Use four labels consistently
Identify hardware, software, method, environment, sample size, units, date, uncertainty and retained evidence.
Publish equations, inputs, sources, sensitivity and the conditions under which the result changes.
Name the exact part and operating conditions internally; do not present a datasheet value as integrated-system performance.
State the acceptance test and threshold that would convert the target into measured evidence.
Do not jump from simulation to operational proof
Each step answers a different question. A simulator can test algorithms and failure logic. Bench hardware can measure power and timing. Environmental testing examines survival under controlled stress. Only flight evidence measures behaviour in the intended orbital context; operational history then addresses reliability over time.
Preserve enough context to rerun the result
- A stable claim identifier and plain-language statement.
- Repository or document revision, configuration and dependency versions.
- Input data, provenance, exclusions and preprocessing.
- Test protocol, equipment, calibration and operator record.
- Raw results, derived calculations and machine-readable outputs.
- Expected result, acceptance threshold and negative cases.
- Known limitations, uncertainty and unresolved anomalies.
- A public summary that removes confidential detail without changing the conclusion.
Different claims require different proof
- Thermal: correlated model plus integrated thermal-vacuum test
- Radiation: mission analysis plus relevant component and board evidence
- Links: calculated budget plus measured service under adverse cases
- Consensus: reproducible runs, negative security cases and independent review
- Economics: common workload, full lifecycle and sensitivity
- Sustainability: declared boundary, resource inputs and lifecycle effects
- Sovereignty: law, keys, authority, audit and remedy
- Readiness: contracts, approvals and operational evidence—not intent
Make corrections visible and versioned
Every material public research release should show publication date, revision date, authoring organisation, maturity, evidence class and limitations. If inputs or conclusions change, retain a dated correction note instead of silently replacing the earlier claim.
Patent-sensitive, personal, security-critical and commercially controlled material can remain protected. The public summary should still state what was tested, what result was observed, what was not tested and who performed any independent review.
Transparency is not publishing every secret. It is making the public conclusion traceable to an honest method and boundary.