Identifiability and matched throughput — 4 November 2026
Computational re-analysis of the Incoprea Research archive. NO EXPERIMENT PERFORMED.

Contents
--------
a3_rule.py                    H0199: the D rule (bias = mask-mean / boundary-mean - 1)
a2_h0199_factorial.py         4 layouts x 5 estimators x 7 field families
a_h0199_design.py             BLUP / MSE-optimal witness designs
b_i2_pde.py                   independent finite-volume solve of the I2 heat problem
c_n0751_feasibility.py        N0751: gain-match rule, CRLB, drift amplification
d1b_h034_ident.py             H034: elastic-matching null space + phonon constraint
d2b_sodium.py                 #15: duty-tax theorem and the Sand channel
d3_recovery.py                #15: recovery-rate sweep on the supplied lumped model
out/*.json                    every number quoted on the page
figs/*.png                    the figures

Reproduce
---------
python3 -m venv .venv && .venv/bin/pip install numpy scipy matplotlib
.venv/bin/python a3_rule.py
.venv/bin/python b_i2_pde.py
... etc. Each script writes its JSON to out/ and its figure to figs/.

Reproduction of the supplied dossiers (run first, to confirm the baseline):
  further/02-H034-materials-error/code/lattice_cutoff.py   -> exponents +0.247 / -2.994
  further/07-15-sodium-pulse/code/sodium_pulse.py          -> peak ratios 3.11-3.17 at 10 % duty
  further/04-H0199-reference-placement/code/h0199.py       -> matches saved JSON exactly

Status
------
Independent computational re-analysis by an external AI auditing system.
Not an experiment, not a measurement, not peer review. Kinetics parameters in
the #15 analysis are order-of-magnitude and are not measured for any cell.
