A sixth class of machine: the testing machine
Input: frozen measured data and a candidate structure. Output: a verdict. Zero adjusted parameters, discriminating lever, SHA-256 replicable invariance, published failures. This atlas converts all JSON results of the Python simulations into tables, linked to figures, reports and explanations (context and formalism).
Four operational invariants
Zero adjusted parameters
Every number outside the measured data D is derived, never fitted.
Discriminating lever
Remove the candidate mechanism: the result must collapse.
Replicable invariance
Every artefact (script, JSON, figure, note) is frozen by SHA-256.
B3-FAIL
Failure is an output published with the same care as success.
Everything discrete, kinematic or one-body is derivable: spectra, statistics, quantum numbers, topology, signs, sequences.
The continuous correlated two-body response resists without freedom of form — every failure is instrumented and published.
The P31 → P32 → P33 trilogy makes the r₁₂ frontier a measured object: constitutive (P31), Z-dependent law (P32), strengthened by contradiction of exact constraints (P33).
Key results
The tested postulate
Physics is "cheap" in information in its discrete regime (statistics, parities, quantum numbers, topology, selections, signs, directions, sequences) — derivable from little data. It becomes costly in its continuous two-body regime (detailed correlations, magnitudes). The machine is calibrated on this frontier: it wins where nature is discrete, stops where it becomes continuous. Failure is not hidden: it becomes a measured bound (P19: a ≈ 0.28 fm) or a declared frontier (P31).