BLBlind Challenge Ledger AuditorLedger AuditorLOCAL CSV / SEQUENCE PROOFLOCAL

Timestamped verification, structurally tested

Blind Challenge Ledger Auditor

A verdict is not a chain. Test challenge, solution, independent audit, and outcome rows before calling a public record verified.

Synthetic sample; supplied-row rules only. “Verified” means these local checks passed. No actor identity, blind commitment, evidence reference or outcome is authenticated.
Load the bundled ledger or paste your own rows.
CHALLENGE -> SOLUTION -> AUDIT -> OUTCOME

The result will expose ordering, independence, and missing-stage evidence.

Worked distinction

Independent review is a relationship, not a label.

Verified

All four supplied stages exist in order, the challenge is marked blind, and reviewer/solver names differ. No identity or blinding is authenticated.

Compromised

A declared timestamp inversion, a no blind flag, or matching solver/reviewer names blocks the verified label.

Incomplete

Supplied rows without detected rule faults remain incomplete when a required stage is absent.

Should BC-102 count as verified because its audit says “passed”?

Run the sample, then test the rule.

Structure is necessary. It is not scientific truth.

This tool audits supplied rows. It cannot authenticate actors, timestamps, hashes, source documents, model output, or real-world outcomes.

Read the chain before trusting its label

What does a structurally complete ledger establish, and what evidence still needs independent checking?

This local auditor groups supplied CSV rows by case and checks a four-stage sequence: challenge, solution, audit and outcome. Use it to locate gaps and contradictions before inspecting the actual records behind each reference. A “verified” classification means the supplied rows pass this tool’s rules; it is not authentication of the challenge, people or result.

Actor names, timestamps, blind_commitment values and evidence references are declarations supplied by the CSV author. No evidence URL, commitment hash, identity or scientific claim is checked. Different actor strings do not demonstrate reviewer independence. The engine does not judge whether an outcome is successful.

Try a worked example

Load the synthetic sample and audit it. BC-101 has all four stages in increasing timestamp order and different solver/reviewer names. BC-102 has its solution before the challenge, a no blind flag, and the same solver/reviewer name; it is compromised even though its audit says passed. BC-103 has only challenge and solution, so it is incomplete. The sample totals are 9 rows, 3 cases, 1 verified, 1 compromised and 1 incomplete.

Connect a row to the evidence it names

W3C PROV describes provenance using entities, activities and agents. A useful review asks which record was produced, by what activity, and who was responsible. Here, an evidence_ref is only text: a hash: or doi: prefix does not prove a real document exists. Retrieve and inspect supporting records separately before relying on their contents. This CSV format is not a PROV conformance validator. W3C: Provenance overview

Order timestamps without claiming they are trusted

This parser expects second-resolution UTC strings such as 2026-01-10T09:00:00Z. It rejects offsets and fractional seconds even though RFC 3339 permits broader representations. Adjacent existing stages must strictly increase; equal timestamps count as an order fault. The tool sorts rows for display and export, but sorting does not cure an inverted declared sequence. RFC 3339: Internet timestamps

The engine uses browser Date.parse after a shape check; it is not a complete calendar-validity validator. When an intermediate stage is missing, its nonadjacent neighbors are not compared. Confirm actual dates and chronology in the original records, rather than treating an accepted timestamp as an attestation.

Distinguish a blind flag from a commitment

Only the challenge row’s yes/no blind_commitment field affects this rule. A yes flag does not demonstrate that a solver was unaware of the answer, that a commitment preceded disclosure, or that a hash was later opened correctly. The reviewer comparison trims actor strings and ignores case; affiliations, shared access and aliases remain outside the check.

Interpret gaps and downloads carefully

Each case permits at most one row for each stage. A detected fault takes precedence over missing stages, and the missing-stage total counts only cases classified incomplete. In the sample, BC-102 lacks an outcome but contributes no missing-stage count; BC-103 contributes two. Read its event rail as well as the totals.

The exported JSON includes the supplied canonical rows, classifications, findings, method and limitations. It is a reproducible record of these local rules, not a signed certificate or verification of referenced evidence. Outcome text must be present but any nonempty value is accepted, including failed. After an invalid audit, the last valid result remains visible; its JSON describes that last valid input. The successful input is saved in browser local storage; Reset clears that saved value.

Sources and further reading
Super generates helpful tools and automates fact-checking across the internet proactively. If you enjoyed this tool, build your own with Super and share it with a friend.