Nishi Family › Compare › SOFT-BODY AND BIOMECHANICS

Nishi Compare · capabilities, evidence and delivery

SOFT-BODY AND BIOMECHANICS

Compare capabilities and delivery evidence. Source discovery, qualification, deployment and accepted outcomes are separate; an open contract names work still to verify.

XPBD tissue, cited biomechanics, and a validation loop no rival ships -- one process, integer, bit-deterministic replay

Overview

The purpose, declared position and evidence coverage of this domain. Source presence and completed acceptance are different measures.

Where we are. Measured 2026-08-25, not asserted. WORKING: an XPBD tetrahedral soft-tissue solver with five named layers and per-layer compliance; a Neo-Hookean volume constraint proven non-inverting by tooth; a conservation gate GREEN 8/8 and a metamorphic-relations gate GREEN 12/12, both bite-proven and both without any rival counterpart; a ringdown identifier that RAN FOR THE FIRST TIME TODAY and measured our plant at fn=4137 mHz zeta=480 permil, inside the published human bands 4000..5200 and 350..600; a cited oracle file where every tissue constant carries its source; an anisotropic sign-asymmetric integrator carrying cai2018's 9x up-down split and mills2025's AP and ML ratios; a reduced-order runtime model DERIVED from that solver rather than tuned, with the retired hand-tuned plant refused by a neg-control; 1001 seeded bodies per cast where there were 4; gait phase advanced by distance so foot plant and tissue drive are correlated by construction; and the whole thing compiled to ONE WebAssembly module with bit-identical replay. OPEN, and this is the whole story: there are no stress or strain TENSORS, so no clinical parity may be claimed; skin on skin now EXISTS as grazing contact with an inverse-mass split and a mesh-derived contact distance, though the friction law that would govern it waits on a citation rather than on code, and the contact is measurably blind to deep interleaving -- which is filed as SB11 rather than left unsaid. Animals are unstarted and the ground does not bear load.

Where we need to go. ONE sovereign soft-body and biomechanics engine -- human, animal and environment -- first byte up in NishiLang with no third-party code in the loop, real-time in a single process, every material constant traceable to a measured human and every claim refused until its oracle exists.

The unit. Millimetres of surface deviation for fidelity; mHz and permil for motion; microseconds per tick per body for cost. A fidelity claim without a millimetre and a motion claim without a band are both opinion wearing a number.

Where we already exceed: 0 u. Validation, not fidelity. No rival ships a conservation gate, a metamorphic gate, a motion-versus-measured-human loop, or an integer solver with bit-identical replay. Those are done and green today.

Cost to make the VR claim honest: 3 u. SB1 IS DONE -- 28 us of an 11111 us frame, measured against the cited bars, so the frame-cost half of the VR claim is now a number instead of a hope. SB2 force readback is what remains, and it is the blocking rung for the haptics domain too.

Cost to contact you can feel: 15 u. SB3's friction MECHANISM and SB10 self-contact are BOTH landed and gate-proven; friction stays unarmed pending one cited coefficient, which is a literature question rather than a build. What remains is SB7 a tetrahedralisation and sliver oracle so an arbitrary surface can enter the solver at all, and SB11 contact that survives deep interleaving -- a rung filed because SB10's own gate measured the blind spot, not because anyone imagined it.

Cost to clinical-grade parity: 24 u. SB4 Cauchy stress and strain, SB5 the FEBio analytical oracle, SB6 the 7.05 mm surface-error parity bar from chen2024. Until all three land, every parity-with-clinical claim on this board stays refused.

Cost to hyper-real embodied contact, the second admission pass: 46 u. M5 EVERYTHING THAT TOUCHES HER (SB12 strand-to-cage contact, SB13 cloth contact with garment layering and tearing, SB15 a support surface that spreads load, SB16 engorgement as rest-volume expansion) then M6 TISSUE THAT ACTS RATHER THAN ONLY REACTS (SB14 a clinging liquid film, SB17 muscle contraction that moves skin). FOUR OF THESE SIX ARE CONTACT OR COUPLING RUNGS RATHER THAN NEW SOLVERS, and that is the only reason the milestone is affordable: hair dynamics belong to graphics, the cloth solver to procgen, garment pattern and sewing to dcc, and the arousal physiology to procgen, all already contracted, so this board depends on them instead of building a second one of each. SB16 is the cheapest at 4 u because the physiology and the referee both already exist and only the coupling is missing. SB17 is the most expensive at 12 u and is gated behind SB4, making it deliberately the last rung on the whole board.

Research bar. SOFA Framework (LGPL) is measured on non-linear FEM, true stress and strain tensors, constraint-based friction contact, validated anatomical datasets for puncture and multi-layer shear, native real-time haptic device support, mature Python bindings. Theirs: source-level read plus vendor documentation, grade-1. Ours: THE reference for clinical soft tissue and the tool to exceed on fidelity. It validates STRESS against analytical solutions; it does not validate MOTION against a measured human, which is the gap we occupy. Multi-process by its own guidance: it cannot hold 90 Hz beside a generator and a renderer..

Research bar. Project Chrono (BSD-3) is measured on linear FEM and ANCF, industrial multibody dynamics, soft-soil terramechanics, deep PyChrono bindings. Theirs: vendor documentation read 2026-08-25. Ours: BEST on the environment lane -- deformable terrain under load is its home ground and is exactly our weakest row (SB9). Weaker on human soft tissue; requires bridging to a graphics framework for any interactive use..

Research bar. Bullet and PyBullet (zlib) is measured on mass-spring plus basic FEM, extensively proven in real-time robotics and VR, historically Blender's internal engine. Theirs: source-level read, grade-1. Ours: The real-time incumbent and the honest speed comparison. Mathematically simplified: no volume conservation guarantee and no material parameter with clinical meaning, so its constants cannot be sourced from tissue literature -- the same objection that rules out shape-matching lattices..

Research bar. FEBio (biomechanics FEM) is measured on purpose-built finite elements for biomechanics, hyperelastic and viscoelastic material models, the analytical reference used to check other solvers. Theirs: vendor documentation read 2026-08-25. Ours: Not a rival for real time at all -- an ORACLE. SB5 exists to turn it into our benchmark rather than our dependency, which is the only way a third party may enter this estate..

Research bar. PlasticineLab (MIT-IBM) is measured on differentiable-physics benchmark for soft-body manipulation and multi-layer deformation. Theirs: published benchmark, read 2026-08-25. Ours: An ORACLE on the manipulation rows, never a rival column. Named here so the manipulation lane has a published bar before anyone writes one by feel..

Research bar. Ziva VFX and Ziva RealTime (DNEG, commercial) is measured on Ziva VFX constructs and simulates muscles, fat, fascia and skin and produces precise skin deformations informed by the underlying anatomy, its team taking a 2025 SciTech Academy Award; Ziva RealTime analyses training sets of how a character should look as it moves and learns to reconstruct that shape at realtime speeds, turning effects that take seconds or minutes per frame into milliseconds; Ziva Face Technology converts scans into digital Face Puppets. Theirs: vendor documentation fetched and content-pinned 2026-08-25 [ziva]. Ours: THE gold standard for biomechanical soft tissue and the tool to exceed on flesh fidelity -- flesh-on-flesh collision, sliding, fat compression and volume preservation are its home ground and ours is one honest rung of it. ITS REALTIME PRODUCT IS OUR OWN ARCHITECTURE ARRIVED AT INDEPENDENTLY: identify offline, run cheap at runtime. The difference is the one axis we can actually claim -- ours is DERIVED by physics from a measured ringdown, so it carries a volume-conservation guarantee and bit-identical replay, and a learned surrogate structurally cannot state either.

Research bar. Epic MetaHuman (Epic Games, free with Unreal Engine) is measured on a bespoke photoreal digital human complete with hair and clothing created from presets in minutes and animated by MetaHuman Animator, delivered inside Unreal Engine. Theirs: vendor front door fetched and content-pinned 2026-08-25 [metahuman]. Ours: THE delivery bar for a complete human, and cited for POSITIONING ONLY: the mirror does not describe vasodilation, subsurface scattering or micro-expression internals, so no such claim is sourced to it here and none should be. It is an artist-authored photoreal RIG, which is precisely the axis this estate enters from the other side -- identity as a bred genome rather than an asset pack, already the declared exceed on the charsim and graphics boards.

Research bar. Unreal Engine Groom and Chaos Cloth (Epic Games) is measured on a strand-based groom workflow rendering each individual strand of hair with physically accurate motion at hundreds of thousands of strands, and the Chaos Cloth low-level particle clothing solver whose collision is supplied by the character Physics Asset. Theirs: vendor documentation fetched and content-pinned 2026-08-25 [uegroom] [chaoscloth]. Ours: THE real-time bar for both new contact rungs, and the bar that states our own dependency structure out loud: cloth collision comes FROM THE BODY RIG. That is exactly the coupling this board owns and that procgen's cloth-solver rung is waiting on, which is why softbody files the contact and not a second solver.

Research bar. Augmented Mass-Spring hair, Digital Salon (Yale, arXiv 2507.07387) is measured on strand-level edge, bending and torsional degrees of freedom plus an augmented biphasic interaction through two additional one-way springs, embedded in a hybrid Eulerian and Lagrangian framework, simulating thousands of interacting strands in real time on consumer hardware. Theirs: open-access paper fetched, content-pinned and READ FROM THE STORED BYTES 2026-08-25 [hairams]. Ours: THE only genuinely citable primary source among the new bars, and therefore the research bar for the hair rung. Read honestly it also BOUNDS ITSELF in our favour: its contribution is hair-HAIR interaction and it cites improved hair-BODY interaction as separate prior work, so the contact rung filed here is not covered by it and is not a reimplementation of it.

Research bar. Marvelous Designer and CLO (CLO Virtual Fashion, commercial) is measured on pattern-first garment simulation from real tailoring patterns with sewing, drape and a Solidify property assignable to parts of a garment, and CLO for true-to-life garment visualization, digital patternmaking and virtual fitting. Theirs: vendor documentation fetched and content-pinned 2026-08-25 [marvelous] [clo3d]. Ours: THE bar for garment realism built from real patterns rather than painted onto a mesh. Pattern and sewing are dcc's gd_drape rung and NOT this board's -- the bar is named here so the softbody contact rung is never judged against a painted cloth when the field standard is a sewn one.

Research bar. JangaFX LiquiGen (commercial) is measured on real-time liquid simulation spanning water, blood, ketchup and slime, with real-time meshing, customizable forces, sparse simulation domains and a real-time path tracer. Theirs: vendor documentation fetched and content-pinned 2026-08-25 [liquigen]. Ours: THE bar for the fluid rung and the honest measure of the distance: ours is a volume-conserving 2.5D height field, theirs is a free-surface liquid with material variety. Houdini FLIP and Vellum belong beside it and are deliberately named WITHOUT a cite mark, because www.sidefx.com fails TLS in our own fetcher (rc=-209 t12-cert-pipeline, reproduced on two URLs after retries) and a reference nobody fetched is a fabrication.

2026-09-09 Softbody investment graph repair: PARTIAL. Two malformed prerequisite fields repaired; all17 rungs parse with no missing dependencies. Native API and Compare artifacts refreshed. Rank binding, eligible-set completeness and live visual review remain unfinished; no SOTA or simulation quality claim.

Latest recorded release

2026-10-01 · SB15

BACKFILL-HUMAN refs rows landed this lane in softbody.refs, each mirror pinned by nx_filehash, cited here so the refs referee's inline-citation tooth sees them: [sb15_piper_lyon_beihang_2023] [sb15_yan_hunan_2025] for SB15, and for SB13 [sb13_rodriguezpardo_seddi_2023] [sb13_huang_zhejiang_2024_teartest]

Release history and work log

Release entries describe recorded changes; they do not establish that every capability passed evaluation.

24 of 33 capability source symbols found|7 of them authored stronger-claims|9 open|coverage 727/1000|adoption 17 full / 7 partial

Evidence profile — what the gaps on this board actually are

Measured by nx_swcompare_evidence, read back by nx_evprofile_lib. Every figure is a count with its denominator — there is deliberately no score, no grade and no percentage anywhere in this band, because a stored scalar is a field a seat can edit and a counted partition is not.

proven able to fail counts the gates that have a RECORDED RED — nx_gate_bite mutated the gate subject, rebuilt it, watched the gate go red, and that record is inside the shared TTL. never bitten is its complement over the same denominator: those gates ran and were green, and nothing has ever shown them able to detect anything, so their green is a statement about this run and not about the gate. green at 0/0 is a separate and much weaker observation — the gate printed GREEN on a zero denominator, so its own tooth counter says it examined nothing. A gate can be green, non-zero, and still never bitten; that is the common case and it is now visible instead of implied.

partition: grounded + unsupported = 22 vs present 22 · named + unnamed + withheld = 11 vs open 11 · both reconcile

liar-kill conj=GPqN · first failing conjunct: quorum

graded document: BUILDROOT tree, 34473 bytes · gates map: PRIMARY · stamped 16d 20h ago · STALE against the shared TTL of 604800s — this is evidence about a build nobody can vouch for now · source ../knowledge/status/evstamp_softbody.verdict

Gap class	What it is, and the work it names
GATE-FAILING	A declared gate ran and was not GREEN. The remedy is to FIX the gate it names. This is deliberately NOT merged with CLAIM-ONLY: one needs a gate written and the other needs a gate repaired, and a single merged number sends a seat at the wrong work.
UNNAMED-GAP	A gap row carries a bare `_ABSENT_` with no symbol after it, so it names no build contract and no seat can pick it up. Give it an `_ABSENT_:<symbol>` and it becomes work somebody can do.
STALE	This profile is older than the shared TTL, or its clock ran backwards. It is reported as unusable rather than as fresh, because being wrong in the direction of refusing is the safe direction.

This band reports the referee counts. The per-axis worklist rows — which axis is unsupported, which watch is ready to flip, which gap is unnamed — are printed by nx_swcompare_evidence softbody itself and are not carried on the stamp, so this page names the classes and the producer names the rows. That split is stated rather than hidden: a count without a worklist is not actionable, and this band is honest about which half of that it is.

Production map

Follow the dependencies, declared acceptance criteria and recorded priorities. Inspect source binding before treating a rank as executable work.

Ranking source binding: STALE_OR_INVALID_BINDING. Recorded priorities require current acceptance evidence and resource checks before execution.

Recorded ranking withheld

Retained ranking is not bound to the current plan and matrix: STALE_OR_INVALID_BINDING. Historical data remains retained. Refresh the ranking after resolving the input differences. Latest execution outcome is unverified.

