Exact or formal
Definitions, symbolic derivations, logical arguments, proof assumptions, and counterexamples.
Trust and Rigor
Polish does not establish correctness.
Every public object should make it possible to distinguish exact mathematics, numerical approximation, educational simplification, simulated output, physical observation, hypothesis, and reviewed knowledge.
Evidence Classes
Different claims require different proof.
The platform will not collapse computation, peer review, reproduction, and physical validation into one generic verified badge.
Computational
Algorithms, versions, tolerances, precision, tests, convergence evidence, and reproducible execution state.
Empirical
Measurement methods, instruments, calibration, uncertainty, datasets, provenance, and independent replication.
Publication requirements
- Define symbols, domains, units, coordinate frames, conventions, and assumptions.
- Separate raw input and output from explanation and interpretation.
- Identify exact results, approximations, tolerances, and known failure modes.
- Cite definitions, constants, algorithms, datasets, and scientific claims.
- Record object, engine, method, and dependency versions needed for reproduction.
- Preserve corrections, superseded revisions, negative results, and failed runs.
- State what was reviewed, by whom or what role, under which criteria, and when.
Three public integrity zones
| Zone | Meaning |
|---|---|
| Verified knowledge | Established material reviewed against stated sources, scope, and version. |
| Experimental space | Hypotheses, simulations, provisional interpretations, and reproducibility attempts. |
| User-created future | Community builds and extensions with attribution and lineage, without inherited verification. |
AI boundaryAI may explain, organize, or propose. It must not silently modify canonical state, fabricate a proof or source, convert simulation into observation, or mark its own output as verified knowledge.