The manual console on two engines

Every manual preset through the engine the console uses today (the frozen seed) and through the one pipeline the rest of the house uses (profile source → placement → field solve) — side by side, measured. Station SHAPE in mm at the nine motors (each extension relative to the centre motor — placement-independent; each path's own placement is named on its verdict line), lower section; both paths locked to the centreline so stations align; glass thickness 0 for a clean read; field stations on the middle glass row of a 3000 mm sheet in the WIRE direction (the seed's own question — the free pipeline outcome is on the verdict line); tolerance 3 mm. Off-glass motors differ BY DESIGN (the seed's owned continuation vs the field's free-wire physics) — Δ is read on in-glass stations only. Seed = the console's own default read ('max-margin', the owned continuation).

What this sheet concludes — and what it does not ask. The comparison surfaced ONE structural divergence: CHAINS REST DIFFERENTLY between the engines. The seed births a chain heading-horizontal and then SOLVES the resting tilt — the restoring-balance physics you ruled in the project's first week (a formed piece must sit stably on its wires). The pipeline renders the authored tilt as given: it never absorbed that physics. The divergence is not a convention to choose between — it is the symptom of a missing law. So the engineering conclusion is already made: the flip is NOT YET. The resting/balance law lands in the field first (the same placement machinery that owns direction, lateral and the transfer datum gains the section tilt, solved under balance + margins); then manual ≡ auto holds by construction and chain files change only where ruled physics says so. Singles agree by pin today. The italic Δ values below are this missing-law divergence, not solver error — each engine follows its own target within the wire band. The question this sheet asked you was ANSWERED 20.07 — and it was a study, not a switch. Your terrain: the more lopsided the piece, the more it wants to settle; each roller-row hand-off lets it WALK a step toward that rest; a piece off its minimum-rectangle orientation may (or may not) also right itself — and none of this is directly measurable from the floor. So neither engine's assumption becomes law: the first-week tilt rule is a thesis-model, render-as-commanded is the other, and THE STUDY decides (KB-006): every solve gains a plain settling-exposure readout, every fidelity scan already measures the walk (the fitted pose — now declared as that reading), and production accumulates the evidence for free. The flip follows once the exposure readout and the declared walk reading land — with the study watching, nobody's assumption baked in. Nothing on this sheet needs a decision any more.

Single arc — the console defaultR500 sag, developed 1200

motor	1	2	3	4	5	6	7	8	9
seed · shape	941.7	933.8	817.2	475.4	0.0	-227.6	-29.7	455.9	817.0
field · shape	—	—	—	—	0.0	—	—	—	—
Δ	—	—	—	—	0.00	—	—	—	—

seed forms (per-wire read; centre extension 248.2)
field does not form at tolerance 3 — reported in the wire direction; worst deviation 87.8 mm
field wire-direction reading: refuses (3 distinct causes)

The field refuses this one in the wire direction: an R500 section sits EXACTLY on the glass floor, and the sampled read lands in the measured gray band (R450 gates / R510 clears — an authored R500 reads ≈486). Behind that stands the open wire-floor question: R500 glass implies a wire bend near 389 against the 500 wire floor. Both belong to the wire-yield ruling. Along travel the 45.5 mm row-link law limits the shape to ≈4.7 mm (forms at tolerance 6). The seed answers the per-wire question and is satisfied; the field answers the whole-machine question and is not — at tolerance 3, with today’s constants, honestly.

S with straight middleR900 sag 700 · straight 300 · R900 hog 700

motor	1	2	3	4	5	6	7	8	9
seed · shape	315.3	311.5	297.0	195.9	0.0	-71.9	69.3	66.9	-128.9
field · shape	—	—	—	35.5	0.0	-36.4	—	—	—
Δ	—	—	—	160.38	0.00	35.54	—	—	—

seed forms (per-wire read; centre extension 520.8)
field forms — shape along travel, worst deviation 0.31 mm, station agreement worst Δ 160.38 mm over 3 stations
field wire-direction reading: refuses (1 distinct causes)

In the wire direction an S rides the wire’s own transition band between the two bends (the commissioning-template class the steel gauge measures — several mm). At tolerance 3 the field therefore forms this ALONG TRAVEL instead (stepped straight wires) and says so; the station table compares the wire direction, the seed’s own question. An S also has no single up-side — the transfer-datum law names mixed sense open, so the field margin-places vertically.