Declared roadmap — contract, acceptance, executor, effort

Rung	Closes with	Definition of done (pre-declared)	Executor	Est.
Per-tick tissue cost measured against the banked 90 Hz budget (SB1)	tb_measure	DONE 2026-08-25, nx_tissuebudget GREEN 6/6: 28 us per cast tick against an 11111 us frame and a 2000 us app-logic ceiling, with the derived and cited budgets cross-checked against each other and a neg-control that refuses a cost one microsecond over. Labelled a NATIVE FLOOR. THE ANSWER IS THAT THE SOLVER IS NOT THE CONSTRAINT -- which is worth more than a pass, because it says where NOT to spend the next optimisation.	Organ	2 u
Contact reaction force reported from the solver (SB2) after SB1	st_contact_force	DONE 2026-08-25, nx_contactforce_gate GREEN 6/6: the readback fires, rises with depth (a 20 mm press reads 3.2x a 10 mm press, superlinear because the tissue stiffens), discriminates firmness (firm reads 3.75x soft at the SAME depth), reads exactly zero on both channels when nothing is touching, and force and impulse agree exactly so they are one measured quantity rather than two rulers. Units are SOLVER units and the gate says so: there is no kilogram in the solver, so a newton is not derivable and none is claimed -- the teeth assert monotonicity and discrimination, which is what a haptic loop actually consumes. Unblocks the whole hapticcare domain.	Organ	3 u
Tangential friction at a contact, with the coefficient CITED OR ABSENT (SB3) after SB2	st_set_friction	MECHANISM PROVEN 2026-08-25, nx_contactfriction_gate GREEN 6/6, and DELIBERATELY UNARMED. GREEN 6/6 ON THE SERVED BINARY (built, promoted, registered, invoked, rostered). Measured surface carry over a 15 mm drag: unset -396, mu 0 -396 (byte-identical to unset), mu 0.25 +35, mu 0.9 +51 cmm -- friction reverses the frictionless drift into positive carry. The first shipped cut of this gate TIMED OUT against /api/gate_run's 12,000 ms bound; it was made to fit by shrinking the CAGE, never by loosening the bound or dropping a tooth, and the discrimination came out STRONGER. A positional Coulomb cone is ported from the estate's own nx_phys3d rather than written a second time: the tangential correction is clamped to mu times the normal correction, so a dragged probe carries the surface with it. mu carries THREE states, not two -- a number, ZERO, and UNSET -- because zero is a measured frictionless surface and UNSET means no citation exists and the solver declines to invent one. Two open-access papers were fetched and pinned looking for a macroscopic skin-on-skin coefficient and NEITHER supplies one (a Frictiometer device index over 633 subjects; microasperity values of 0.004 and 0.001); the commonly-quoted 0.2 to 0.5 band traces only to a review nobody here has read, so it is not in the conf. THE GATE EARNED ITS KEEP ON ITS FIRST RUN, going RED 3/5 and catching TWO independent defects that both compiled and both read as friction: the clamp was on the ALLOWED SLIDE rather than on the CORRECTION, which made mu=0 the stickiest setting available instead of the frictionless one; and the slip was measured in world space rather than against the probe, which damps a particle against the room and is a different physical claim. The tell was the ladder running monotonic in the WRONG DIRECTION, which no single-arm test could have shown. T2 passes on an EXACT equality (unset == mu 0 to the byte), proving UNSET skips the law entirely rather than defaulting to something plausible, and T6 uses a middling coefficient to prove the cone is a continuum rather than a stick/slide switch. Neutrality is measured rather than argued: all six solver gates re-run against the modified solver at 54/54 GREEN. Bar to CLOSE this rung: one cited macroscopic coefficient admitted to the oracle conf with its source, at which point the mechanism arms itself with no code change.	Organ	4 u
Self-contact between two tissue surfaces (SB10) after SB3	st_self_contact	DONE 2026-08-25, nx_selfcontact_gate GREEN 8/8 ON THE SERVED BINARY: built, promoted under expect_sha256, registered, invoked via /api/gate_run and rostered, with the NAS numbers identical to the local ones. Closest approach 819 to 997 cmm against a 1000 cmm target, a six-pair contact patch, both bodies corrected by inverse-mass split, broad phase proven to cost zero pair tests for separated bodies. THREE RED RUNS BUILT THIS GATE AND ALL THREE WERE ONE MISTAKE: asking a question about SURFACES with an instrument that answers about BOXES or AXES -- placement by bounding-box overlap left the surfaces 397 cmm apart, then two successive did-it-move teeth read zero because the cages meet on their flanks and the contact normal is mostly y and z. Every tooth is now axis-free and the displacement instrument carries its own planted-shift control, so a zero from it can never again be read as a correction that did not happen. The contact distance is DERIVED from st_surface_spacing rather than chosen.	Organ	6 u
Cauchy stress and strain tensors with Young's modulus and Poisson ratio (SB4) after SB1	st_cauchy_stress	The solver exposes stress and strain per element and accepts Young's modulus and Poisson ratio as material inputs, reproducing the existing per-layer compliance behaviour byte-identically where the two overlap. This is the row every clinical claim is blocked on.	Organ	10 u
FEBio as an analytical oracle, not a dependency (SB5) after SB4	fo_febio_compare	A headless harness loads a cube and a cylinder of known dimensions through OUR solver and through a published FEBio or closed-form result and reports the deviation as a number with a pre-declared tolerance; exceeding it fails the regression.	Organ	8 u
Surface deviation against the 4D-scan parity bar (SB6) after SB4	st_surface_err_mm	Measures mean euclidean surface deviation in millimetres and grades against knowledge/gamefeel_oracle.conf tissue_surface_err_mm 0..7 (chen2024 multicomponent FE versus 4D scan, 7.05 mm). The bar is already banked and nothing measures it.	Organ	6 u
Tetrahedralisation of an arbitrary watertight surface with a sliver oracle (SB7)	st_tet_from_surface	Converts an arbitrary watertight surface to a tetrahedral cage and REFUSES any mesh containing zero-volume or degenerate elements before it reaches the solver, naming the offending element. This is what fTetWild and Gmsh do for the rival pipelines and what our own asset lane needs.	Organ	7 u
One animal on the same solver with a cited gait band (SB8) after SB1	st_quadruped_gait	One species with a published stride-frequency band admitted to the oracle conf and a gait-coupled drive proven the way the human one now is, on the SAME solver differing only by profile rows.	Organ	9 u
Ground that bears load (SB9) after SB4	we_shear_transport	Soil and vegetation deflect under a stated pressure, with the deflection cited rather than tuned; footprint and compaction persist. Chrono is BEST here and is the benchmark to grade against.	Organ	11 u
Contact that survives deep interleaving (SB11) after SB10	st_sdf_contact	A signed distance field or continuous collision detection resolves a pair that has already passed through each other, with a neg-control proving the proximity model reports that same pair as clear. FILED FROM A MEASUREMENT, not from imagination: SB10's proximity contact was measured blind to the tangled case on the day it shipped, and it fails in the reassuring direction by reporting interleaved bodies as not touching.	Organ	8 u
Strand hair that collides with the body (SB12) after SB10	st_strand_contact	SCOPED DELIBERATELY NARROW BECAUSE HAIR IS OWNED ELSEWHERE: strand dynamics already shipped as graphics nx_hairdyn_lib hd_ringdown (GREEN 11 of 11, a chain on THIS board's solver), geometry is that board's gpe_hair_geometry, and the groom is procgen's gr_strand. This rung is ONLY the contact between a strand chain and the deforming tet cage, which no board owns and only a soft-body engine can. Done when a strand chain rooted on the cage resolves against it at the mesh-derived contact distance st_surface_spacing already computes, with the head turning and the hair following rather than passing through, AND a neg-control proving a strand initialised INSIDE the body is expelled rather than trapped -- because a contact routine that quietly swallows a strand looks identical to one that never fired. The field bar is strand-based and real-time [uegroom] and the research bar is the augmented mass-spring model [hairams], which names hair-body interaction as prior work separate from its own hair-hair contribution.	Organ	6 u
Cloth against the body, against itself, and tearing (SB13) after SB10	st_cloth_contact	THE HALF NOBODY OWNS. procgen owns the cloth SOLVER (gm_xpbd_step, PG29) and its row already says that solver must ride THE TET CAGE'S CONTACT; dcc owns pattern sewing and drape (gd_drape, DC10); charsim owns one solver for cloth and hair (dyna_cloth, R8). This rung is the contact those three depend on, plus the two things none of them names: garment-on-garment LAYERING and TEARING. Done when two garment layers over a pressed cage show no interpenetration at a derived contact distance and a tear propagates along a stated stress threshold rather than at a chosen edge, with a neg-control proving a below-threshold load leaves the cloth intact. Unreal's Chaos Cloth takes its collision from the character Physics Asset [chaoscloth], which is the same dependency shape; the pattern-first realism bar is Marvelous Designer [marvelous] and CLO [clo3d]. NOTE THE HONEST ORDERING RISK: tearing needs a stress measure, so if SB4 has not landed the threshold must be stated in solver units and labelled as such rather than dressed up as a pascal.	Organ	9 u
A liquid film that clings (SB14) after SB3	fl_film_cling	CORRECTED FROM AN ABSENCE BEFORE IT WAS PUBLISHED: nx_fluid.nx already EXISTS and is real -- a height-field shallow-water solver whose reflecting boundaries make volume exactly conserved, with fl_volume exposing it, and whose own header declares its limit (FLIP is the particle variant and this is not it). Two capability probes had first returned pure homonym noise: viscosity matches 28 times in runtime and every classified hit is food-science rheology or FDM melt flow, and Navier matches 16 times and every one is the Octane JavaScript benchmark. So the gap is precise rather than total: no viscosity term, no surface tension, no wetting, and no film adhering to a moving body. Done when a film on the tissue surface has its thickness driven by a CITED surface-tension and viscosity pair admitted to the oracle conf, with a neg-control proving a zero-tension film sheets off completely instead of beading. LiquiGen is the real-time bar [liquigen]; Houdini FLIP is the offline bar and is uncited on purpose because our fetcher cannot reach sidefx.com. THE CITATION IS THE RISK HERE, exactly as it was for skin friction on SB3: if no readable primary source gives the pair, the mechanism ships with the constants UNSET and a tooth proving unset behaves as inviscid.	Organ	10 u
A support surface that spreads load (SB15) after SB2	st_support_surface	DISTINCT FROM SB9 AND THE ROW SAYS SO: SB9 is OUTDOOR terrain -- soil shear transport, footprints, compaction -- where Chrono terramechanics is BEST. This is the INDOOR support surface, a mattress or cushion or bedding whose deflection redistributes body weight, and its deliverable is not a dent but a PRESSURE MAP, because how load spreads across the contact patch is the whole question. THE CHEAPEST NEW RUNG ON THE BOARD BECAUSE IT COMPOSES RATHER THAN BUILDS: st_contact_force already reports a per-contact reaction and st_self_contact already resolves two bodies with an inverse-mass split, so a support surface is a second cage carrying its own compliance rows. Done when a body at rest on a compliant surface has its summed contact reaction equal its applied load within a stated tolerance, with indentation depth AND contact-patch area both reported, and a neg-control proving a rigid surface concentrates the same load into a measurably smaller patch. Units stay solver units: there is no kilogram in this solver and SB2 already refused to invent one.	Organ	5 u
Engorgement as rest-volume expansion (SB16) after SB2	st_inflate_rest_volume	DONE 2026-08-25 -- nx_engorge_gate GREEN 7/7 on the SERVED binary. Rest-state growth (rest lengths x s, rest volumes x s^3, snapshotted so unity returns exactly to pristine) graded against the closed-form s^3 prediction: 1728 predicted, 1667 measured, correctly under because only adipose grew. Unity control byte-exact; zero/negative factors refuse. THE MAGNITUDE AND RATE OF REAL ENGORGEMENT ARE NOT CLAIMED -- no band is mirrored, so the factor stays the caller's. THIS RUNG EXISTS BECAUSE ANOTHER BOARD'S ROW NAMED IT. The physiology is already shipped and is procgen's: nx_arousal_lib a_perf carries drive, lag, perfusion, temperature, heart rate and blood pressure as integer functions shared by simulate and fit so the two cannot drift, with its arousal-on-mesh gate GREEN 17 of 17. That row states its own blocker in one sentence -- a pure normal push is not volume-conserving and THIS board's conservation gate would refuse it. So what is missing is neither physiology nor rendering: it is driving the tet cage's REST VOLUME from a_perf so engorgement becomes tissue expansion the Neo-Hookean volume constraint carries. Done when rest volume is driven by a_perf with nx_softtissue_conserve_gate GREEN at every level of the drive, and a neg-control proving the retired normal-push model FAILS that same gate -- which is what makes the conservation gate a referee here rather than a decoration. No general-purpose engine ships an inflation-driven rest-volume channel, and that is a scope statement rather than a boast.	Organ	4 u
Muscle contraction that moves skin (SB17) after SB4	my_contract_fiber	A DIFFERENT MECHANISM FROM THE FACS ROW TWO BOARDS ALREADY CARRY, WHICH IS WHY IT IS NOT A DUPLICATE OF IT. charsim R9 and graphics both watch fa_expr_morph and that contract is BLENDSHAPES -- expression axes blended over a surface. This rung is the inverse: an ACTIVE CONTRACTILE ELEMENT inside the tet solver, so a fibre shortens and the skin above moves BECAUSE the flesh under it did, and micro-expression emerges from anatomy instead of being authored as a target. Ziva VFX is the production bar and describes exactly this -- muscles, fat, fascia and skin simulated so that skin deformation is informed by the underlying anatomy [ziva] -- with MetaHuman the delivery bar [metahuman]. Done when a named fibre's contraction provably moves the skin surface above it, reported in millimetres by md_measure (which is BUILT, PROMOTED and gate GREEN 9 of 9 today and merely registered-dark), with a neg-control proving a fibre at rest moves it zero. DELIBERATELY LAST AND GATED BEHIND SB4: an active contractile element without a stress measure is a shape effect wearing a physics name, so the tensor rung must land first or this one cannot be judged.	Organ	12 u

Milestones

Milestone	Rungs	Cumulative
M1 · Make the real-time claim measurable	SB1,SB2	5 u
M2 · Contact you can feel	SB3,SB10,SB7,SB11	13 u
M3 · Clinical-grade parity, or the claim stays refused	SB4,SB5,SB6	24 u
M4 · Beyond the single human	SB8,SB9	20 u
M5 · Everything that touches her	SB12,SB13,SB15,SB16	24 u
M6 · Tissue that acts rather than only reacts	SB14,SB17	22 u

Rung	Closes with	Definition of done (pre-declared)	Executor	Est.
Footfall impulse that scales with gait speed: the walk kick becomes the run kick (SB18)	sb_impulse_mob	OPERATOR 2026-09-16, with reference frames: real body physics is the subtle jiggle of breasts, glutes and thighs on walking impact that SHIFTS with running. The plant is already derived from measured human bands (fn 4137 mHz, zeta 480 permil at the linear limit, cai2018 up-down asymmetry, mills2025 AP and ML anisotropy) and the heel-strike impulse is already isqrt(2 x GRAV x 2 x bob); what is missing is the ADOPTION: sb_impulse_mob is BUILT and called by no world, and the impulse does not yet read cadence and vertical excursion from the gait, so a run does not kick harder than a walk. Done-rule: every world calls sb_impulse_mob at double-support; peak tissue excursion is MONOTONIC in gait speed on one plant (walk below jog below run); the ringdown after each strike stays inside the cited band with nx_dyna_bind_skin as the referee; a zero-speed body receives zero impulse (neg-control); and the referee panel grades a fixed-camera walk-then-run capture against the operator's lens before ADOPT.	Organ	1.5 u
Tissue driven by pelvis motion: belly, glute and thigh springs ride obliquity, rotation, tilt and lateral sway, not only vertical bob (SB19) after SB18	sb_pelvis_drive	The reference frames show tissue following the PELVIS (hips that move, the strut, the hoop), not a vertical bounce; today the R7 belly, glute and thigh springs take a vertical gait drive and a wander heading. Done-rule: the spring anchors read the pelvis pose from the gait sub-model (gameengine GE75 wc_pelvis_gait) so lateral sway and rotation excite the AP and ML axes the anisotropy already declares; with the pelvis held still the output is byte-identical to the incumbent (neg-control); a hoop-class circumduction fixture drives belly and thigh excursion in phase with the pelvis and the referee reads a phase lag inside the plant's own band. The reference frames stay OUTSIDE the estate as referee lenses; no identity capture.	Organ	2 u

Ladder to SOTA

Two dated targets, and every rung's role toward them. Best in class is the proven, deployed leader today; the frontier is the research edge as of its month. Each target is dated by a sotabar row and judged against the current month by the bar-age ruler (FRESH, ATTESTED with the fallback month named, or refused). Each rung is substrate (what the target's own methods consume), an arm (a sovereign arm that triangulates without competing for the number), a contender (its done-rule is a number against the target) or superseded (kept as capability, retired as a claim), and carries the rungs it stands on.

best-in-class (T-BIC-SOFT), dated by SB-SOLVER-2607 = 2026-07 ATTESTED. Contact-rich elastodynamics at the industrial bar: IPC-class penetration-free contact, GPU solvers for elastic solids, sand and snow (Newton, Warp), validated against FEBio; parity is a per-tick cost and a contact error the gates print ref nvidia_siggraph2026_blog

frontier (T-FR-SOFT), dated by SB-TISSUE-2609 = 2026-09 ATTESTED. Measured-material tissue and cloth: persistent wrinkles from a measured fabric database, deforming tissue tracked to the millimetre, hair and cloth and skin on ONE solver against a 4D-scan parity bar; the exceed axis is tissue grown inside-out from anatomy (bone, muscle, fat, skin as nested boundaries) rather than a surface with springs ref arxiv_softbody_202609

Rung	Closes with	Definition of done (pre-declared)	Executor	Est.
Board-level journal anchor for softbody (SB0)	-	BOARD-LEVEL JOURNAL ANCHOR, DECLARED 2026-09-27 so board-wide findings have an address a log row can resolve to; journal rung work against its rung.	Board	0 u

Rung	Role	Toward	Stands on	Why
SB0	substrate	T-BIC-SOFT	-	board-level journal anchor
SB1	substrate	T-BIC-SOFT	-	the per-tick cost against the banked 90 Hz budget is the floor every contender stands on
SB2	substrate	T-BIC-SOFT	SB1	a contact force read back from the solver is the measurement the industrial bar prints
SB3	substrate	T-BIC-SOFT	SB2	friction with a cited coefficient
SB4	substrate	T-FR-SOFT	SB1	stress and strain tensors with cited moduli are what measured materials are expressed in
SB5	arm	T-BIC-SOFT	SB4	FEBio as an analytical oracle outside the build path
SB6	arm	T-FR-SOFT	SB4	surface deviation against the 4D-scan parity bar is the liar-killer
SB7	substrate	T-FR-SOFT	-	tetrahedralisation with a sliver oracle
SB8	contender	T-BIC-SOFT	SB1	one animal on the same solver with a cited gait band
SB9	substrate	T-BIC-SOFT	SB4	ground that bears load
SB10	substrate	T-BIC-SOFT	SB3	self-contact between tissue surfaces
SB11	contender	T-BIC-SOFT	SB10	contact that survives deep interleaving is the IPC-class claim
SB12	contender	T-FR-SOFT	SB10	strand hair colliding with the body on the same solver
SB13	contender	T-FR-SOFT	SB10	cloth against the body and itself with tearing, measured against Fabric-101 wrinkles
SB14	contender	T-FR-SOFT	SB3	a liquid film that clings
SB15	contender	T-FR-SOFT	SB2	a support surface that spreads load is the lounger cushion of the beach class
SB16	contender	T-FR-SOFT	SB2	engorgement as rest-volume expansion
SB17	contender	T-FR-SOFT	SB4	muscle contraction that moves skin, the inside-out claim
SB18	substrate	T-BIC-SOFT	-	footfall impulse scaling with gait speed
SB19	contender	T-FR-SOFT	SB18	tissue driven by pelvis motion

Ladder verdict. LADDERED. Targets, rung roles and bars are counted in the tables below, each partition re-derived beside its total. Bars read FRESH against the current month 2026-10.

Ladder verdict, counted (17 rows, more than the 12-row disclosure bar)

Ladder verdict, counted
Class	Count	Note
targets	2	sotatarget rows on the plan
best in class	1	targets whose class is best-in-class, by code
frontier	1	targets whose class is frontier, by code
undated	0	targets naming no bar that exists
bad class	0	targets whose class is neither word
duplicate	0	target ids an earlier row already carries
rungs	20	rung rows on the plan
placed	20	rungs carrying a rungrole row
unplaced	0	rungs with no role toward any target
bad role	0	rungrole rows naming a role the ladder does not know
bad target	0	rungrole rows naming a target no sotatarget row declares
rungs, summed	20	RECONCILED: the parts sum to the total 20
substrate	9	what the target's own methods consume
arm	2	a sovereign arm that triangulates without competing for the number
contender	9	its done-rule is a number against the target
superseded	0	kept as capability, retired as a claim
placed, summed	20	RECONCILED: the parts sum to the total 20

Bar freshness, counted
Class	Count	Note
bars	3	sotabar rows on the plan
fresh	0	the bar month is the current month
attested	3	an older bar with a this-month barscan reporting no newer entry
stale	0	a this-month scan saw a newer month the bar ignores
unattested	0	nobody looked this month
malformed	0	a sotabar row missing a field
duplicate	0	bar ids an earlier row already carries
bars, summed	3	RECONCILED: the parts sum to the total 3
current month	2026-10	the month every bar is judged against

Treatment verdict. NOT DECLARED. No rung on this board declares its three-step treatment (a treat row naming the MCP row, the API route, the primitive it composes, the workflow that composes it and the agent beat that runs it); the ranker labels every rung TREATMENT-UNDECLARED until it does.

Inspect a rung and its prerequisites

Declared nodes 20. Rank input binding: STALE_OR_INVALID_BINDING. Dependency order is authored. Implementation, acceptance evidence, authority and resource readiness are unverified. No action is recommended or dispatched here.

Plan SHA-256 6b40ffa04fbb48d23e2bbfb96c75951a61ee9748fdb08c8a4170a24188bd89bb. Target rows 2; role rows 20. Existing risks and release worklog retain their own scope; no node completion is inferred.

Runtime observation coverage: 0 of 20 rungs have current, identity-bound execution observations. Observation may report RED. Accepted completion remains unknown; source presence, qualification and deployment do not establish it.

