1754 — The First Measurement Rees Grade Restores Degenerate Tomography
Degeneration
Degenerate Entry 1753's fourth labelled effect toward its third:
At , the two effects coincide. The ordinary scalar packet then has rank three and loses the antisymmetric off-diagonal direction.
The exact readout expands as
Therefore
The zeroth grade already supplies
Combining the two grades gives
Thus the first normal/Rees grade restores exactly the one matrix direction lost at ordinary specialization. The second grade is , which is already present in the zeroth packet and introduces no new direction.
Narrow result
The existing flagged-normal calculus is sufficient for a codimension-one loss of tomographic rank:
No post hoc dual-frame splitting is used. The conclusion requires the source-labelled parameter , including its orientation and scale. If only the unparametrized tangent line is retained, the missing direction is known projectively but its scalar calibration remains absent, in agreement with Entries 1735--1736.
This is coefficient/readout Rees data over the existing degeneration locus, not a new carrier stratum.
Durable artifacts
- research/benincasa/checkers/degenerate_tomography_rees.rs
- research/benincasa/results/degenerate-tomography-rees.json
- research/benincasa/degenerate-tomography-rees.md
Next falsifier
Force two independent measurement directions to disappear at the same ordinary specialization. Test whether the labelled two-normal Rees packet recovers both directions and whether the order of specialization has a nonzero commutator.