E6 nonlinear-union audit and branch closure
The small e6 attenuation can support the bit18-to-R6 line only if the complete nonlinear monomial union, not a selected subset, replicates across original and fresh cohorts.
Diagnostic work, exact algebra and local capabilities are not treated as end-to-end mining advantage.
What was tested?
The small e6 attenuation can support the bit18-to-R6 line only if the complete nonlinear monomial union, not a selected subset, replicates across original and fresh cohorts.
Why the test is meaningful
A complete-union endpoint prevents cherry-picking: all eleven nonlinear monomials and their lane multiplicities are included. A robust structural reduction should reproduce both aggregate direction and a sufficiently coherent term-level pattern.
U(e6.low4)=⋃_{lane=0}^{3} N(e6[lane])Δ_U=E[|U||survivor]−E[|U||non-survivor]closure rule: close unless complete monomial replication supports the reductionHow it was tested
Enumerate the entire e6.low4 nonlinear union for each exact record, measure presence and lane multiplicity for all eleven monomials, compare original and untouched fresh cohorts, and apply the preregistered closure rule.
What happened
Semantic errors were zero. Mean union difference was −0.116 monomials in the original cohort and −0.145 fresh, but only 6 of 11 monomials agreed in sign across the complete lane-presence audit. The closure rule therefore failed and the bit18-to-R6 line was closed.
Exactness and statistical controls
All monomials were included, all 1,361 original and 300 fresh records were retained, and the closure decision used the complete map rather than the favorable aggregate alone. The negative conclusion is canonical.
What the result means
A small aggregate attenuation replicated, but its internal composition did not replicate strongly enough to justify a mechanism or selector. EXP-040 closes this research line without erasing its exact intermediate capabilities.
Limitations
- Closure is specific to the registered bit18-to-R6 route and endpoint.
- It does not prove that no cheaper mathematical mining method exists.
- No end-to-end ASIC, energy or hashrate advantage was demonstrated.
Evidence trail
Rebuild all e6 low-four-bit ANFs, enumerate every nonlinear term and lane multiplicity, preserve the original/fresh split and apply the same complete-map closure criterion.
Canonical variants
SUBENGINE-V40ASUBENGINE-V40BSUBENGINE-V40CSource: internally audited canonical reports. Local filesystem structure, private headers and operational identifiers are excluded from publication.