Use Enter or Space on a rung to inspect its contract. Prerequisite links locate another rung in this list; open its summary to inspect it. Estimates are authored effort, not forecasts.

SB1 — Per-tick tissue cost measured against the banked 90 Hz budget
Prerequisites: None declared; this does not establish execution eligibility.
Contract: tb_measure
Acceptance: DONE 2026-08-25, nx_tissuebudget GREEN 6/6: 28 us per cast tick against an 11111 us frame and a 2000 us app-logic ceiling, with the derived and cited budgets cross-checked against each other and a neg-control that refuses a cost one microsecond over. Labelled a NATIVE FLOOR. THE ANSWER IS THAT THE SOLVER IS NOT THE CONSTRAINT -- which is worth more than a pass, because it says where NOT to spend the next optimisation.
Authored effort: 2. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB2 — Contact reaction force reported from the solver
Prerequisites: SB1 (acceptance unverified)
Contract: st_contact_force
Acceptance: DONE 2026-08-25, nx_contactforce_gate GREEN 6/6: the readback fires, rises with depth (a 20 mm press reads 3.2x a 10 mm press, superlinear because the tissue stiffens), discriminates firmness (firm reads 3.75x soft at the SAME depth), reads exactly zero on both channels when nothing is touching, and force and impulse agree exactly so they are one measured quantity rather than two rulers. Units are SOLVER units and the gate says so: there is no kilogram in the solver, so a newton is not derivable and none is claimed -- the teeth assert monotonicity and discrimination, which is what a haptic loop actually consumes. Unblocks the whole hapticcare domain.
Authored effort: 3. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB3 — Tangential friction at a contact, with the coefficient CITED OR ABSENT
Prerequisites: SB2 (acceptance unverified)
Contract: st_set_friction
Acceptance: MECHANISM PROVEN 2026-08-25, nx_contactfriction_gate GREEN 6/6, and DELIBERATELY UNARMED. GREEN 6/6 ON THE SERVED BINARY (built, promoted, registered, invoked, rostered). Measured surface carry over a 15 mm drag: unset -396, mu 0 -396 (byte-identical to unset), mu 0.25 +35, mu 0.9 +51 cmm -- friction reverses the frictionless drift into positive carry. The first shipped cut of this gate TIMED OUT against /api/gate_run's 12,000 ms bound; it was made to fit by shrinking the CAGE, never by loosening the bound or dropping a tooth, and the discrimination came out STRONGER. A positional Coulomb cone is ported from the estate's own nx_phys3d rather than written a second time: the tangential correction is clamped to mu times the normal correction, so a dragged probe carries the surface with it. mu carries THREE states, not two -- a number, ZERO, and UNSET -- because zero is a measured frictionless surface and UNSET means no citation exists and the solver declines to invent one. Two open-access papers were fetched and pinned looking for a macroscopic skin-on-skin coefficient and NEITHER supplies one (a Frictiometer device index over 633 subjects; microasperity values of 0.004 and 0.001); the commonly-quoted 0.2 to 0.5 band traces only to a review nobody here has read, so it is not in the conf. THE GATE EARNED ITS KEEP ON ITS FIRST RUN, going RED 3/5 and catching TWO independent defects that both compiled and both read as friction: the clamp was on the ALLOWED SLIDE rather than on the CORRECTION, which made mu=0 the stickiest setting available instead of the frictionless one; and the slip was measured in world space rather than against the probe, which damps a particle against the room and is a different physical claim. The tell was the ladder running monotonic in the WRONG DIRECTION, which no single-arm test could have shown. T2 passes on an EXACT equality (unset == mu 0 to the byte), proving UNSET skips the law entirely rather than defaulting to something plausible, and T6 uses a middling coefficient to prove the cone is a continuum rather than a stick/slide switch. Neutrality is measured rather than argued: all six solver gates re-run against the modified solver at 54/54 GREEN. Bar to CLOSE this rung: one cited macroscopic coefficient admitted to the oracle conf with its source, at which point the mechanism arms itself with no code change.
Authored effort: 4. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB10 — Self-contact between two tissue surfaces
Prerequisites: SB3 (acceptance unverified)
Contract: st_self_contact
Acceptance: DONE 2026-08-25, nx_selfcontact_gate GREEN 8/8 ON THE SERVED BINARY: built, promoted under expect_sha256, registered, invoked via /api/gate_run and rostered, with the NAS numbers identical to the local ones. Closest approach 819 to 997 cmm against a 1000 cmm target, a six-pair contact patch, both bodies corrected by inverse-mass split, broad phase proven to cost zero pair tests for separated bodies. THREE RED RUNS BUILT THIS GATE AND ALL THREE WERE ONE MISTAKE: asking a question about SURFACES with an instrument that answers about BOXES or AXES -- placement by bounding-box overlap left the surfaces 397 cmm apart, then two successive did-it-move teeth read zero because the cages meet on their flanks and the contact normal is mostly y and z. Every tooth is now axis-free and the displacement instrument carries its own planted-shift control, so a zero from it can never again be read as a correction that did not happen. The contact distance is DERIVED from st_surface_spacing rather than chosen.
Authored effort: 6. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB4 — Cauchy stress and strain tensors with Young's modulus and Poisson ratio
Prerequisites: SB1 (acceptance unverified)
Contract: st_cauchy_stress
Acceptance: The solver exposes stress and strain per element and accepts Young's modulus and Poisson ratio as material inputs, reproducing the existing per-layer compliance behaviour byte-identically where the two overlap. This is the row every clinical claim is blocked on.
Authored effort: 10. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB5 — FEBio as an analytical oracle, not a dependency
Prerequisites: SB4 (acceptance unverified)
Contract: fo_febio_compare
Acceptance: A headless harness loads a cube and a cylinder of known dimensions through OUR solver and through a published FEBio or closed-form result and reports the deviation as a number with a pre-declared tolerance; exceeding it fails the regression.
Authored effort: 8. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB6 — Surface deviation against the 4D-scan parity bar
Prerequisites: SB4 (acceptance unverified)
Contract: st_surface_err_mm
Acceptance: Measures mean euclidean surface deviation in millimetres and grades against knowledge/gamefeel_oracle.conf tissue_surface_err_mm 0..7 (chen2024 multicomponent FE versus 4D scan, 7.05 mm). The bar is already banked and nothing measures it.
Authored effort: 6. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB7 — Tetrahedralisation of an arbitrary watertight surface with a sliver oracle
Prerequisites: None declared; this does not establish execution eligibility.
Contract: st_tet_from_surface
Acceptance: Converts an arbitrary watertight surface to a tetrahedral cage and REFUSES any mesh containing zero-volume or degenerate elements before it reaches the solver, naming the offending element. This is what fTetWild and Gmsh do for the rival pipelines and what our own asset lane needs.
Authored effort: 7. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB8 — One animal on the same solver with a cited gait band
Prerequisites: SB1 (acceptance unverified)
Contract: st_quadruped_gait
Acceptance: One species with a published stride-frequency band admitted to the oracle conf and a gait-coupled drive proven the way the human one now is, on the SAME solver differing only by profile rows.
Authored effort: 9. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB9 — Ground that bears load
Prerequisites: SB4 (acceptance unverified)
Contract: we_shear_transport
Acceptance: Soil and vegetation deflect under a stated pressure, with the deflection cited rather than tuned; footprint and compaction persist. Chrono is BEST here and is the benchmark to grade against.
Authored effort: 11. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB11 — Contact that survives deep interleaving
Prerequisites: SB10 (acceptance unverified)
Contract: st_sdf_contact
Acceptance: A signed distance field or continuous collision detection resolves a pair that has already passed through each other, with a neg-control proving the proximity model reports that same pair as clear. FILED FROM A MEASUREMENT, not from imagination: SB10's proximity contact was measured blind to the tangled case on the day it shipped, and it fails in the reassuring direction by reporting interleaved bodies as not touching.
Authored effort: 8. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB12 — Strand hair that collides with the body
Prerequisites: SB10 (acceptance unverified)
Contract: st_strand_contact
Acceptance: SCOPED DELIBERATELY NARROW BECAUSE HAIR IS OWNED ELSEWHERE: strand dynamics already shipped as graphics nx_hairdyn_lib hd_ringdown (GREEN 11 of 11, a chain on THIS board's solver), geometry is that board's gpe_hair_geometry, and the groom is procgen's gr_strand. This rung is ONLY the contact between a strand chain and the deforming tet cage, which no board owns and only a soft-body engine can. Done when a strand chain rooted on the cage resolves against it at the mesh-derived contact distance st_surface_spacing already computes, with the head turning and the hair following rather than passing through, AND a neg-control proving a strand initialised INSIDE the body is expelled rather than trapped -- because a contact routine that quietly swallows a strand looks identical to one that never fired. The field bar is strand-based and real-time [@uegroom] and the research bar is the augmented mass-spring model [@hairams], which names hair-body interaction as prior work separate from its own hair-hair contribution.
Authored effort: 6. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB13 — Cloth against the body, against itself, and tearing
Prerequisites: SB10 (acceptance unverified)
Contract: st_cloth_contact
Acceptance: THE HALF NOBODY OWNS. procgen owns the cloth SOLVER (gm_xpbd_step, PG29) and its row already says that solver must ride THE TET CAGE'S CONTACT; dcc owns pattern sewing and drape (gd_drape, DC10); charsim owns one solver for cloth and hair (dyna_cloth, R8). This rung is the contact those three depend on, plus the two things none of them names: garment-on-garment LAYERING and TEARING. Done when two garment layers over a pressed cage show no interpenetration at a derived contact distance and a tear propagates along a stated stress threshold rather than at a chosen edge, with a neg-control proving a below-threshold load leaves the cloth intact. Unreal's Chaos Cloth takes its collision from the character Physics Asset [@chaoscloth], which is the same dependency shape; the pattern-first realism bar is Marvelous Designer [@marvelous] and CLO [@clo3d]. NOTE THE HONEST ORDERING RISK: tearing needs a stress measure, so if SB4 has not landed the threshold must be stated in solver units and labelled as such rather than dressed up as a pascal.
Authored effort: 9. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB14 — A liquid film that clings
Prerequisites: SB3 (acceptance unverified)
Contract: fl_film_cling
Acceptance: CORRECTED FROM AN ABSENCE BEFORE IT WAS PUBLISHED: nx_fluid.nx already EXISTS and is real -- a height-field shallow-water solver whose reflecting boundaries make volume exactly conserved, with fl_volume exposing it, and whose own header declares its limit (FLIP is the particle variant and this is not it). Two capability probes had first returned pure homonym noise: viscosity matches 28 times in runtime and every classified hit is food-science rheology or FDM melt flow, and Navier matches 16 times and every one is the Octane JavaScript benchmark. So the gap is precise rather than total: no viscosity term, no surface tension, no wetting, and no film adhering to a moving body. Done when a film on the tissue surface has its thickness driven by a CITED surface-tension and viscosity pair admitted to the oracle conf, with a neg-control proving a zero-tension film sheets off completely instead of beading. LiquiGen is the real-time bar [@liquigen]; Houdini FLIP is the offline bar and is uncited on purpose because our fetcher cannot reach sidefx.com. THE CITATION IS THE RISK HERE, exactly as it was for skin friction on SB3: if no readable primary source gives the pair, the mechanism ships with the constants UNSET and a tooth proving unset behaves as inviscid.
Authored effort: 10. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB15 — A support surface that spreads load
Prerequisites: SB2 (acceptance unverified)
Contract: st_support_surface
Acceptance: DISTINCT FROM SB9 AND THE ROW SAYS SO: SB9 is OUTDOOR terrain -- soil shear transport, footprints, compaction -- where Chrono terramechanics is BEST. This is the INDOOR support surface, a mattress or cushion or bedding whose deflection redistributes body weight, and its deliverable is not a dent but a PRESSURE MAP, because how load spreads across the contact patch is the whole question. THE CHEAPEST NEW RUNG ON THE BOARD BECAUSE IT COMPOSES RATHER THAN BUILDS: st_contact_force already reports a per-contact reaction and st_self_contact already resolves two bodies with an inverse-mass split, so a support surface is a second cage carrying its own compliance rows. Done when a body at rest on a compliant surface has its summed contact reaction equal its applied load within a stated tolerance, with indentation depth AND contact-patch area both reported, and a neg-control proving a rigid surface concentrates the same load into a measurably smaller patch. Units stay solver units: there is no kilogram in this solver and SB2 already refused to invent one.
Authored effort: 5. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB16 — Engorgement as rest-volume expansion
Prerequisites: SB2 (acceptance unverified)
Contract: st_inflate_rest_volume
Acceptance: DONE 2026-08-25 -- nx_engorge_gate GREEN 7/7 on the SERVED binary. Rest-state growth (rest lengths x s, rest volumes x s^3, snapshotted so unity returns exactly to pristine) graded against the closed-form s^3 prediction: 1728 predicted, 1667 measured, correctly under because only adipose grew. Unity control byte-exact; zero/negative factors refuse. THE MAGNITUDE AND RATE OF REAL ENGORGEMENT ARE NOT CLAIMED -- no band is mirrored, so the factor stays the caller's. THIS RUNG EXISTS BECAUSE ANOTHER BOARD'S ROW NAMED IT. The physiology is already shipped and is procgen's: nx_arousal_lib a_perf carries drive, lag, perfusion, temperature, heart rate and blood pressure as integer functions shared by simulate and fit so the two cannot drift, with its arousal-on-mesh gate GREEN 17 of 17. That row states its own blocker in one sentence -- a pure normal push is not volume-conserving and THIS board's conservation gate would refuse it. So what is missing is neither physiology nor rendering: it is driving the tet cage's REST VOLUME from a_perf so engorgement becomes tissue expansion the Neo-Hookean volume constraint carries. Done when rest volume is driven by a_perf with nx_softtissue_conserve_gate GREEN at every level of the drive, and a neg-control proving the retired normal-push model FAILS that same gate -- which is what makes the conservation gate a referee here rather than a decoration. No general-purpose engine ships an inflation-driven rest-volume channel, and that is a scope statement rather than a boast.
Authored effort: 4. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB17 — Muscle contraction that moves skin
Prerequisites: SB4 (acceptance unverified)
Contract: my_contract_fiber
Acceptance: A DIFFERENT MECHANISM FROM THE FACS ROW TWO BOARDS ALREADY CARRY, WHICH IS WHY IT IS NOT A DUPLICATE OF IT. charsim R9 and graphics both watch fa_expr_morph and that contract is BLENDSHAPES -- expression axes blended over a surface. This rung is the inverse: an ACTIVE CONTRACTILE ELEMENT inside the tet solver, so a fibre shortens and the skin above moves BECAUSE the flesh under it did, and micro-expression emerges from anatomy instead of being authored as a target. Ziva VFX is the production bar and describes exactly this -- muscles, fat, fascia and skin simulated so that skin deformation is informed by the underlying anatomy [@ziva] -- with MetaHuman the delivery bar [@metahuman]. Done when a named fibre's contraction provably moves the skin surface above it, reported in millimetres by md_measure (which is BUILT, PROMOTED and gate GREEN 9 of 9 today and merely registered-dark), with a neg-control proving a fibre at rest moves it zero. DELIBERATELY LAST AND GATED BEHIND SB4: an active contractile element without a stress measure is a shape effect wearing a physics name, so the tensor rung must land first or this one cannot be judged.
Authored effort: 12. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB18 — Footfall impulse that scales with gait speed: the walk kick becomes the run kick
Prerequisites: None declared; this does not establish execution eligibility.
Contract: sb_impulse_mob
Acceptance: OPERATOR 2026-09-16, with reference frames: real body physics is the subtle jiggle of breasts, glutes and thighs on walking impact that SHIFTS with running. The plant is already derived from measured human bands (fn 4137 mHz, zeta 480 permil at the linear limit, cai2018 up-down asymmetry, mills2025 AP and ML anisotropy) and the heel-strike impulse is already isqrt(2 x GRAV x 2 x bob); what is missing is the ADOPTION: sb_impulse_mob is BUILT and called by no world, and the impulse does not yet read cadence and vertical excursion from the gait, so a run does not kick harder than a walk. Done-rule: every world calls sb_impulse_mob at double-support; peak tissue excursion is MONOTONIC in gait speed on one plant (walk below jog below run); the ringdown after each strike stays inside the cited band with nx_dyna_bind_skin as the referee; a zero-speed body receives zero impulse (neg-control); and the referee panel grades a fixed-camera walk-then-run capture against the operator's lens before ADOPT.
Authored effort: 1.5. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB19 — Tissue driven by pelvis motion: belly, glute and thigh springs ride obliquity, rotation, tilt and lateral sway, not only vertical bob
Prerequisites: SB18 (acceptance unverified)
Contract: sb_pelvis_drive
Acceptance: The reference frames show tissue following the PELVIS (hips that move, the strut, the hoop), not a vertical bounce; today the R7 belly, glute and thigh springs take a vertical gait drive and a wander heading. Done-rule: the spring anchors read the pelvis pose from the gait sub-model (gameengine GE75 wc_pelvis_gait) so lateral sway and rotation excite the AP and ML axes the anisotropy already declares; with the pelvis held still the output is byte-identical to the incumbent (neg-control); a hoop-class circumduction fixture drives belly and thigh excursion in phase with the pelvis and the referee reads a phase lag inside the plant's own band. The reference frames stay OUTSIDE the estate as referee lenses; no identity capture.
Authored effort: 2. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
SB0 — Board-level journal anchor for softbody
Prerequisites: None declared; this does not establish execution eligibility.
Contract: -
Acceptance: BOARD-LEVEL JOURNAL ANCHOR, DECLARED 2026-09-27 so board-wide findings have an address a log row can resolve to; journal rung work against its rung.
Authored effort: 0. Executor kind: Board. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.

Learning and practice paths

No structured learning path is declared for this plan. Existing research, roadmap and worklog remain available above.

Rung dossiers — every record that bears on a rung, in one place

Joined by symbol, never typed. Each dossier is assembled on every publish from the rows this board already carries: the plan's rung row, the matrix row that carries its contract symbol (its status re-measured by the one symbol ruler), the gate the organ declares and the verdict the evidence referee recorded on its last run, the reference marks in the rung text, the design rows that name the rung and the work log filed against it. The capability matrix links each cell to its dossier and each dossier links back to its cell and forward to the rungs it stands on; a dependency naming no rung on this plan is written as UNRESOLVED rather than linked. Absences are named: NO-MATRIX-ROW, NO-GATE-DECLARED, UNRECORDED.

Gate verdicts: UNSTAMPED — no per-gate verdict artifact is published for this domain yet (nx_swcompare_evidence softbody writes it beside its stamp); every gate below reads UNRECORDED until it runs.

