classifyThreshold_raw
plain-language theorem explainer
Packages the classification cutoff 0.9 (±0.05) as a provenance-tracked real measurement for the audited classification interface. It is an explicit calibration constant from manual tuning on canonical bases, not a derived RS quantity. Downstream code cites it in the calibration inventory and in the optimality hypothesis wrapper. The body is a structure literal with fixed fields.
Claim. The classification threshold is the real measurement result with value $0.9$, uncertainty $0.05$, and calibrated provenance (source description: classification threshold for exact token match; date $2026$-$01$-$06$; method: manual tuning on canonical bases). It is a calibration constant, not a derived value.
background
The ambient module is the quarantine layer for empirical inputs. Certified theorem surfaces must not import raw measurements; every number that comes from data or tuning is wrapped so source, method, and optional error bound travel with the value.
A measurement result is a triple: a value of some type, an optional real uncertainty, and a provenance record. Provenance marks whether the entry is calibrated, mined, or otherwise generated, and carries human-readable source text. Unwrapping discards provenance and returns only the bare value.
This particular constant is the cutoff used by the audited classification interface (exact token match). The doc-comment states flatly that it is calibration, not derivation: the number is fixed by manual tuning on canonical bases rather than forced by the Recognition chain (T0–T8, RCL, or the phi ladder).
proof idea
Definitional structure construction, not a proof. The body fills MeasurementResult ℝ with value 0.9, uncertainty some 0.05, and a calibrated provenance triple (description string, date string, method string). No lemmas are applied; there is nothing to reduce or discharge.
why it matters
Keeps the classification cutoff out of the certified surface while still making it machine-readable inside the quarantined measurement layer. Downstream, calibrationSummary lists it among the other calibration constants (stability threshold, network/projection/energy coefficients) with provenance source strings. The sibling hypothesis wrapper turns the same record into a DataHypothesis: claim that 0.9 is optimal, with stated falsifier path (a different threshold yields better accuracy) and a suggested derivation route (stability analysis plus information theory).
In framework terms this is bookkeeping, not a forcing step. It does not touch J-uniqueness, phi, the eight-tick octave, or D = 3. Its job is to enforce the quarantine rule so empirical cutoffs cannot silently enter theorem-level claims.
Switch to Lean above to see the machine-checked source, dependencies, and usage graph.