Reverse engineering
of processes

Reconstruct the whole process Research protocol

Begin with independently measured effects, transitions and constraints. Seek one model that accounts for them together. A familiar outcome is a reconstruction target, not an input disguised as a prediction.

Closure and saturation Conditionally derived

Closure asks which values of the target quantity remain after imposing the conditions and removing physical duplicates. Zero-dimensional does not mean unique: several isolated solutions can remain. Saturation asks whether proven lower bounds can be attained together; choosing a minimum is an additional step.

The working protocol Research protocol

1. Fix effect, preparation, system boundary, duration and uncertainty.
2. Separate measured inputs, established laws and new assumptions.
3. Find all minimally sufficient sets of conditions for the same target quantity; report existence separately from uniqueness.
4. Search for simultaneous saturation and list every equality case. Keep degeneracies and unattained infima visible.
5. Measure transitions and response as well as occupancy. Check whether the proposed process preserves its stated constraints and stationary weights.
6. Declare what remains free. Test a new perturbation without fitting again; record failures and inconclusive outcomes.

A small impossibility test Conditionally derived

For binary A, B and C, demand A=B, B=C and A≠C with certainty. Each pair is separately possible, even with uniform one-variable marginals. Together the demands contradict each other. Adding hidden variables cannot repair the contradiction while retaining all three exact demands.

Negative controls are results too Conditionally derived

The present architecture does not fix Yukawa magnitudes, mixing angles, θ_QCD or the cosmological constant. Consistency and topology alone leave these quantities free within the tested packages. An added hypothesis needs its own registration and possible failure.