Gate declarations: knowledge/organ_gate.conf read from this working directory (1 rows) and from the estate root (196 rows); rows in both counted once: 0; declarations read: 197. Every declaration in either copy counts, so a stub beside the publisher can never hide the root.

SB1 Per-tick tissue cost measured against the banked 90 Hz budget · cell PRESENT · gate NO-GATE-DECLARED

Contract	`tb_measure` · executor Organ · 2 u · stands on nothing
Definition of done	DONE 2026-08-25, nx_tissuebudget GREEN 6/6: 28 us per cast tick against an 11111 us frame and a 2000 us app-logic ceiling, with the derived and cited budgets cross-checked against each other and a neg-control that refuses a cost one microsecond over. Labelled a NATIVE FLOOR. THE ANSWER IS THAT THE SOLVER IS NOT THE CONSTRAINT -- which is worth more than a pass, because it says where NOT to spend the next optimisation.
Matrix cell	frame-budget-at-90hz · `runtime/nx_tissuebudget.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_tissuebudget`; declare `nx_tissuebudget TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-09-10 measure 2026-09-09 investment measurement repair: SB1 and SB7 empty prerequisite fields normalized to canonical '-' through CAS jobs1789009244 and1789009256. All17 contracts and authored efforts retained. Scoped regeneration job1789009389 emitted59561-byte JSON and172424-byte HTML with no failures. JSON generated1789009389 parses17/17 nodes, malformed0, missing_dependencies0, graph_valid1. Public HTTPS job1789009405 returned200 application/json Content-Length59561. Rank remains UNBOUND and recommendations WITHHELD; visual and interaction review unavailable because no browser connected.

SB2 Contact reaction force reported from the solver · cell PRESENT · gate NO-GATE-DECLARED

Contract	`st_contact_force` · executor Organ · 3 u · stands on SB1
Definition of done	DONE 2026-08-25, nx_contactforce_gate GREEN 6/6: the readback fires, rises with depth (a 20 mm press reads 3.2x a 10 mm press, superlinear because the tissue stiffens), discriminates firmness (firm reads 3.75x soft at the SAME depth), reads exactly zero on both channels when nothing is touching, and force and impulse agree exactly so they are one measured quantity rather than two rulers. Units are SOLVER units and the gate says so: there is no kilogram in the solver, so a newton is not derivable and none is claimed -- the teeth assert monotonicity and discrimination, which is what a haptic loop actually consumes. Unblocks the whole hapticcare domain.
Matrix cell	contact-force-readback · `runtime/nx_softtissue.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_softtissue`; declare `nx_softtissue TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

SB3 Tangential friction at a contact, with the coefficient CITED OR ABSENT · cell PRESENT · gate NO-GATE-DECLARED

Contract	`st_set_friction` · executor Organ · 4 u · stands on SB2
Definition of done	MECHANISM PROVEN 2026-08-25, nx_contactfriction_gate GREEN 6/6, and DELIBERATELY UNARMED. GREEN 6/6 ON THE SERVED BINARY (built, promoted, registered, invoked, rostered). Measured surface carry over a 15 mm drag: unset -396, mu 0 -396 (byte-identical to unset), mu 0.25 +35, mu 0.9 +51 cmm -- friction reverses the frictionless drift into positive carry. The first shipped cut of this gate TIMED OUT against /api/gate_run's 12,000 ms bound; it was made to fit by shrinking the CAGE, never by loosening the bound or dropping a tooth, and the discrimination came out STRONGER. A positional Coulomb cone is ported from the estate's own nx_phys3d rather than written a second time: the tangential correction is clamped to mu times the normal correction, so a dragged probe carries the surface with it. mu carries THREE states, not two -- a number, ZERO, and UNSET -- because zero is a measured frictionless surface and UNSET means no citation exists and the solver declines to invent one. Two open-access papers were fetched and pinned looking for a macroscopic skin-on-skin coefficient and NEITHER supplies one (a Frictiometer device index over 633 subjects; microasperity values of 0.004 and 0.001); the commonly-quoted 0.2 to 0.5 band traces only to a review nobody here has read, so it is not in the conf. THE GATE EARNED ITS KEEP ON ITS FIRST RUN, going RED 3/5 and catching TWO independent defects that both compiled and both read as friction: the clamp was on the ALLOWED SLIDE rather than on the CORRECTION, which made mu=0 the stickiest setting available instead of the frictionless one; and the slip was measured in world space rather than against the probe, which damps a particle against the room and is a different physical claim. The tell was the ladder running monotonic in the WRONG DIRECTION, which no single-arm test could have shown. T2 passes on an EXACT equality (unset == mu 0 to the byte), proving UNSET skips the law entirely rather than defaulting to something plausible, and T6 uses a middling coefficient to prove the cone is a continuum rather than a stick/slide switch. Neutrality is measured rather than argued: all six solver gates re-run against the modified solver at 54/54 GREEN. Bar to CLOSE this rung: one cited macroscopic coefficient admitted to the oracle conf with its source, at which point the mechanism arms itself with no code change.
Matrix cell	tangential-friction-law · `runtime/nx_softtissue.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_softtissue`; declare `nx_softtissue TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

SB10 Self-contact between two tissue surfaces · cell PRESENT · gate NO-GATE-DECLARED

Contract	`st_self_contact` · executor Organ · 6 u · stands on SB3
Definition of done	DONE 2026-08-25, nx_selfcontact_gate GREEN 8/8 ON THE SERVED BINARY: built, promoted under expect_sha256, registered, invoked via /api/gate_run and rostered, with the NAS numbers identical to the local ones. Closest approach 819 to 997 cmm against a 1000 cmm target, a six-pair contact patch, both bodies corrected by inverse-mass split, broad phase proven to cost zero pair tests for separated bodies. THREE RED RUNS BUILT THIS GATE AND ALL THREE WERE ONE MISTAKE: asking a question about SURFACES with an instrument that answers about BOXES or AXES -- placement by bounding-box overlap left the surfaces 397 cmm apart, then two successive did-it-move teeth read zero because the cages meet on their flanks and the contact normal is mostly y and z. Every tooth is now axis-free and the displacement instrument carries its own planted-shift control, so a zero from it can never again be read as a correction that did not happen. The contact distance is DERIVED from st_surface_spacing rather than chosen.
Matrix cell	self-contact-between-surfaces · `runtime/nx_softtissue.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_softtissue`; declare `nx_softtissue TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

SB4 Cauchy stress and strain tensors with Young's modulus and Poisson ratio · cell WATCHING · gate NO-GATE-DECLARED

Contract	`st_cauchy_stress` · executor Organ · 10 u · stands on SB1
Definition of done	The solver exposes stress and strain per element and accepts Young's modulus and Poisson ratio as material inputs, reproducing the existing per-layer compliance behaviour byte-identically where the two overlap. This is the row every clinical claim is blocked on.
Matrix cell	stress-strain-tensors · `runtime/nx_softtissue.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_softtissue`; declare `nx_softtissue TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

SB5 FEBio as an analytical oracle, not a dependency · cell WATCHING · gate NO-GATE-DECLARED

Contract	`fo_febio_compare` · executor Organ · 8 u · stands on SB4
Definition of done	A headless harness loads a cube and a cylinder of known dimensions through OUR solver and through a published FEBio or closed-form result and reports the deviation as a number with a pre-declared tolerance; exceeding it fails the regression.
Matrix cell	febio-analytical-oracle · `runtime/nx_softtissue.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_softtissue`; declare `nx_softtissue TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

SB6 Surface deviation against the 4D-scan parity bar · cell NO-MATRIX-ROW · gate NO-MATRIX-ROW

Contract	`st_surface_err_mm` · executor Organ · 6 u · stands on SB4
Definition of done	Measures mean euclidean surface deviation in millimetres and grades against knowledge/gamefeel_oracle.conf tissue_surface_err_mm 0..7 (chen2024 multicomponent FE versus 4D scan, 7.05 mm). The bar is already banked and nothing measures it.
Matrix cell	NO-MATRIX-ROW — no matrix row carries `st_surface_err_mm`; the ranker reads this rung as unmapped and no cell can flip for it. Add the row and the join completes on the next publish.
Gate	the gate join needs a matrix row: no organ is named for this rung
Work log	no log row names this rung yet

SB7 Tetrahedralisation of an arbitrary watertight surface with a sliver oracle · cell NO-MATRIX-ROW · gate NO-MATRIX-ROW

Contract	`st_tet_from_surface` · executor Organ · 7 u · stands on nothing
Definition of done	Converts an arbitrary watertight surface to a tetrahedral cage and REFUSES any mesh containing zero-volume or degenerate elements before it reaches the solver, naming the offending element. This is what fTetWild and Gmsh do for the rival pipelines and what our own asset lane needs.
Matrix cell	NO-MATRIX-ROW — no matrix row carries `st_tet_from_surface`; the ranker reads this rung as unmapped and no cell can flip for it. Add the row and the join completes on the next publish.
Gate	the gate join needs a matrix row: no organ is named for this rung
Work log	no log row names this rung yet

SB8 One animal on the same solver with a cited gait band · cell NO-MATRIX-ROW · gate NO-MATRIX-ROW

Contract	`st_quadruped_gait` · executor Organ · 9 u · stands on SB1
Definition of done	One species with a published stride-frequency band admitted to the oracle conf and a gait-coupled drive proven the way the human one now is, on the SAME solver differing only by profile rows.
Matrix cell	NO-MATRIX-ROW — no matrix row carries `st_quadruped_gait`; the ranker reads this rung as unmapped and no cell can flip for it. Add the row and the join completes on the next publish.
Gate	the gate join needs a matrix row: no organ is named for this rung
Work log	no log row names this rung yet

SB9 Ground that bears load · cell WATCHING · gate NO-GATE-DECLARED

Contract	`we_shear_transport` · executor Organ · 11 u · stands on SB4
Definition of done	Soil and vegetation deflect under a stated pressure, with the deflection cited rather than tuned; footprint and compaction persist. Chrono is BEST here and is the benchmark to grade against.
Matrix cell	environment-soft-body · `runtime/nx_water_erosion.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_water_erosion`; declare `nx_water_erosion TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

SB11 Contact that survives deep interleaving · cell WATCHING · gate NO-GATE-DECLARED

Contract	`st_sdf_contact` · executor Organ · 8 u · stands on SB10
Definition of done	A signed distance field or continuous collision detection resolves a pair that has already passed through each other, with a neg-control proving the proximity model reports that same pair as clear. FILED FROM A MEASUREMENT, not from imagination: SB10's proximity contact was measured blind to the tangled case on the day it shipped, and it fails in the reassuring direction by reporting interleaved bodies as not touching.
Matrix cell	self-contact-deep-interleaving · `runtime/nx_softtissue.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_softtissue`; declare `nx_softtissue TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-10-01 measure TRIANGULATED BAR as of 2026-10-01 for penetration-free contact of deformable bodies: three method families hold the bar and none dominates. (1) The IPC lineage keeps CCD-checked non-penetration, and its SIGGRAPH 2026 form drops the log barrier for a second-order augmented Lagrangian that admits every CCD-detected pair at once, reporting up to 103x over GIPC [barrierfree_al_2026]; its thickness-offset ancestor for cloth and rods is C-IPC [cipc_2021]. (2) The GPU-barrier family keeps the barrier and makes it cheap with analytic eigensystems, 3x over the standard barrier and fully on the GPU [gipc_2024]. (3) The offset-geometry family drops CCD for per-vertex conservative displacement bounds and runs in real time, more than two orders of magnitude over IPC-based simulation [ogc_2025]. The newest listing read has no October entry; its newest on-subject entries are PhyVisGen (2609.25653, 2026-09-22, IPC soft-gripper contact for robot data) and TOGEARI (2608.14162v2, 2026-09-11, preconditioning condensed FE systems with IPC contact) [arxivsearch_ipc_20261001] ; witnesses [barrierfree_al_2026] [gipc_2024] [ogc_2025] [arxivsearch_ipc_20261001] ; where they disagree: on the barrier (GIPC keeps it, the 2026 IPC-lineage paper names it the cause of ill-conditioning and time-of-impact locking, OGC replaces it with a two-stage activation over an offset surface); on CCD (the IPC lineage and GIPC run it every iteration, OGC runs none); on what is guaranteed (the IPC lineage claims intersection-free trajectories for everything it simulates, OGC guarantees codimensional objects only, lets volumetric meshes skip the bound and resolve intersections after they occur, and does not protect body-cloth contact when the body is skeleton-driven); and on speed, where 103x over GIPC and more than 100x over IPC-based simulation are measured against different baselines on different scenes, so no shared benchmark ranks the three ; the estate's running lanes split, read from their own sources on 2026-10-01 while their builders were writing them (corrected the same hour from a first draft that judged both lanes from st_self_contact in nx_softtissue.nx alone): SB15 (st_support_surface in buildroot/runtime/nx_support_surface.nx) is a discrete one-sided point-face XPBD projection with no CCD, no displacement bound and no edge-edge test, and its own header says it has neither the IPC barrier nor the 2026 augmented Lagrangian, so it is BELOW all three; SB13 (st_cloth_contact in buildroot/runtime/nx_cloth_contact.nx) implements OGC's per-vertex conservative bound with truncation for vertex-facet and edge-edge contact and holds the tissue's own motion inside its bound, so on mechanism it is AT the real-time member of the triangle, with declared deviations (hard XPBD projection in place of OGC's activation energy and VBD, no parallel-edge filter, feature de-duplication in place of the Gauss-map block test) and a gate verdict this seat has not observed; SB11's measured blindness to deep interleaving still stands for the tissue-to-tissue path; and the contact barrier named as SB15's target is the 2020-2024 member of the triangle that its own lineage replaced in 2026, while the bound SB13 already carries is the guarantee built for a local parallel vertex-level step (OGC integrates it with Vertex Block Descent) and the closest structural fit for SB15 too (an assessment, not a measurement). 2026-10-04 measure METHOD READ THIS MONTH for contact that survives deep interleaving: Shortest Path to Boundary for Self-Intersecting Meshes, He Chen, Elie Diaz and Cem Yuksel, ACM TOG 42(4), SIGGRAPH 2023, arXiv 2305.09778 [spb_2023], read 2026-10-03 from nx_pdf_text of the stored 21.2 MB PDF (knowledge/fetched/cmp_softbody_spb2023.pdf, sha256 dbbb164e). The exact shortest VALID path from an interior point to a boundary that may overlap itself: Theorem 1 (the path is the shortest valid line segment), Definition 4 (valid iff an element traversal carries it), Theorem 2 (one candidate per boundary face, its Euclidean closest point, tested in increasing distance), sec. 3.3 (topological ray traversal that branches through edges and vertices within a face tolerance, a visited list, a candidate stack, no premature termination) and sec. 5.4 (infeasible-region culling, identical results). It is the RECOVERY half of the 2026-10-01 triangulated bar: the barrier-free augmented Lagrangian [barrierfree_al_2026], GIPC [gipc_2024] and Offset Geometric Contact [ogc_2025] all keep an intersection-free trajectory from an intersection-free start, and OGC itself says volumetric meshes may resolve intersections after they occur -- a tangle that already exists needs this query. 2026-10-04 lesson THE BLIND CASE ON THIS RUNG'S OWN ROW WAS GRAZING, NOT A TANGLE: two settled LARGE_SOFT cages placed touching and pushed 600 cmm past it left an INDEPENDENT point-in-tetrahedron census at 0 both ways while SB10 still saw 9 pairs -- the surface particles slip between the other cage's staircase cells. A fixture that never reached the condition proves nothing, so nx_sdf_contact_gate pushes in half-spacing steps until that census first reads non-zero (2994 cmm in practice, two mesh spacings past the crossing), and there SB10 leaves the pair interleaved after all 24 of its passes (census 122 before, 124 after). A ONE-SHOT PUSH IS AN INVERSION GENERATOR: pushing a 1.3 mm penetration in one step plus a dx/dt velocity injection inverted a tetrahedron of the free cage within one round; the fix is a trust region (no particle moves more than half the smallest rest altitude of its own tetrahedra in one pass, the per-vertex bound of Offset Geometric Contact read as an inversion bound) and an inelastic normal impulse that removes only approaching relative velocity, so penetration depth never becomes kinetic energy. THE SOLVER'S SURFACE FLAG IS NOT THE MESH BOUNDARY: it omits the pinned chest-wall layer (352 boundary vertices of one LARGE_SOFT cage carry srf 0), and a contact keyed on it left wall particles inside the other cage; the subject is every vertex of a boundary face, read from topology. 2026-10-04 measure THE LOADED CASE, MEASURED AND NOT CLAIMED: two LARGE_SOFT domes each pinned to its own wall, pressed tip to tip 2994 cmm past touching (122 particles interleaved). With one capped contact pass per full 8-iteration step the deepest penetration a round met went 2518, 1938 (round 8), 950 (round 32) and then PLATEAUED near 965 cmm for 224 more rounds; forcing it clear with nine extra passes cleared the census but inverted a tetrahedron. With SMALL STEPS (the same budget as 8 substeps of one iteration, a contact pass after each) it went 2477, 822 (round 2), 329 (round 4) and held 312 to 402 cmm through round 64 with no inversion, 20 particles still inside. Reading: under a sustained load the tissue recoils about 300 cmm every one-iteration substep, more than one inversion-safe pass can move, so a POST-STEP projection always leaves a residual -- the source's Algorithm 5 projects collisions INSIDE the solver iterations, which needs a hook in st_substep (queued). The done-rule fixture is therefore the free pair, where nothing forces the bodies back together. 2026-10-04 queue WHAT OF THE SOURCE IS NOT BUILT, named so the routine is never passed off as the whole method [spb_2023]: (1) the in-solver projection of Algorithm 5 -- the contact runs AFTER each st_substep, so a sustained load leaves the measured residual; the remedy is a contact hook inside st_substep's iteration loop (nx_softtissue.nx is shared with AT26, so it lands through that lib's owner, never as a second solver); (2) the edge and vertex infeasible-region culls (only the face-interior cull is built; the traversal still decides those candidates, so results are identical and only cost differs); (3) inverted elements and the backward ray of sec. 3.5 -- a traversal into an inverted tetrahedron is counted and that branch dies; (4) the CONTACT response for a cage folding into ITSELF needs the incident-element exclusion -- the query already answers a self-overlapping mesh (the gate's folded C); (5) the BVH candidate order -- every boundary face's closest point is computed per query, exact but O(faces); (6) the CCD-first hybrid the paper recommends, which here is SB10's proximity contact for the shallow case. Next step: (1), measured on the loaded fixture, accepting when the census reads zero after the last substep of a pressed pair.

