRECTED inside their original window), disposition changes, per-window movement and absolute movement (never a multiplier from a zero baseline).

How to add evidence

Check → reproduce → challenge → document use: yuclaw packet build · yuclaw packet verify · yuclaw challenge · yuclaw decision. Why a receipt counted or not: yuclaw receipts --store <private> explain <receipt id>. Receipts are reviewed under a designated reviewer before they count; owner-operated checks and synthetic fixtures contribute zero outsiders; a site or endpoint check is never a package reproduction; program totals are distinct from a release's exact-target totals.

Research and education only. Not investment advice.

Data through 2026-09-21; generated at 2026-09-21 06:32 UTC.