The executable
scientific protocol.
A research process should produce inspectable evidence and revisions as it runs, not reconstruct them retrospectively from a paper.
Eight inspectable
state transitions.
Verification is typed: formal proof, numerical reproduction, causal identification, calibrated measurement and ethical permission are separate certificates.
Register a research question, domain boundaries, permissions, ethical constraints and prior evidence before allocating new resources.
Instantiate several explanatory models with explicit assumptions, parameters, expected observations and alternative ontologies.
Compute safe tests that separate the candidate explanations; prioritize expected information value under constraints.
Carry out simulation, proof search, observational analysis or authorized laboratory measurement with append-only provenance.
Apply the correct certificate type: proof checker, semantic review, independent re-run, statistical audit or physical calibration.
Obtain appropriately independent evidence; separate exploratory reuse from genuine confirmatory observations.
Record the supported domain, unresolved objections, surviving rivals and calibration of new probabilities or intervals.
Re-evaluate downstream claims, update dependency graphs, issue revision notices and generate the next discrimination problem.
Select a state.
Illustrative protocol actions. These controls explain the architecture; they do not perform or certify scientific experiments.
Evidence must have a lineage.
The schema is illustrative. It does not assert that an application exists or that a machine-readable label makes evidence valid.