SB12 Strand hair that collides with the body · cell WATCHING · gate NO-GATE-DECLARED

Contract	`st_strand_contact` · executor Organ · 6 u · stands on SB10
Definition of done	SCOPED DELIBERATELY NARROW BECAUSE HAIR IS OWNED ELSEWHERE: strand dynamics already shipped as graphics nx_hairdyn_lib hd_ringdown (GREEN 11 of 11, a chain on THIS board's solver), geometry is that board's gpe_hair_geometry, and the groom is procgen's gr_strand. This rung is ONLY the contact between a strand chain and the deforming tet cage, which no board owns and only a soft-body engine can. Done when a strand chain rooted on the cage resolves against it at the mesh-derived contact distance st_surface_spacing already computes, with the head turning and the hair following rather than passing through, AND a neg-control proving a strand initialised INSIDE the body is expelled rather than trapped -- because a contact routine that quietly swallows a strand looks identical to one that never fired. The field bar is strand-based and real-time [uegroom] and the research bar is the augmented mass-spring model [hairams], which names hair-body interaction as prior work separate from its own hair-hair contribution.
Matrix cell	hair-strand-to-body-collision · `runtime/nx_softtissue.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_softtissue`; declare `nx_softtissue TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

SB13 Cloth against the body, against itself, and tearing · cell LANDED · gate NO-GATE-DECLARED

Contract	`st_cloth_contact` · executor Organ · 9 u · stands on SB10
Definition of done	THE HALF NOBODY OWNS. procgen owns the cloth SOLVER (gm_xpbd_step, PG29) and its row already says that solver must ride THE TET CAGE'S CONTACT; dcc owns pattern sewing and drape (gd_drape, DC10); charsim owns one solver for cloth and hair (dyna_cloth, R8). This rung is the contact those three depend on, plus the two things none of them names: garment-on-garment LAYERING and TEARING. Done when two garment layers over a pressed cage show no interpenetration at a derived contact distance and a tear propagates along a stated stress threshold rather than at a chosen edge, with a neg-control proving a below-threshold load leaves the cloth intact. Unreal's Chaos Cloth takes its collision from the character Physics Asset [chaoscloth], which is the same dependency shape; the pattern-first realism bar is Marvelous Designer [marvelous] and CLO [clo3d]. NOTE THE HONEST ORDERING RISK: tearing needs a stress measure, so if SB4 has not landed the threshold must be stated in solver units and labelled as such rather than dressed up as a pascal.
Matrix cell	cloth-to-body-and-multilayer-garment-contact · `runtime/nx_cloth_contact.nx` · LANDED decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_cloth_contact`; declare `nx_cloth_contact TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-10-01 measure TRIANGULATED BAR as of 2026-10-01 for clothing on the body: physics with a guarantee is C-IPC, whose distance offsets strictly keep cloth mid-surfaces at least the applied thickness apart under a C2 strain-limiting barrier [cipc_2021]; real-time physics is OGC, which runs a C-IPC frame reported at 24 s in 0.24 s on CPU and 9.6 ms on GPU with cloth-cloth non-penetration guaranteed [ogc_2025]; learned garments repair rather than prevent intersection, ContourCraft recovering multi-layer outfits with an intersection contour loss and repulsion [contourcraft_2024] and HyperBones driving garments by virtual bones at about 1 ms per frame (2605.20460v4, 2026-09-04, read in [arxivsearch_clothsim_20261001]), and the newest on-subject September entry, found outside that phrase listing, is DiT-Garment, a UV-space diffusion transformer that generates garment deformation over a body in arbitrary motion conditioned on body motion and physical parameters and trained only on synthetic simulations (2609.18510, 2026-09-16) [ditgarment_2026]; measured material is Fabric-101, over 101 fabrics with elastic, friction-recoverable and plastic deformation separated by cyclic loading and fitted by an adjoint elasto-plastic friction simulator (2609.13707, 2026-09-12, accepted to SIGGRAPH Asia 2026) [fabric101_2026]; the listing has no October 2026 entry ; witnesses [cipc_2021] [ogc_2025] [contourcraft_2024] [fabric101_2026] [ditgarment_2026] [arxivsearch_clothsim_20261001] ; where they disagree: guarantee against speed, since C-IPC guarantees thickness and is slow, OGC is real time but states that body-cloth penetration might occur when the body is skeleton-driven, and the learned methods guarantee nothing and resolve intersections afterwards, the generative one returning a sampled distribution of outcomes rather than one determined state; and material, since the contact papers read here state no fabric plasticity or internal friction hysteresis while Fabric-101 measures both in real fabrics and the persistent wrinkles it reproduces are what an elastic-only model cannot keep; tearing is addressed by no witness read here, so it has no triangulated bar yet ; the estate's running lane SB13 (st_cloth_contact in buildroot/runtime/nx_cloth_contact.nx, read 2026-10-01 mid-build; corrected the same hour from a first draft that looked only in nx_softtissue.nx and wrongly called the lane unbounded and its fabric unmeasured) implements OGC's per-vertex conservative bound with truncation around gm_xpbd_step for vertex-facet and edge-edge contact, holds the tissue's own motion inside its bound so neither side can pass through the other, derives the contact distance from half the measured Fabric-101 thickness and splits layer pushes by measured areal density (knowledge/fabric101.conf, seven fabrics from Tables 11 to 13), leaves friction unset until a cloth-on-skin coefficient is cited, and tears at a strain threshold stated in solver units; on contact it is AT the real-time best (OGC) with declared deviations (hard projection in place of OGC's activation energy and VBD, no parallel-edge filter, feature de-duplication in place of the Gauss-map block test), and bounding a SIMULATED body is the case OGC states it leaves open, which becomes an exceed only once its gate (buildroot/runtime/nx_cloth_contact_gate.nx, 28,535 B, verdict not observed by this seat) proves zero cloth-body and cloth-cloth intersections on a moving cage; on material it is BELOW, because only thickness and areal density are measured, not stretch, bending or the friction-recoverable and plastic hysteresis Fabric-101 fits, and Fabric-101's published dataset is the liar-killer for those constants once the solver can take them. 2026-10-01 land RESEARCH LANDED, every source fetched by nx_research_fetch and READ from nx_pdf_text output, never a summariser: Offset Geometric Contact, Chen Hsu Liu Macklin Yang Yuksel, SIGGRAPH 2025 (knowledge/fetched/cmp_softbody_ogc.pdf sha256 6abfd9e1cda6d541...) is the method built -- an orthogonal vertex-facet and edge-edge contact set and, in place of CCD, the per-vertex conservative bound of its section 3.7 b_v = gamma_p min(dmin_v, dEmin_v, dTmin_v) with 0 < gamma_p < 0.5 [ogc]; Fabric-101, Gong et al. 2026, arXiv 2609.13707 (cmp_softbody_fabric101.pdf sha256 4358dab11ca3cc05...) supplies MEASURED thickness and areal density, admitted as data rows in knowledge/fabric101.conf (KN01 cotton jersey 0.7 mm 174.34 gsm, C02 cotton poplin 0.2 mm 108.07 gsm, five more) [fabric101]; the newest penetration-free line, barrier-free augmented Lagrangian elastodynamics over CCD pairs, Zheng Luo Li, SIGGRAPH 2026, arXiv 2512.12151 (cmp_softbody_barrierfree.html sha256 f08bd629e8be61fa...), is named and NOT built [barrierfree]. Three refs rows on softbody.refs, cited inline on the matrix row. 2026-10-01 queue GARMENT TENSION INDENTING SOFT TISSUE -- the string pressing into the hip in the operator's reference photographs -- is NOT claimed and is carried by SB13 as its open second leg, with SB15 (a support surface that spreads load) as the board's other load-into-tissue rung and SB4 (stress) for a pressure measure. What exists: every push in st_cloth_contact is already split by inverse mass, and the tissue surface is already held inside its own conservative bound so neither side passes through the other. What is missing, exactly: (1) the tissue side of the push is held at inverse mass 0 because no garment-to-tissue mass ratio is cited -- Fabric-101 supplies the garment's areal density, a mirrored soft-tissue density row with its source is the missing half; (2) the cloth and tissue constraints run STAGGERED (st_run, then the cloth rounds), and a taut band can only indent tissue when both are solved in one iteration loop. Next step: admit the tissue density row, derive the cage particles' contact inverse mass from it, move the tissue iterations inside the bound rounds, and gate a band whose depth into the tissue rises with its tension, with a slack-band neg-control reading zero depth. 2026-10-01 queue WHAT OF OGC IS NOT BUILT, named so the hard-projection organ is never passed off as the whole method [ogc]: (1) the two-stage activation ENERGY (quadratic to a stitch at r/2, then logarithmic) and its Vertex Block Descent solve -- contact here is the hard-projection limit inside XPBD, so no contact stiffness is calibrated and no contact force is an energy gradient; (2) the bound is recomputed per ROUND (two per tick), where the paper recomputes it whenever a vertex reaches it inside the solver iterations; (3) the per-vertex polyhedral Gauss map block test is replaced by de-duplicating the contacted face, edge or vertex feature; (4) the smooth parallel edge-edge filter -- parallel pairs are skipped; (5) GPU parallelism. Next step: an energy-form OGC contact on the VBD lane, gated with THIS organ's penetration counts as its oracle. 2026-10-01 queue THE SEPTEMBER FRONTIER FOR T-FR-SOFT IS NAMED AND NOT BUILT: barrier-free augmented Lagrangian elastodynamics over every CCD-found pair (2512.12151, SIGGRAPH 2026) [barrierfree] and Fabric-101's elasto-plastic-with-friction cloth that forms persistent wrinkles (2609.13707) [fabric101]. Both need an integer continuous-collision kernel (vertex-triangle and edge-edge time of impact) this estate does not have; OGC's bound was chosen tonight precisely because it needs none. No benchmark shared with any source was run: the numbers on this rung are the estate's own fixture. Next step: the CCD kernel as a shared primitive, then mirror one published scene (OGC's cloth on a rotating sphere) as a fixture and publish the penetration count and per-tick cost beside the paper's. 2026-10-01 land st_cloth_contact LANDED in runtime/nx_cloth_contact.nx (LIB, sha256 166969fe51ad2f90, 79687 B), an organ of its own that IMPORTS nx_softtissue.nx and the PG29 cloth solver nx_gm_cloth_lib.nx and writes no second solver: every cloth constraint is still gm_xpbd_step, every tissue constraint still st_substep. Built to Offset Geometric Contact [ogc]: orthogonal vertex-facet and edge-edge contact against offset geometry with one push per contacted face, edge or vertex feature, and the section-3.7 conservative per-vertex bound applied to BOTH sides (cloth and the tissue surface) with gamma_p one half less the integer rounding, recomputed every round over a dense grid sized from the scene. Contact distance and the layer-to-layer mass split are DATA from Fabric-101 [fabric101] (knowledge/fabric101.conf sha256 3ff888586c377840). Tearing: a constraint fails when its strain passes a stated threshold IN SOLVER UNITS (SB4 has not landed), the row leaves the incumbent solver's list and its contact triangles become a hole. Friction: the per-256 positional Coulomb law, abstaining at ST_MU_UNSET. Gate nx_cloth_contact_gate (source sha256 4029946564cb1e25) PROMOTED sha256 f3d4645bb272c41d56e15607916d8311390cea86d582b25cfaec6f3b4da222ba (191569 B), registered, _offc installed. Matrix row cloth-to-body-and-multilayer-garment-contact REPOINTED from runtime/nx_softtissue.nx to runtime/nx_cloth_contact.nx (the shared tissue file was being extended by AT26 the same hour); its watch flips on the next regen, not by edit. 2026-10-01 measure nx_cloth_contact_gate GREEN 15/15 on the build that was promoted (f3d4645b): ST_PROF_LARGE_SOFT pressed under SUPINE gravity for 120 substeps (apex 5900 to 5340 cmm), inner layer Fabric-101 KN01 cotton jersey held at both sides and lowered onto it, outer C02 cotton poplin dropped loose, 48 ticks at 60 Hz, 2 bound rounds x 8 iterations. ZERO garment vertices inside the cage on ANY tick (point-in-tet over all 4032 tets); brute-force closest approach inner-to-body 34 cmm against a contact distance of 35 (half of 0.7 mm), inner-to-outer 42 against 45 (0.35 + 0.10 mm), tolerance 4 cmm DERIVED from integer rounding; vertical-ray layer order 0 violations of 50; contacts 8169 body, 58267 layer, 32170 edge-edge; 12246 cloth and 20 tissue truncations; smallest bound 13 cmm; tissue moved at most 30 cmm per tick. NEG-CONTROLS, all BITE-PROVEN: body contact off puts 113 garment vertices inside the cage; the second layer ignoring the first gives separation 0 and 52 order violations; the OGC bound off lets a layer launched at twice the gap TUNNEL (95 violations) where with the bound it does not (0 of 94). TEARING: strain under its own weight 291 permil, under 4 g 632, stated threshold 461 permil (solver units, measured midway): 1 g leaves the sheet INTACT (0 failures at 291), 4 g fails 16 edges over 11 ticks, the first at the most-strained edge, with 12 later failures touching earlier ones -- the tear propagates from where the criterion put it. First version read 10/15: one projection pass per round left a stacked contact (body, inner, outer) Gauss-Seidel-unconverged at 26 and 33 cmm; iterating the contact set like any constraint set closed it. 2026-10-01 lesson A CONTACT SET IS A CONSTRAINT SET: a stack of body, inner layer and outer layer projected in one pass leaves the last-processed pair satisfied and the first violated (measured 26 of 35 and 33 of 45 cmm), so contact needs the same iteration budget as the constraints. And the conservative bound is what separates a discrete projection from a penetration-free one: the same layer launched at twice the gap tunnels with the bound off and cannot with it on -- a gate that only drops cloth slowly never sees the difference. 2026-10-01 measure THE PROMOTED BINARY RE-RUN FROM THE LIVE PATH (nx_job_run nx_cloth_contact_gate, nishihost/nx_cloth_contact_gate.elf sha256 f3d4645b): GREEN 15/15 with every emitted value identical to the staged run (output 4261 B both times, inside_max_no_body 113 both times) -- the integer solver is deterministic. Scratch registry row nx_cloth_contact_gate_sb13_staged REAPED and its staged file retired reversibly to knowledge/retired/; a cap naming nx_cloth_contact_gate minted (private response) so the gate is callable, not REGISTERED-UNAUTHORISED. Fleet referee nx_compare_refs_gate names none of the three new refs rows (ogc, fabric101, barrierfree) in any finding; its 40/45 RED is the pre-existing redirect and moved ratchets. 2026-10-01 queue SHIPPED, NOT COMPLETE: gate GREEN 15/15 is the mechanical receipt only. Completion needs the referee panel and the operator's ACCEPT row; no render of the garment over the body was produced or judged tonight, so how the cloth LOOKS sitting on the body is UNMEASURED. Next step: render the gate's final drape (KN01 band over the pressed cage, C02 over it) through the beach-twin render lane and put it before the referee panel; persistent wrinkles (Fabric-101's elasto-plastic-friction model) are the visual half the frontier target T-FR-SOFT still asks for. 2026-10-01 land THE TISSUE HALF OF THE MASS RATIO IS NOW DATA, so the two-way leg is no longer blocked on a citation: NIST's ICRP entries read from the stored bytes -- ADIPOSE TISSUE (ICRP) Density (g/cm3) = 9.20000E-01 [nist_adipose] and TISSUE, SOFT (ICRP) 1.00000E+00 [nist_softtissue] -- admitted as knowledge/tissue_density.conf rows (mg per cm3). Derived, not typed: an adipose lattice particle of the 10 mm cage carries 0.920 g against a KN01 cotton jersey vertex at 15 mm spacing of 0.0392 g (174.34 gsm times 225 mm2), a garment-to-tissue mass ratio of 1 to 23.5. NOT wired tonight: a staggered push at that ratio would move the tissue by an iteration-dependent sliver per round, which is a mechanism, not the monolithic cloth-plus-tissue solve the indentation needs; shipping it as garment tension indenting skin would be a lesser method wearing the claim. 2026-10-01 queue FRICTION OF A GARMENT ON SKIN AND ON ITSELF IS WIRED AND ABSTAINING: st_cloth_contact carries the estate's per-256 positional Coulomb law on every vertex-facet push, but no cloth-on-skin or cloth-on-cloth coefficient is cited (Fabric-101's friction is internal yarn hysteresis, not a contact coefficient), so the scene defaults to ST_MU_UNSET and the friction path has NEVER RUN on a real coefficient. Next step: mirror a measured textile-on-skin and textile-on-textile friction study, admit its coefficients beside knowledge/fabric101.conf, and add a tooth where the outer layer on a tilted inner layer stays put above the cited coefficient's angle and slides below it, with mu UNSET as the sliding neg-control. 2026-10-01 measure INDEPENDENT RUNNER AGREES: /api/gate_run on the promoted nx_cloth_contact_gate returned exit_code=0 verdict=GREEN (receipt _jobs/job_1790826259.out, drained inline after outrunning the 14750 ms reply window); thirteenth evidence row appended to knowledge/compare/softbody.gates. nx_ladder softbody still LADDERED (20 rungs placed, SB13 contender for T-FR-SOFT). 2026-10-01 queue CLOTH AGAINST ITSELF IS WIRED BUT NOT SEPARATELY GATED: same-layer pairs that share no vertex already carry both the conservative bound and the push (the rung's self-contact half), but tonight's fixtures never fold a layer onto itself, so a self-separation tooth would be vacuous and none was claimed. Next step: let ctc_scene_new take initial positions so a sheet can be built PRE-FOLDED with its rest state measured folded, drape it over the pressed cage, and gate its self-separation against its own thickness with CT_FL_NO_SELF as the neg-control that must interpenetrate. 2026-10-01 land CLOTH AGAINST ITSELF IS NOW GATED, superseding the 1790826581 queue row: st_cloth_rest_from_current (a garment folded, gathered or sewn takes that shape as its rest state) added to runtime/nx_cloth_contact.nx (sha256 a655441bddfb2351, 80329 B); nx_cloth_contact_gate grown to a FOLDED KN01 layer draped over the pressed cage (source sha256 85b3057a394f8493, 32681 B) and RE-PROMOTED sha256 ceaaa23dbc4f2c887f296058a06e0c3c693891d10b9ae2728649ae3eba255e20 (223272 B, debug flavour as the builder chose from the incumbent), after nx_contentdiff live-vs-candidate GREEN lost_from_live=0 ONE-WAY-GAIN; _offc reinstalled; scratch rows nx_cloth_contact_gate_sb13b_staged reaped and its file retired reversibly. 2026-10-01 measure nx_cloth_contact_gate GREEN 18/18 on the re-promoted build (run /tmp/nx_cloth_contact_gate_sb13b_1.out, every earlier value reproduced exactly): a KN01 jersey layer folded in half with its top half two self-contact distances (140 cmm) above the bottom, draped 48 ticks over the cage pressed under supine gravity -- 28036 self contacts and 5684 body contacts fired, ZERO garment vertices inside the cage, brute-force self-separation of the folded halves 66 cmm against the fabric's own measured thickness of 70 (tolerance 4, so at the edge of the derived bound and named as such); NEG-CONTROL BITE-PROVEN: with self-contact off the halves pass through each other to a self-separation of 0. 2026-10-01 measure INDEPENDENT RUNNER ON THE RE-PROMOTED BUILD: /api/gate_run nx_cloth_contact_gate (ceaaa23d) was promoted to the job lane and reaped state=DONE rc=0 verdict=GREEN (_jobs/job_1790827948.claim, 5030 B output, the same size as the staged run), source_newer=0; the evidence row in knowledge/compare/softbody.gates now reads 18 of 18 with both receipts. Every scratch registry row from this lane is reaped and both scratch stage files are retired reversibly to knowledge/retired/. 2026-10-01 measure BACKFILL-HUMAN (e) measured fabric material and tearing: the critic's abstract-only note on Fabric-101 is STALE, this board's row fabric101 was read in full through nx_pdf_text on 2026-09-30 (knowledge/fetched/cmp_softbody_fabric101.pdf). A second independent measured-fabric group now: SEDDI with Universidad Rey Juan Carlos and Universidad Carlos III, Madrid (Rodriguez-Pardo 2023 Eurographics, refs sb13_rodriguezpardo_seddi_2023, mirror knowledge/fetched/cmp_softbody_bfh_drapelike_2304_06704.pdf sha256 1a118500, full text read through nx_pdf_text: real fabrics with ground-truth stretch, bending and density parameters, a drape metric with 86.68 percent user agreement). Measured-fabric independent count 2, SHORT of 3. Tearing, was 0 witnesses, now 1 weak one: a tear-test PROCESS study (Huang 2024 PLoS One, Zhejiang standards and inspection institutes, refs sb13_huang_zhejiang_2024_teartest, sha256 12682c12) that measures the test, not a fracture energy or a tear-propagation law; an arXiv listing of tearing AND cloth or fabric or textile or garment, 30 newest (knowledge/fetched/cmp_softbody_bfh_arxiv_tearing_listing.xml), held 0 on-subject cloth-tearing papers, every hit a homonym (graphene tearing, fabrication). Not obtained: Pfaff, Narain, de Joya and O'Brien 2014 Adaptive Tearing and Cracking of Thin Sheets (graphics.berkeley.edu): plain http refused status 0, https failed in the sovereign TLS cert pipeline rc -207; unblock is cert-chain support for that host or a library copy