Bi-arc SR900 sag 700 · R900 hog 700

motor	1	2	3	4	5	6	7	8	9
seed · shape	275.6	272.5	254.5	164.8	0.0	-70.7	53.5	-48.9	-197.8
field · shape	—	—	—	-44.5	0.0	6.4	—	—	—
Δ	—	—	—	209.36	0.00	77.08	—	—	—

seed forms (per-wire read; centre extension 578.8)
field forms — shape along travel, worst deviation 0.17 mm, station agreement worst Δ 209.36 mm over 3 stations
field wire-direction reading: refuses (3 distinct causes)

Same wire-transition-band and mixed-sense notes as the S. The chain flows through both engines deterministically.

Deep S — the recorded divergence caseR500 sag 700 · R500 hog 700

motor	1	2	3	4	5	6	7	8	9
seed · shape	480.3	474.7	442.6	283.7	0.0	-145.5	71.8	-75.3	-322.2
field · shape	—	—	—	-26.7	0.0	327.6	—	—	—
Δ	—	—	—	310.40	0.00	473.06	—	—	—

seed forms (per-wire read; centre extension 541.5)
field forms — shape along travel, worst deviation 0.21 mm, station agreement worst Δ 473.06 mm over 3 stations
field wire-direction reading: refuses (4 distinct causes)

The 17.07 record said "seed solves it, field refuses". Decomposed 19.07: at glass length 3000 the field solves it CLEAN (shape along travel, deviation 0.6). At the 8350 default the across-the-wires layout exceeds the 3600 forming width (that direction is geometrically unavailable) and the wire-direction solve is the deep-relief row-differential case that already owns its own wave (the alternating-field work, Prada acceptance). Not a defect — a boundary, named.

The standing facts behind the divergences (measured 19.07)

Chains rest by convention. The seed's heading-horizontal birth and the pipeline's canonical-frame + datum-law placement tilt the same S differently on the bed. The machine slumps the sheet from its transfer plane (the ruled transfer-datum law) — the pipeline's rest is the law's rest; the seed's was an authoring convention. The ruling decides which rest the manual console's files carry.
The two engines answer different questions. The seed answers "can a wire form this profile" — one row, one profile. The field answers "can the machine form this sheet" — placement, direction, row-to-row articulation, stroke, laws. Where they disagree on these cases, the field is carrying constraints the seed never sees.
Along travel, the row-link law is real: stepping an R500 profile down the bed demands ≈51 mm of row-to-row travel at the steep ends against the 45.5 mm the link hardware allows — the shape gives ≈4.7 mm and the verdict says so. At tolerance 6 it forms.
At the glass floor, the sampled radius reads under: an authored R500 section reads ≈486 through the section reconstruction (the measured band: R450 gates, R510 clears). With the open wire-floor question (R500 glass ⟹ wire ≈389 vs floor 500) this belongs to the wire-yield ruling — queued, on the table.
Long sheets: beyond 3600 mm developed the across-the-wires layout is geometrically unavailable, and deep alternating shapes land in the alternating-field work already scoped (Prada acceptance). The field says this honestly rather than solving past it.

Generated by kernel/verify/gen-w4-sheet.ts · 2026-07-22 10:36 · engine E-021 @ 102c310 · seed at the console default 'max-margin' (the 'standard' oracle read wraps its raw arc extension on tight radii — the recorded seed-era boundary); comparison is relative station shape, placement named per path · agreement band 1.0 mm (provisional) · station grid 510 mm.