Parent: #650. Leftover of #633, not a new production and not a second #622.
Why this is a separate ticket
#633 restored spin-0 N725 Q/Z (PR #655) and retracted 159.6° → 20.4° unoriented on the tiny-torus g vs ΔZ reading. It left N725 g-versus-ΔZ undetermined (no second production size).
#632/#653 repaired the exact laboratory (dual_fail=0 on site; do not copy PR #628's broken bond Z).
The comparison #633 asked for can now be done on existing objects:
DZ_Q[g](u) = (g(u)−g(a))/W − (Q(u)−Q(a))(g(b)−g(a))/W²
W = Q(b)−Q(a)
Default shared anchors 0.2 and 0.8 (both exist on g's 9-decile grid). 1/4 and 3/4 only as a labelled sensitivity if you interpolate g. Compare unoriented lines (|cosine|). Tiny-torus ΔZ cannot prove an asymptotic mismatch — say so if that is still all you have.
Assets (unmerged; fetch branches)
Do not rerun N725. Do not become #622 (no new S theory). One comparison, one notes file.
Deliverable
notes/issue-633-glift-exact-lab-YYYYMMDD.md + JSON of the angles/vectors. Draft PR against main. Equal weighting is sensitivity, not a second vote.
If an anchor or covariance is missing, leave the angle undetermined. Do not invent a conclusion to fill the ticket.
No STATUS, no merge, do not close #633/#622. Wait for ASSIGNED_MACHINE.
Related: #633, #632, #653, #655, #622.
Parent: #650. Leftover of #633, not a new production and not a second #622.
Why this is a separate ticket
#633 restored spin-0 N725 Q/Z (PR #655) and retracted 159.6° → 20.4° unoriented on the tiny-torus g vs ΔZ reading. It left N725 g-versus-ΔZ undetermined (no second production size).
#632/#653 repaired the exact laboratory (
dual_fail=0on site; do not copy PR #628's broken bond Z).The comparison #633 asked for can now be done on existing objects:
Default shared anchors 0.2 and 0.8 (both exist on g's 9-decile grid). 1/4 and 3/4 only as a labelled sensitivity if you interpolate g. Compare unoriented lines (|cosine|). Tiny-torus ΔZ cannot prove an asymptotic mismatch — say so if that is still all you have.
Assets (unmerged; fetch branches)
results/probe-invariant-shape/n725-zflow-corrected.json,scripts/threshold_quantile_lineage.pyresults/probe-invariant-shape/census-exact.jsonresults/type582-residual/latest.json/results/p612-n725-score/latest.json— frozen g (725 never in the freeze)results/probe-invariant-shape/blindness-and-glift.json— historical 93.9° / 159.6° (withdrawn reading)Do not rerun N725. Do not become #622 (no new S theory). One comparison, one notes file.
Deliverable
notes/issue-633-glift-exact-lab-YYYYMMDD.md+ JSON of the angles/vectors. Draft PR against main. Equal weighting is sensitivity, not a second vote.If an anchor or covariance is missing, leave the angle undetermined. Do not invent a conclusion to fill the ticket.
No STATUS, no merge, do not close #633/#622. Wait for
ASSIGNED_MACHINE.Related: #633, #632, #653, #655, #622.