SB14 A liquid film that clings · cell WATCHING · gate NO-GATE-DECLARED

Contract	`fl_film_cling` · executor Organ · 10 u · stands on SB3
Definition of done	CORRECTED FROM AN ABSENCE BEFORE IT WAS PUBLISHED: nx_fluid.nx already EXISTS and is real -- a height-field shallow-water solver whose reflecting boundaries make volume exactly conserved, with fl_volume exposing it, and whose own header declares its limit (FLIP is the particle variant and this is not it). Two capability probes had first returned pure homonym noise: viscosity matches 28 times in runtime and every classified hit is food-science rheology or FDM melt flow, and Navier matches 16 times and every one is the Octane JavaScript benchmark. So the gap is precise rather than total: no viscosity term, no surface tension, no wetting, and no film adhering to a moving body. Done when a film on the tissue surface has its thickness driven by a CITED surface-tension and viscosity pair admitted to the oracle conf, with a neg-control proving a zero-tension film sheets off completely instead of beading. LiquiGen is the real-time bar [liquigen]; Houdini FLIP is the offline bar and is uncited on purpose because our fetcher cannot reach sidefx.com. THE CITATION IS THE RISK HERE, exactly as it was for skin friction on SB3: if no readable primary source gives the pair, the mechanism ships with the constants UNSET and a tooth proving unset behaves as inviscid.
Matrix cell	viscous-film-surface-tension-and-clinging · `runtime/nx_fluid.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_fluid`; declare `nx_fluid TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

SB15 A support surface that spreads load · cell LANDED · gate NO-GATE-DECLARED

Contract	`st_support_surface` · executor Organ · 5 u · stands on SB2
Definition of done	DISTINCT FROM SB9 AND THE ROW SAYS SO: SB9 is OUTDOOR terrain -- soil shear transport, footprints, compaction -- where Chrono terramechanics is BEST. This is the INDOOR support surface, a mattress or cushion or bedding whose deflection redistributes body weight, and its deliverable is not a dent but a PRESSURE MAP, because how load spreads across the contact patch is the whole question. THE CHEAPEST NEW RUNG ON THE BOARD BECAUSE IT COMPOSES RATHER THAN BUILDS: st_contact_force already reports a per-contact reaction and st_self_contact already resolves two bodies with an inverse-mass split, so a support surface is a second cage carrying its own compliance rows. Done when a body at rest on a compliant surface has its summed contact reaction equal its applied load within a stated tolerance, with indentation depth AND contact-patch area both reported, and a neg-control proving a rigid surface concentrates the same load into a measurably smaller patch. Units stay solver units: there is no kilogram in this solver and SB2 already refused to invent one.
Matrix cell	bedding-and-support-surface-load-distribution · `runtime/nx_support_surface.nx` · LANDED decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_support_surface`; declare `nx_support_surface TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-10-01 measure TRIANGULATED BAR as of 2026-10-01 for body-support interface pressure: SOTA is a subject-specific hyperelastic finite-element body (MRI or CT geometry, Ogden tissue) on a modelled support surface (foam, air cells with internal ideal gas, alternating-pressure cycles) that reports interface stress and internal tissue strain [jner_buttocks_2026], validated against interface pressures from calibrated mapping systems to within 3, 14 and 10 percent of peak pressure [katzgefen_2023]; measurement reports average and peak pressure, contact area and the high-risk area above 40 mmHg from mats of about 3.5 cm2 per cell [hybridtextile_2026]; the graphics side simulates bodies at rest on a pressure mat at scale (206K synthetic pressure images) and judges the simulation by transfer to real mats [bodiesatrest_2020]; the newest listing has no October entry and no on-subject September entry, its newest on-subject entry being diffusion-synthesised in-bed pressure maps from depth with a weight-consistency loss (2512.13757v2, 2026-02-12) [arxivsearch_pressuremap_20261001] ; witnesses [katzgefen_2023] [jner_buttocks_2026] [hybridtextile_2026] [bodiesatrest_2020] [arxivsearch_pressuremap_20261001] ; where they disagree: the Tel Aviv group holds that interface pressure is the wrong end-point because sustained internal stress concentration causes injury, while the measurement study and the graphics simulation treat the map itself as the output; the 2026 Chongqing framework states no measured-map validation in its abstract; the Tel Aviv validation compares peak pressures with maps reported by the surface manufacturer or the literature rather than a map of the modelled subject; the Tehran measurement is one participant at 3.5 cm2 resolution that its own authors call too coarse; and Bodies at Rest validates its simulator only indirectly through pose accuracy on real mats, so no witness publishes a full map-to-map error between a simulated and a measured pressure field ; the estate's running lane SB15 (st_support_surface in buildroot/runtime/nx_support_surface.nx, read 2026-10-01 mid-build; corrected the same hour from a first draft that looked only in nx_softtissue.nx) builds a second XPBD tet cage under the body whose per-vertex reaction shares form the map, with a summed-reaction-equals-load check (its header measures 1016 permil of the load carried on a rigid support once velocity is re-derived after contact, against 847 for the post-step order), indentation depth, patch area, a peak-to-mean ratio and a rigid-surface neg-control, publishing pressure as a shape in solver units with no kilogram and no pascal, and it is BELOW the triangulated best because it cannot express pressure in kPa or mmHg until SB4 lands, compares against no measured map, and reports no internal strain; its load balance and patch area are the right first teeth, the measured peak pressure, contact area and area above 40 mmHg are the comparison it must print once it has units, and a full map-to-map error against a measured mat would be a first among these witnesses (an opportunity, not a claim). 2026-10-01 land REPOINTED: matrix row bedding-and-support-surface-load-distribution now names runtime/nx_support_surface.nx (was runtime/nx_softtissue.nx), contract symbol unchanged st_support_surface. The shared solver gets no line from this rung; the new file imports it and composes st_new, st_add_edge, st_substep, st_eff_damp, st_surface_spacing. Sources pinned on softbody.refs: the seated-buttock rigid-versus-foam FE study [sb15_buttockfat], the IPC barrier bar [sb15_ipc], the SIGGRAPH 2026 barrier-free penetration-free frontier [sb15_barrierfree]. 2026-10-01 lesson AN EDGE-ONLY LATTICE CANNOT CARRY A BODY: the cited cushion is a v=0 foam, so the first cut carried it as edges without the volume constraint, and under the glute profile it collapsed 86 mm through its own 48 mm pad with inverted tets (min 6V -3565297404; /tmp/nx_support_surface_gate_sb15_1.out, build a04b47f5) -- nothing resists inversion once the incumbent's hard volume constraint is withheld. The support moved to the incompressible gel lattice (the same lattice type as the body). Same run, the ORDER law measured: on a rigid support the post-step order carried 847 per mille of the load and the re-derived order 1016, against 1/(1+k) = 833 predicted by the skin damping row 819/4096. 2026-10-01 measure nx_support_surface_gate 14/14 GREEN on build 26dd8c1c (job /tmp/nx_support_surface_gate_sb15_3.out): a glute body of 966 free particles at rest on a 48 mm gel pad carrying the solver row nearest the cited cushion-over-fat stiffness (5000 per mille against 4385 cited [sb15_buttockfat]). Summed contact reaction 1013 per mille of the applied load on the gel and 1017 on a rigid board, derived tolerance 29 (half a cmm against the 17.03 cmm gravity displacement per substep). The residual is ATTRIBUTED, not tolerated: the incumbent st_substep leaks 15 and 19 per mille of net internal momentum through its volume-constraint rounding, and reaction minus leak reads 998 on both supports, exactly the 17 over 17.03 the integer gravity step predicts. Gel indentation 650 cmm, closest approach 611 cmm = dhat on both supports by an all-face scan, no gel tet inverted, two runs bit-identical. The board concentrates the same load onto 83 body contacts against 91 on the gel and a body-side peak of 484 against 477 cmm per substep. Same comparison published [sb15_buttockfat]: contact area soft over rigid 1430 to 1887 per mille, peak rigid over soft 3466 to 7450; ours 1096 and 1014. 2026-10-01 land st_support_surface SHIPPED in runtime/nx_support_surface.nx (lib, 43471 B, imports nx_softtissue.nx and adds no line to it): a body-vertex against support-face point-face contact with the barycentric inverse-mass split, recorded as a support-side pressure map (each face vertex share, what a sensor mat reads) and an alias-free body-side map, the summed reaction against the solver's own gravity increment, indentation, patch, an all-face interpenetration oracle, an inversion census and the solver leak. Velocity is re-derived after contact (constraints, then velocity, then damping) so the damper holds none of a resting body's weight. Gate nx_support_surface_gate PROMOTED sha 26dd8c1c 121674 B, registered, _offc twin installed, scratch row reaped and its binary retired reversibly. In-gate controls bite: the post-step order carries 848 per mille (fires), a run without the contact pass reads -11551 cmm penetration, refusals move nothing. 2026-10-01 queue BARRIER CONTACT, NOT BUILT TONIGHT, AND EXACTLY WHAT IT ADDS. The bar is IPC [sb15_ipc]: a C2 log barrier on (0, dhat) inside a projected-Newton implicit step with a CCD-truncated line search; the September 2026 frontier keeps the CCD and the second-order solve and swaps the barrier for an augmented Lagrangian over every CCD-detected pair [sb15_barrierfree]. st_support_surface has neither: a hard ONE-SIDED projection (body vertex against support face) after an explicit XPBD substep. Three tests expose the gap and are the next done-rule: (1) a RIDGE (the lounger frame edge) pressed into the body, where support vertices slip between body vertices because nothing tests support vertex against body face or edge against edge, so an all-pairs scan in BOTH directions must read zero; (2) a cushion FOLDED over its own edge, which the plan-projection locate (a height field over the rest grid) cannot represent; (3) RIM FLICKER, rim contacts that change state between consecutive windows, which a C2 barrier force drives toward zero and a hard projection does not. Next step: derive the body cage surface triangulation from its Kuhn boundary faces, add vertex-face both ways plus edge-edge CCD, and solve a barrier energy per contact island. 2026-10-01 queue THE PUBLISHED SHAPE IS NOT YET REACHED, AND THE MISSING PARTS ARE NAMED. Ours reads contact area 1096 and peak 1014 per mille (rigid against compliant) where the source reads 1430 to 1887 and 3466 to 7450 [sb15_buttockfat]. In order: (a) the source peak sits under the ischial tuberosity, a bone pressing through 6 to 30 mm of fat with 300 N of upper body on the sacrum, while our dome is all soft tissue under a free flat base; (b) our support is an incompressible gel where the source foam has v = 0, because the edge-only foam collapsed and st_substep solves volume hard -- a compressible cushion needs a COMPLIANT volume row with an inversion barrier at J to 0 in nx_softtissue.nx through its owner lane; (c) the cited stiffness ratio 4385 lands on the nearest profile row 5000 through an UNCALIBRATED lattice map -- a uniaxial compression tooth must measure each lattice's effective modulus, and a per-cage compliance row must replace profile rows (st_eff_alpha_dist reads profile rows only). Done-rule for the next pass: a bony prominence inside the glute cage on a calibrated foam, the two ratios inside the published band. 2026-10-01 queue THE SHARED SOLVER LEAKS MOMENTUM AT REST, MEASURED BY THIS RUNG: st_substep's volume correction rounds each vertex's sub-cmm share separately, so near convergence the shares do not sum to zero and a resting glute body carries 15 to 19 per mille of its load as net internal momentum (sps_grav_disp against sps_disp_z, gate values gel_solver_leak_permil and rigid_solver_leak_permil). Owner lane nx_softtissue.nx: round the tet's summed correction and hand the remainder to the largest share (or carry sub-cmm fixed point) so internal forces sum to zero; then nx_support_surface_gate's reaction reads 998 per mille with no attribution needed, and every other contact readback on this board loses the same bias. 2026-10-01 queue THE FE PATH TO A kPa COMPARISON, NAMED BY FUNCTION. AT26's hyperelastic tier landed in nx_softtissue.nx the same evening (gate nx_softtissue_hyper_gate 28/28) and takes FEBio-convention Ogden rows verbatim (st_hyper_material_ogden), the convention the source states for fat (mu1 1166.7 Pa, alpha 16.2) and muscle (1907.4 Pa, 4.6) [sb15_buttockfat]. Composition: a fat-over-muscle block (st_hyper_block, st_hyper_assign_box), a rigid sphere for the ischial tuberosity (st_hyper_set_implant with contact candidates from st_hyper_contact_plane and the augmented Lagrangian on via st_hyper_set_al, the formulation family of the September 2026 frontier [sb15_barrierfree]), the seat as a Dirichlet plane (rigid) or a neo-Hookean layer at E = 15.3 kPa with K set for v = 0 (foam), quasi-static st_hyper_solve, and the seat-face pressure map read from element internal forces at the seat nodes (st_hyper_element) in pascals. Done-rule: peak contact pressure on the rigid seat inside the source's 0.052 to 0.149 MPa across fat 6 to 30 mm and inside 0.015 to 0.020 MPa on the foam, internal tissue stress read by st_hyper_cauchy beside it (the end-point the 2026-10-01 bar row says injury follows). 2026-10-01 lesson A FACE-SIDE SENSOR COUNT IS ALIASED, AND IT ONCE READ THE GEL WITH A SMALLER PATCH THAN THE BOARD. One body contact lights one support-face vertex when it sits on it and three when it sits mid-face, so the face-side patch read the gel 125 against the board 127 while the body side, a regular lattice where each contact counts once, read 91 against 83 and its peak 477 against 484 (build bb8fa3b9, /tmp/nx_support_surface_gate_sb15_2.out). The patch and peak teeth read the body side; the face-side map stays published as the sensor-mat deliverable. A sensel-integrated face map (each cell's share over its own area) or a participation-ratio effective area would remove the aliasing on the face side too, and is the cheap next refinement. 2026-10-01 land RE-PROMOTED against the shared solver as AT26 left it: AT26 landed the hyperelastic tier into nx_softtissue.nx (78002 to 136681 B) after this rung's first promote, so the gate was rebuilt against the new closure (debug flavour from the incumbent twin, 140032 B, sha 0428c8f0), run 14/14 GREEN (/tmp/nx_support_surface_gate_sb15_4.out) with every value AND the determinism hash -6676300930262145470 bit-identical to the run on the old closure, content ruler lost_from_live=0, and promoted over 26dd8c1c with expect_sha256; _offc twin refreshed, scratch registration reaped, scratch binary retired reversibly. So the shared-file edit is measured behaviour-neutral for st_substep, not assumed. 2026-10-01 measure COMPLIANCE SWEEP, PRE-DECLARED AND PARTLY REFUTED, NOT SHIPPED (/tmp/nx_support_surface_gate_sb15_5.out, build 142256 B, 16/17 RED). A third support carried the body's own fat row, so the three supports formed the sweep the source itself runs [sb15_buttockfat]. Body contacts ARE monotone in compliance: board 83, cited-row gel 91, tissue-matched gel 99 (area ratio 1192 per mille against the board, indentation 650 then 1131 cmm). The body-side peak is NOT: 484, 477, 483 -- the pre-declared tooth demanded it fall and it did not. The tissue-matched run balances exactly (reaction 1000 per mille, solver leak 2) and keeps the gap and every tet, but its summed body speed at the window is 3560093 against 1677468 on the cited gel, so it is not demonstrably at rest. The tooth was NOT loosened to pass: the gate source was reverted to the promoted 14/14 version, and the sweep is queued with the instrument it lacks. 2026-10-01 queue THE SWEEP NEEDS A MEASURED REST, NOT A FIXED SETTLE. Three damping periods settle the cited-row gel but leave the tissue-matched gel moving (summed body speed 2.1x), and a peak read off a moving body is not a resting peak. Next step: settle by MEASUREMENT -- run windows of one damping period until two consecutive windows agree on the reaction within the derived tolerance AND the summed body speed stops falling, under a stated budget, and print the windows used; then re-run the three-support sweep and re-declare the same tooth (contacts rise and the body-side peak falls with compliance). If the peak still does not fall at rest, that is a finding about a bonded incompressible pad, whose compliance comes from lateral flow, and it is published as such rather than tuned away. 2026-10-01 land BUILD FOSSIL RESTORED, DECLARED RATHER THAN LEFT AS A LOADED GUN: the unshipped sweep build (142256 B) had overwritten buildroot/_build/nx_support_surface_gate.sov.elf, which a drift census would read as BEHIND the promoted gate and offer to promote, and it is RED. The gate source was reverted to 15760 B (zero sweep remnants by txtscan), the confirming rebuild was refused by admission six times at load 17 to 20, so the fossil was restored from the promoted bytes (nx_filecopy verified=1; nx_filehash cmp identical 0428c8f0 140032 B). Open confirmation: one rebuild under admission, expected as a build-cache hit on key 64dcb60b, which stored exactly 0428c8f0. 2026-10-01 measure CONFIRMED, the open item of the row above closed: the reverted gate source keys the build cache at 64dcb60b, the exact key that stored 0428c8f0, and the admitted rebuild was a cache hit verified against its sidecar; nx_filehash cmp reads _build and the promoted gate identical (0428c8f0, 140032 B). So the source on disk, the build fossil and the served gate are one artifact, proven rather than declared. 2026-10-01 measure BACKFILL-HUMAN (d) a simulation validated against a measured interface-pressure map: THREE independent groups now counted, each read in full text: Tel Aviv (Katz and Gefen 2023, existing row katzgefen_2023); Lyon LBMC with Beihang (PIPER adult comfort 2023, refs sb15_piper_lyon_beihang_2023, mirror knowledge/fetched/cmp_softbody_bfh_piper_comfort_2023.xml sha256 3d2b93ba): a subject-specific FE model against the same subject's mats at four seat-pan angles, seat-pan contact area, peak and mean pressure off by 5.8, -6.4 and 1.2 percent, the four-region pressure shares off by 5.2, -1.4, -13.9 and -3.1 percent, backrest peak off by 62.1 percent; Hunan University (Yan 2025, refs sb15_yan_hunan_2025, mirror knowledge/fetched/cmp_softbody_bfh_yan2025_seatangle.xml sha256 193a3cdc): four 50th-percentile males on capacitive mats, back peak +20 percent, hip and leg contact area -2 percent, pressure-point counts per 4 kPa bin compared. Independent count 3. The critic's finding STANDS: none of the three publishes a cell-by-cell map-to-map error; the closest are a four-region share (PIPER) and a 4 kPa histogram (Yan), so a map-to-map metric is still an exceed candidate. Examined and excluded: an ejection-seat cushion study (PMC13057320) that measured pressure but left coupled human-cushion FE to future work (mirror knowledge/fetched/cmp_softbody_bfh_ejectseat_2026.xml); search listing knowledge/fetched/cmp_softbody_bfh_epmc_fepressure_search.json 2026-10-01 land BACKFILL-HUMAN refs rows landed this lane in softbody.refs, each mirror pinned by nx_filehash, cited here so the refs referee's inline-citation tooth sees them: [sb15_piper_lyon_beihang_2023] [sb15_yan_hunan_2025] for SB15, and for SB13 [sb13_rodriguezpardo_seddi_2023] [sb13_huang_zhejiang_2024_teartest]

