Claim–Evidence
Ledger
Every central statement is connected to assumptions, formulas, source hashes, code, tests, results, limitations and a falsification condition. A software PASS is never promoted into empirical confirmation.
Search the public scientific record
Deep links use the Claim-ID itself. Hashes identify the reviewed source bytes; commit SHAs identify code state. An empty or open field is evidence of incompleteness, not permission to infer the missing result.
Loading evidence ledger…
Which statements depend on which locks?
An arrow points from a dependent claim toward a prerequisite. Select a node to open its ledger record. The graph represents declared logical/provenance dependencies, not causal physics.
Evidence classes are not interchangeable
Audit a claim without trusting the portal
Resolve the Claim-ID
Read the precise statement, domain, observer scope and explicitly excluded implications.
Verify source bytes and commit
Compare the SHA-256 hash, repository path and commit with the cited public source.
Reproduce the code path
Run the recorded command with the declared inputs and inspect numerical tolerances.
Challenge the evidence class
Decide whether the result is an identity, numerical check, compatibility result, fit or genuinely discriminating observation.
Apply the falsification condition
Record negative results and dependencies instead of averaging them into a green aggregate.
How to read Evidence
This page separates the declared definition, its computation or visualisation, the evidence supporting it, and the conclusions that remain outside its scope.
Start with the formula or control, identify its domain and inputs, then follow the linked implementation and evidence record.
A passing identity, numerical limit, plot or comparison supports only the stated relation under its recorded assumptions and provenance.
It is not automatically an independent experiment, a complete physical theory, or a proof beyond the explicit claim boundary.
Definitions, evidence and limitations stay linked
The portal keeps current canonical locks above historical descriptions and keeps software verification separate from empirical confirmation.
Canonical source
Use the current P0, JIF or mathematical lock for the formula and domain shown on this page.
Evidence class
Read tests, convergence, reference compatibility, dataset-conditioned comparisons and independent replication as different evidence classes.
Boundary
Every result retains its assumptions, limits and explicit non-claim so a visual or passing assertion is not over-promoted.