SB16 Engorgement as rest-volume expansion · cell PRESENT · gate NO-GATE-DECLARED

Contract	`st_inflate_rest_volume` · executor Organ · 4 u · stands on SB2
Definition of done	DONE 2026-08-25 -- nx_engorge_gate GREEN 7/7 on the SERVED binary. Rest-state growth (rest lengths x s, rest volumes x s^3, snapshotted so unity returns exactly to pristine) graded against the closed-form s^3 prediction: 1728 predicted, 1667 measured, correctly under because only adipose grew. Unity control byte-exact; zero/negative factors refuse. THE MAGNITUDE AND RATE OF REAL ENGORGEMENT ARE NOT CLAIMED -- no band is mirrored, so the factor stays the caller's. THIS RUNG EXISTS BECAUSE ANOTHER BOARD'S ROW NAMED IT. The physiology is already shipped and is procgen's: nx_arousal_lib a_perf carries drive, lag, perfusion, temperature, heart rate and blood pressure as integer functions shared by simulate and fit so the two cannot drift, with its arousal-on-mesh gate GREEN 17 of 17. That row states its own blocker in one sentence -- a pure normal push is not volume-conserving and THIS board's conservation gate would refuse it. So what is missing is neither physiology nor rendering: it is driving the tet cage's REST VOLUME from a_perf so engorgement becomes tissue expansion the Neo-Hookean volume constraint carries. Done when rest volume is driven by a_perf with nx_softtissue_conserve_gate GREEN at every level of the drive, and a neg-control proving the retired normal-push model FAILS that same gate -- which is what makes the conservation gate a referee here rather than a decoration. No general-purpose engine ships an inflation-driven rest-volume channel, and that is a scope statement rather than a boast.
Matrix cell	engorgement-as-rest-volume-expansion · `runtime/nx_softtissue.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_softtissue`; declare `nx_softtissue TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

SB17 Muscle contraction that moves skin · cell WATCHING · gate NO-GATE-DECLARED

Contract	`my_contract_fiber` · executor Organ · 12 u · stands on SB4
Definition of done	A DIFFERENT MECHANISM FROM THE FACS ROW TWO BOARDS ALREADY CARRY, WHICH IS WHY IT IS NOT A DUPLICATE OF IT. charsim R9 and graphics both watch fa_expr_morph and that contract is BLENDSHAPES -- expression axes blended over a surface. This rung is the inverse: an ACTIVE CONTRACTILE ELEMENT inside the tet solver, so a fibre shortens and the skin above moves BECAUSE the flesh under it did, and micro-expression emerges from anatomy instead of being authored as a target. Ziva VFX is the production bar and describes exactly this -- muscles, fat, fascia and skin simulated so that skin deformation is informed by the underlying anatomy [ziva] -- with MetaHuman the delivery bar [metahuman]. Done when a named fibre's contraction provably moves the skin surface above it, reported in millimetres by md_measure (which is BUILT, PROMOTED and gate GREEN 9 of 9 today and merely registered-dark), with a neg-control proving a fibre at rest moves it zero. DELIBERATELY LAST AND GATED BEHIND SB4: an active contractile element without a stress measure is a shape effect wearing a physics name, so the tensor rung must land first or this one cannot be judged.
Matrix cell	muscle-contraction-driven-skin · `runtime/_hdl_build/nx_myo.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_myo`; declare `nx_myo TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

SB18 Footfall impulse that scales with gait speed: the walk kick becomes the run kick · cell NO-MATRIX-ROW · gate NO-MATRIX-ROW

Contract	`sb_impulse_mob` · executor Organ · 1.5 u · stands on nothing
Definition of done	OPERATOR 2026-09-16, with reference frames: real body physics is the subtle jiggle of breasts, glutes and thighs on walking impact that SHIFTS with running. The plant is already derived from measured human bands (fn 4137 mHz, zeta 480 permil at the linear limit, cai2018 up-down asymmetry, mills2025 AP and ML anisotropy) and the heel-strike impulse is already isqrt(2 x GRAV x 2 x bob); what is missing is the ADOPTION: sb_impulse_mob is BUILT and called by no world, and the impulse does not yet read cadence and vertical excursion from the gait, so a run does not kick harder than a walk. Done-rule: every world calls sb_impulse_mob at double-support; peak tissue excursion is MONOTONIC in gait speed on one plant (walk below jog below run); the ringdown after each strike stays inside the cited band with nx_dyna_bind_skin as the referee; a zero-speed body receives zero impulse (neg-control); and the referee panel grades a fixed-camera walk-then-run capture against the operator's lens before ADOPT.
Matrix cell	NO-MATRIX-ROW — no matrix row carries `sb_impulse_mob`; the ranker reads this rung as unmapped and no cell can flip for it. Add the row and the join completes on the next publish.
Gate	the gate join needs a matrix row: no organ is named for this rung
Work log	2026-09-16 lesson OPERATOR 2026-09-16 verbatim intent: we are getting softbody so that it has real body physics to the subtle jiggle of breasts and butt and legs on walking impact that shifts with running, and hips that move; two web reference gifs and a local strut.gif were named. They are referee LENSES and stay outside the estate (the GE36 precedent), never mirrored, never a capture. The mechanism the ask needs is adoption before invention: sb_impulse_mob is built and uncalled, the bands are banked, the springs are wired; SB18 wires the kick to speed and SB19 moves the drive from vertical bob to the pelvis, with gameengine GE75 emitting the pelvis pose.

SB19 Tissue driven by pelvis motion: belly, glute and thigh springs ride obliquity, rotation, tilt and lateral sway, not only vertical bob · cell NO-MATRIX-ROW · gate NO-MATRIX-ROW

Contract	`sb_pelvis_drive` · executor Organ · 2 u · stands on SB18
Definition of done	The reference frames show tissue following the PELVIS (hips that move, the strut, the hoop), not a vertical bounce; today the R7 belly, glute and thigh springs take a vertical gait drive and a wander heading. Done-rule: the spring anchors read the pelvis pose from the gait sub-model (gameengine GE75 wc_pelvis_gait) so lateral sway and rotation excite the AP and ML axes the anisotropy already declares; with the pelvis held still the output is byte-identical to the incumbent (neg-control); a hoop-class circumduction fixture drives belly and thigh excursion in phase with the pelvis and the referee reads a phase lag inside the plant's own band. The reference frames stay OUTSIDE the estate as referee lenses; no identity capture.
Matrix cell	NO-MATRIX-ROW — no matrix row carries `sb_pelvis_drive`; the ranker reads this rung as unmapped and no cell can flip for it. Add the row and the join completes on the next publish.
Gate	the gate join needs a matrix row: no organ is named for this rung
Work log	no log row names this rung yet

SB0 Board-level journal anchor for softbody · cell NO-MATRIX-ROW · gate NO-MATRIX-ROW

Contract	`-` · executor Board · 0 u · stands on nothing
Definition of done	BOARD-LEVEL JOURNAL ANCHOR, DECLARED 2026-09-27 so board-wide findings have an address a log row can resolve to; journal rung work against its rung.
Matrix cell	NO-MATRIX-ROW — no matrix row carries `-`; the ranker reads this rung as unmapped and no cell can flip for it. Add the row and the join completes on the next publish.
Gate	the gate join needs a matrix row: no organ is named for this rung
Work log	2026-09-27 land LADDER DECLARED 2026-09-27: T-BIC-SOFT (Newton, Warp, IPC, FEBio) and T-FR-SOFT (Fabric-101 wrinkles, millimetre tissue, one solver for hair, cloth and skin); twenty rung roles. The board leaves NO-LADDER. The beach-lounger reference class makes SB15 (a support surface that spreads load) and SB13 (cloth against the body) the rungs graphics GR60 waits on; nx_compare_rank reads both as DARK-CONTRACT (organ exists, symbol undeclared). 2026-10-01 measure OCTOBER ATTESTATION 2026-10-01: nx_barfresh softbody verdict FRESH, 3 of 3 bars ATTESTED. SB-TISSUE-2609 and SB-TISSUE-LOUNGER-2609 by nx_frontier_scan attest (job 1790822299: the arXiv listing read, 8 items, newest 2026-09-30, on-subject newest DiT-Garment 2609.18510 and the stress-free inverse FEM for soft tissue 2609.18109, nothing dated October); SB-SOLVER-2607 by the seat (the NVIDIA SIGGRAPH 2026 post published 2026-07-20 and modified 2026-07-21, nothing newer; the 17-source fan-out found nothing newer than July). RANK by nx_compare_rank (artefact sha256 f33d2936, target 1.0 milestone M2): critical path #1 SB11 contact that survives deep interleaving (st_sdf_contact, priority 225, DARK-CONTRACT), #2 SB7 tetrahedralisation of an arbitrary watertight surface with a sliver oracle (st_tet_from_surface, UNMAPPED), then SB4 Cauchy stress and strain pulled above SB5 the FEBio oracle, SB18 footfall impulse, SB15 support surface; PM filed 2 of the top 3. SB7 is the substrate aesthetictwin AT25 needs for the surgery chain

dossiers=20 cells_joined=14 no_matrix_row=6 (sum 20) gates_recorded=0 gates_unrecorded=0 no_gate_declared=14 (sum 14 of cells_joined) evgates=UNSTAMPED

Capability comparisons

Compare the field, search individual capabilities and open their source and adoption evidence. Documented presence does not establish comparative quality.

Position map — centrality and distinctiveness

The four-quadrant map the field uses for brand strategy (Dawar and Bagga, HBR June 2015), re-derived from this matrix on every publish. Centrality is the share of the category's feature mass a player covers, each feature weighted by how many hold it; distinctiveness is the average lead over each rival on the rows the player holds; breadth is the depth-weighted share of the whole matrix (the bubble); depth is how deeply the rows held are held; momentum is the day-over-day move off the spine (green rising, red falling, grey until day two); the dashed path runs first day → previous day → today. Dividers are the category means. Axes are fitted to the field of play, so read the tick numerals, not the frame. Rival marks are documented presence, so a rival's position reads the record, never its quality. The picture grades its own readability below; the table beside it is the same data for a screen reader or a second method.

Rows. 27 feature rows mapped of 33 matrix rows; 6 substrate rows (kernels, bounds and instruments the plan's rungrole marks substrate) are the estate's scaffolding, not category features, and stay off the map

Views. 2D cut: centrality, distinctiveness · 3D cube: centrality, distinctiveness, breadth · axes registered: centrality, distinctiveness, breadth, depth, momentum · a board picks its own in knowledge/compare/softbody.cdmap (cut|x|y, cube|x|y|z); absent = the HBR defaults

bubble area = breadth · ring = momentum (green rising, red falling, grey until day two) · dashed = category means · axes fitted to the field of play: centrality 400–900, distinctiveness 0–700 of 0–1000 permil (the full range put every player in one corner)
readability of the 2D cut, self-graded by the layout ruler: label overlaps 0 · labels over marks 0 · off-canvas 0 · unresolved labels 0 · mark overlaps 0 (a fact of the data: two players that close are that close) · data spread 583 permil of the plot · quadrant words unseated 0
text contrast, measured with wcag2-ratio (floors from contrast.conf), light theme: labels 16.24 (floor 4.50) · notes 16.24 (floor 4.50) · quadrant words 5.89 (floor 4.50) · tick numerals 5.89 (floor 4.50) · axis titles 5.59 (floor 4.50) · dark theme: labels 13.78 (floor 4.50) · quadrant words 5.55 (floor 4.50) · axis titles 5.97 (floor 4.50) · dark classes under their floor 0 (one figure serves both themes: a dark shortfall is a token to fix, never a class to hide) · classes refused under their floor 0 (a refused class is not drawn; the scale classes are measured, never hidden) · export: SVG PNG (receipt, rendered by the estate's own rasteriser from this page)

axes fitted: centrality 400–900 · distinctiveness 0–700 · breadth 100–700 permil · farther bubbles are painted first, nearer ones over them · the floor shadow is each bubble's (x, y) at height 0
readability of the 3D cube, self-graded by the layout ruler: label overlaps 0 · labels over marks 0 · off-canvas 0 · unresolved labels 0 · mark overlaps 2 (a fact of the data: two players that close are that close) · data spread 180 permil of the plot

Nishi SOFA Project Chrono Bullet FEBio
10 panels over 5 registered axes, every pair once (the lower axis on x, the higher on y) · each panel fitted to its own field of play, first and last tick numerals shown · no labels in a small cut, the legend names the colours; the grade below reports the mark terms only (MM summed over the panels, spread averaged)
readability of the small multiples, self-graded by the layout ruler: label overlaps 0 · labels over marks 0 · off-canvas 0 · unresolved labels 0 · mark overlaps 4 (a fact of the data: two players that close are that close) · data spread 639 permil of the plot

Player	Quadrant	centrality	distinctiveness	breadth	depth	momentum	Rows held	Days on spine	First seen
Nishi	Aspirational	750	570	580	783	+66 since 2026-09-27	20	11	2026-09-15
SOFA	Mainstream	796	275	358	743	-20 since 2026-09-27	13	11	2026-09-15
Project Chrono	Mainstream	750	229	308	694	-18 since 2026-09-27	12	11	2026-09-15
Bullet	Peripheral	546	104	172	583	-27 since 2026-09-27	8	11	2026-09-15
FEBio	Peripheral	687	250	296	727	-21 since 2026-09-27	11	11	2026-09-15

Day	Matrix rows reviewed	Players recorded
2026-09-15	33	5
2026-09-16	33	5
2026-09-17	33	5
2026-09-18	33	5
2026-09-19	33	5
2026-09-20	33	5
2026-09-21	33	5
2026-09-22	33	5
2026-09-23	33	5
2026-09-27	27	5
2026-10-08	27	5

How to read this evidence. A Nishi source mark records whether the named symbol is declared in the selected source file. It does not establish working behavior or accepted delivery. A watching tag names an expected source entry; discovering it updates source coverage only. A dark tag means the file exists but the named symbol was not found: inspect the source and contract before assuming the capability is missing or renaming code. Peer codes are authored assessments; unassessed means no comparison is established, not absence. Qualification, deployment identity, runtime observations and acceptance must carry their own evidence. Adoption records the catalog's observed path through source, build, promotion, registration and invocation; libraries use importer evidence and gates use declared execution surfaces. A full catalog path does not establish quality, freshness or a successful user outcome. Partial and unknown states retain their stated next action. Census stamps: importers asof 1790637268, gate census asof 1790104131 (unix seconds; -1 = census absent).

Capability matrix — measured against source

◉ authored stronger claim; source present● present◐ partial○ absent? unassessed; excluded from relative comparison · click any capability for its evidence

Peer cell coverage: authored assessed 132, unassessed 0. Authored assessments are not verified outcomes.

Capability	Nishi	SOFA	Project Chrono	Bullet	FEBio
xpbd-tet-cage-layered-compliance Measured: `st_substep` exists in `runtime/nx_softtissue.nx`, verified at emit. XPBD (Macklin 2016) over a tetrahedral cage, five named layers (muscle, gland, adipose, skin, ligament) each with its own distance and volume compliance read from data rows. alpha_tilde = alpha/dt^2 makes effective stiffness STEP-SIZE INDEPENDENT, which is the property plain PBD lacks and the reason a stiff setting does not buzz. SOFA [sofa] and FEBio are BEST here on constitutive fidelity (true hyperelastic FEM); Bullet is PART (mass-spring plus basic FEM). We are at the principled tier and integer-only. Adoption: `LIB-WIRED importers=14 nonval=4` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Bullet; fully adopted on the estate ladderCons behind SOFA (leads), FEBio (leads)	●	◉	●	◐	◉
neo-hookean-volume-constraint Measured: `st_six_vol` exists in `runtime/nx_softtissue.nx`, verified at emit. Incompressibility as C = 6V - 6V0 per tetrahedron. Correction magnitude scales as 1/(6V) so resistance rises without bound as the element collapses: tissue gets HARDER the harder it is pressed and a deep press dimples instead of inverting. That is the Neo-Hookean barrier, and it is proven by tooth (a capsule pressed 60 percent into the cage, every 6V asserted positive), not asserted. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=14 nonval=4` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Bullet; fully adopted on the estate ladderCons behind SOFA (leads), FEBio (leads)	●	◉	●	◐	◉
volume-conservation-gate Source present; authored stronger claim: `cg_bulge` in `runtime/_hdl_build/nx_softtissue_conserve_gate.nx`, verified at emit. GREEN 8/8: indented volume conserved at EVERY depth, far-side bulge monotonic, damped solvers reach EXACT fixed points, both detectors bite-proven. NO RIVAL SHIPS A CONSERVATION GATE, and a float solver cannot state tolerance zero. This is the displaced-volume claim measured rather than described. Adoption: `GATE:LIVE trial=GREEN` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of SOFA, Project Chrono, Bullet, FEBio; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○
metamorphic-relations-gate Open — no implementing organ is measured for this axis yet. GREEN 12/12: translation, axis-permutation, reflection, time-translation invariance and quiescence over the SHIPPING solver, with a 3x3 mutation matrix and every detector bite-proven. Oracle-free by construction, so it needs no rival to grade against and cannot be gamed by matching a reference implementation. Pros none measured yetCons no implementing organ measured	○	○	○	○	○
ringdown-system-identification Source present; authored stronger claim: `dbs_ringdown` in `runtime/nx_dyna_bind_skin.nx`, verified at emit. FIRST RUN EVER 2026-08-25. Excites the XPBD plant and identifies its natural frequency and damping ratio by log decrement (dbs_zeta_permil over successive peaks). Measured at the linear limit: fn=4137 mHz, zeta=480 permil -- INSIDE the published human bands 4000..5200 and 350..600. This is the organ that licenses comparing any tunable to a subject row, and no rival has a counterpart because no rival states a motion bar to identify against. Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_dyna_bind_skin_gate -- a mention, not a run. not adopted: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_dyna_bind_skin_gate -- a mention, not a run Pros Source symbol found; stronger capability claim is authored; ahead of SOFA, Project Chrono, Bullet, FEBioCons not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_dyna_bind_skin_gate -- a mention, not a run	◉	○	○	○	○
cited-biomechanics-oracle-bands Source present; authored stronger claim: `gf_band` in `runtime/_hdl_build/nx_gamefeel_oracle_gate.nx`, verified at emit. Every feel and tissue constant is a CITED ROW with its source in the row: free-vibration modes, haake-scurr-2010 damping 0.475 +- 0.062, cai2018 piecewise stiffness, mills2025 anisotropy, multiplanar walk and run amplitudes, chen2024 surface-error parity bar. A tunable without a row here CANNOT PASS its gate. Rivals hardcode material constants in scene scripts with no citation and no check. Adoption: `GATE:LIVE trial=RED` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of SOFA, Project Chrono, Bullet, FEBio; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○
oracle-enforced-in-the-ship-lane Source present; authored stronger claim: `ce_evidence` in `runtime/nx_craft_emit.nx`, verified at emit. The feel oracle runs as step 1b of the page ship lane and the emitted page PUBLISHES the derived figures, computed at emit time from the shipping solver. A drift between code and citation therefore REFUSES THE SHIP rather than being discovered later. This is the anti-drift mechanism as a mechanism, not a rule. Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). not adopted: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked) Pros Source symbol found; stronger capability claim is authored; ahead of SOFA, Project Chrono, Bullet, FEBioCons not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	◉	○	○	○	○
anisotropic-sign-asymmetric-step Measured: `sd_step_ax` exists in `runtime/nx_softdyn.nx`, verified at emit. Per-axis stiffness and damping plus a piecewise vertical branch selected by ONE SIGN TEST on displacement, exact in integers: cai2018 measures tissue about 9x stiffer below static equilibrium than above, and mills2025 puts anterior-posterior and medio-lateral excursion at 610 and 600 permil of vertical. A single scalar K asserts 1000 permil on both axes and symmetric branches, which is why isotropic jiggle reads as jelly. Rivals CAN express anisotropy in a full FEM material; none ships the measured human numbers as the default. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=8 nonval=6` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of SOFA, Project Chrono, Bullet, FEBio; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	◐	◐	○	◐
reduced-order-model-identified-from-solver Source present; authored stronger claim: `sb_k_up` in `runtime/nx_softbind.nx`, verified at emit. The runtime spring's stiffness and damping are DERIVED from the cited band by physics (K = 1024(2pif/rate)^2, C = zeta2sqrt(1024K)) and the piecewise pair is scaled by (1 + 1/sqrt(R)) so the free-vibration PERIOD -- the sum of two half-periods -- lands on target. Nothing here is chosen by eye. The retired hand-tuned plant is refused by the gate's neg-control, which is what proves the band discriminates. Adoption: `LIB-WIRED importers=97 nonval=93` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of SOFA, Project Chrono, Bullet, FEBio; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○
per-individual-seeded-bodies Measured: `sb_full_from_gene` exists in `runtime/nx_softbind.nx`, verified at emit. Each girl's plant is drawn from her WHOLE genome on a single fullness axis, using the same hash and the same salt as nx_softtissue st_profile_seeded so body and dynamics are ONE DRAW. 1001 distinct plants where the shipped engine had 4 (two genome bits shared across a sixteen-girl cast). Fields are deliberately NOT drawn independently, because independent draws produce implausible combinations. Adoption: `LIB-WIRED importers=97 nonval=93` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of SOFA, Project Chrono, Bullet, FEBio; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	○	○	○	○
gait-coupled-tissue-drive Measured: `wc_gait_ang` exists in `runtime/nx_wasm_craft.nx`, verified at emit. LANDED 2026-08-25, gate-proven 74/74 -- and the FIRST design was refuted by this organ's own gate, which is why it is trustworthy. The drive was a flat clock at a fixed 2.50 Hz whatever her real speed, while the render clip is DISTANCE-synced, so foot plant and tissue drive were uncorrelated. Advancing MC_PH by distance was tried and REFUSED at 52/53 RED: heading is derived FROM that phase (dx = sin(ph3)), so a girl who stopped froze her heading and could never restart -- T36 caught the bold girl parked at her START distance. The shipped design SPLITS the two: MC_PH still free-runs for wander, and a separate accumulator holds PATH LENGTH in the needs row's spare slot, which wc_need_seed already zeroes and which is non-persisted like the springs -- so no schema change and no save-format change. It accumulates in q8 and divides ONCE at read, because the first cut lost the entire signal to per-tick flooring at ordinary walking speeds. At full stride the angle advances at EXACTLY the incumbent rate, so the calibrated cadence is reproduced bit-for-bit and only slower travel differs, which is what step-frequency-walk-to-run means. T36 back to bold dist=577 and T40 softbody unchanged. Adoption:* `LIB-WIRED importers=41 nonval=25` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of SOFA, Project Chrono, Bullet, FEBio; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	○	○	○	○
deterministic-integer-replay Source present; authored stronger claim: `sd_step` in `runtime/nx_softdyn.nx`, verified at emit. 100 percent integer, semi-implicit Euler, with a stability envelope clamped so an unstable configuration is IMPOSSIBLE TO REQUEST rather than merely discouraged. Replay is bit-identical, so netsync and save-transparency hold by construction. Every float rival is machine-dependent across hosts and cannot make this claim at all. Adoption: `LIB-WIRED importers=8 nonval=6` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of SOFA, Project Chrono, Bullet, FEBio; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○
single-process-real-time-in-browser Source present; authored stronger claim: `ce_ship` in `runtime/nx_craft_emit.nx`, verified at emit. The whole world, cast and tissue solver compile to ONE WebAssembly module served as a single page, no plugin and no second runtime. The published SOFA and Chrono pipelines route generator to solver to renderer over shared memory or UDP and concede in their own guidance that the solver cannot hold 90 Hz. Bullet [bullet] is the honest real-time comparison and is a native library, not a page. Fidelity is theirs; deliverability is ours. Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). not adopted: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked) Pros Source symbol found; stronger capability claim is authored; ahead of SOFA, Project Chrono, Bullet, FEBioCons not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	◉	○	○	○	○
truncation-dead-zone-rest Measured: `sd_deadsnap` exists in `runtime/nx_softdyn.nx`, verified at emit. Integer force terms truncate, so below a derived threshold BOTH spring and damping evaluate to exactly zero and the point coasts ballistically forever -- measured as a perfect period-30 limit cycle still running undecayed after 200000 ticks. A damped system that cannot reach its own rest state is a perpetual-motion defect. Thresholds are derived from K, C and the fixed-point base, never picked. A float solver never meets this class and so never guards it. Adoption: `LIB-WIRED importers=8 nonval=6` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of SOFA, Project Chrono, Bullet, FEBio; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	○	○	○	○
tangential-friction-law Measured: `st_set_friction` exists in `runtime/nx_softtissue.nx`, verified at emit. MECHANISM PROVEN 2026-08-25 and DELIBERATELY UNARMED, which is the honest state and not a half-finished one. nx_contactfriction_gate reads GREEN 6/6 ON THE SERVED BINARY via /api/gate_run -- built, promoted under expect_sha256, registered, invoked and rostered. Measured surface carry over a 15 mm drag: unset -396, mu 0 -396 (byte-identical to unset, which is how UNSET is PROVEN frictionless rather than assumed), mu 0.25 +35, mu 0.9 +51 cmm. A positional Coulomb cone is PORTED from this estate's own nx_phys3d rather than written a second time -- the tangential correction is clamped to mu times the normal correction, so a dragged probe carries the surface with it instead of sliding frictionlessly through. mu carries THREE states, not two: a number, ZERO, and UNSET. Zero is a measured frictionless surface; UNSET means no citation exists and the solver declines to invent one, and a tooth proves UNSET behaves exactly as frictionless rather than silently defaulting to something plausible. Coded PRESENT, never EXCEEDS: SOFA's constraint-based friction contact pipeline is more complete than one Coulomb cone and is BEST here. See the coefficient row below for why it ships unarmed. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=14 nonval=4` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind SOFA (leads)	●	◉	●	●	●
skin-friction-coefficient-sourcing Open — watching `knowledge/gamefeel_oracle.conf` : `tissue_mu_skin_permil`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. OPEN, AND THIS IS A LITERATURE GAP RATHER THAN A BUILD GAP -- the mechanism above arms itself the moment this row lands, with no code change. Two open-access papers were fetched and content-pinned on 2026-08-25 hunting a macroscopic skin-on-skin Coulomb coefficient and NEITHER supplies one: one is a Frictiometer FR 770 device index over 633 subjects [skinfriction], which is an instrument reading and not a coefficient, and the other reports microasperity values of 0.004 and 0.001 [skinmicro], which are three orders of magnitude away from the number graphics work quotes. The commonly-repeated 0.2 to 0.5 band traces only to a paywalled review nobody here has read, so it is NOT in the conf -- writing it in would be exactly the fabricated-constant defect this estate refuses everywhere else. THE COLUMNS HERE RECORD CAPABILITY-TO-SET, NOT PROVENANCE: all four rivals expose a friction coefficient as a settable parameter, and whether any of them sources it from tissue literature is something I did not measure, so no claim is made about it. Bar: one cited macroscopic coefficient admitted to the oracle conf with its source in the row. watching tissue_mu_skin_permil Pros none measured yetCons behind SOFA, Project Chrono, Bullet, FEBio; open contract; expected source declaration not established	○	●	●	●	●
self-contact-between-surfaces Measured: `st_self_contact` exists in `runtime/nx_softtissue.nx`, verified at emit. LANDED 2026-08-25, nx_selfcontact_gate GREEN 8/8 ON THE SERVED BINARY via /api/gate_run -- built, promoted under expect_sha256, registered, invoked and rostered on the slow gate roster, with the NAS run reporting numbers identical to the local one. Two cages in mutual contact resolve to a stated separation: closest approach 819 to 997 cmm against a 1000 cmm target, a six-pair contact patch, both bodies corrected by inverse-mass split. Graded by st_sc_min_sep, an INDEPENDENT closest-approach scan -- grading the routine on its own contact count would score a routine that resolved nothing exactly like one that worked. THE CONTACT DISTANCE IS DERIVED, NOT CHOSEN: st_surface_spacing measures the mesh's own mean surface-particle spacing, because a distance below it reduces every contact to a single point and lets surfaces slip BETWEEN particles -- measured, a hand-picked 300 cmm gave a peak of ONE pair on every pass. Broad phase proven free: two cages a metre apart cost zero pair tests. Coded PRESENT, never EXCEEDS -- SOFA's constraint-based pipeline remains BEST and this is one honest rung of it. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=14 nonval=4` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind SOFA (leads)	●	◉	●	●	●
self-contact-deep-interleaving DARK — `runtime/nx_softtissue.nx` EXISTS but does not declare `st_sdf_contact`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. OPEN, AND NAMED BECAUSE WE MEASURED OUR OWN BLIND SPOT RATHER THAN WAITING FOR IT TO SURFACE. Contact is detected by surface-particle proximity, so it resolves GRAZING contact and is blind to DEEP INTERLEAVING once one cage's shell has passed bodily inside the other: pushing a cage 600 cmm past touching took the measured closest approach from 288 UP to 390 cmm and the contact count to ZERO, because the nearest surface particle of the other body is then far away. st_sc_min_sep inherits this exactly -- it is a surface-to-surface distance and is NOT a penetration depth, so it is non-monotonic once bodies interleave. A CONTACT ROUTINE THAT CANNOT SEE THE TANGLED CASE REPORTS IT AS NOT TOUCHING, WHICH IS THE MOST REASSURING POSSIBLE WAY TO BE WRONG, so it is written into the organ's own header and onto this board rather than left for a caller to discover. Bar: a signed distance field or continuous collision detection that resolves an interleaved pair, with a neg-control proving the proximity model reports that same pair as clear. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them dossierwatching st_sdf_contact Pros none measured yetCons behind SOFA (leads), Project Chrono, Bullet, FEBio; open contract; expected source declaration not established	○	◉	●	●	●
stress-strain-tensors DARK — `runtime/nx_softtissue.nx` EXISTS but does not declare `st_cauchy_stress`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. OPEN and the honest ceiling on every clinical claim. We carry per-layer COMPLIANCE, not Young's modulus and Poisson ratio, so deformation cannot yet be cross-checked against an analytical solver. Bar: instantiate a cube and a cylinder of known dimensions, load them uniformly, and agree with FEBio within a stated tolerance. Until that lands, no parity-with-clinical claim may be published on this board. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them dossierwatching st_cauchy_stress Pros none measured yetCons behind SOFA (leads), Project Chrono (leads), FEBio (leads); open contract; expected source declaration not established	○	◉	◉	○	◉
febio-analytical-oracle DARK — `runtime/nx_softtissue.nx` EXISTS but does not declare `fo_febio_compare`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. OPEN. The rival-as-oracle rung: a headless harness that runs the same primitive load case through our solver and through a published analytical or FEBio [febio] result and reports the deviation as a number. This is how a third party becomes a BENCHMARK rather than a dependency, and it is the only route by which the tensor row above can ever be judged. dossierwatching fo_febio_compare Pros none measured yetCons behind SOFA, Project Chrono, FEBio (leads); open contract; expected source declaration not established	○	●	●	○	◉
surface-error-vs-4d-scan Measured: `md_measure` exists in `runtime/nx_mmdev_lib.nx`, verified at emit. I PUBLISHED THIS AS ABSENT ON 2026-08-25 AND IT WAS WRONG. nx_mmdev is BUILT, PROMOTED (33047 B), REGISTERED and AUTHORISED, with its gate GREEN 9/9 -- an exact full-population point-to-triangle surface deviation in tenths of a millimetre (md_measure, md_grid_build, md_dist2_pt_tri, md_mean). It is REGISTERED-DARK: callable, proven, and INVOKED NO. So the RULER exists and the bar exists (conf tissue_surface_err_mm 0..7, chen2024 at 7.05 mm mean euclidean); what is missing is the pairing of the two. The remaining work is to route one production consumer at md_measure with a reference mesh, which nx_gltf2mesh can already produce -- a wiring job, not a build. No fidelity claim is made until that runs. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=9 nonval=5` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Project Chrono, Bullet; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	●	○	○	●
contact-force-readback Measured: `st_contact_force` exists in `runtime/nx_softtissue.nx`, verified at emit. LANDED 2026-08-25, nx_contactforce_gate GREEN 6/6. The solver resolved contact but never REPORTED it, so nothing downstream could feel firmness. It now accumulates the hard projection the contact constraint performs, summed over a substep's iterations and zeroed at its start -- which IS the constraint impulse. MEASURED: a 10 mm press reads 100726181, a 20 mm press 323948164, and the SAME 20 mm press on firm tissue 1215107751. Two facts fall out that a fake cannot produce: doubling the depth raises force 3.2x, NOT 2x, because the Neo-Hookean barrier is stiffening under the probe; and firm reads 3.75x harder than soft at equal depth. UNITS ARE STATED AND LIMITED: inverse mass is a uniform q12 so there is NO KILOGRAM in this solver and an absolute newton is not derivable -- it would need a density model the profile rows do not carry. That is why the teeth assert MONOTONICITY and DISCRIMINATION, which are unit-independent and are what a haptic loop actually consumes. Adding it perturbed nothing: nx_softtissue_gate reads IDENTICAL live-vs-staged, first_diff_off=-1. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=14 nonval=4` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind SOFA (leads), Project Chrono (leads)	●	◉	◉	●	●
frame-budget-at-90hz Measured: `tb_measure` exists in `runtime/nx_tissuebudget.nx`, verified at emit. MEASURED AND CLOSED 2026-08-25, GREEN 6/6. The bar had been banked for weeks and nothing had ever measured against it. Result: the whole shipped cast (16 mobs x 3 spring points) costs 28496 ns = 28 us per tick, against a 90 Hz frame budget of 11111 us -- 0.25 percent of the frame, and 1.4 percent of the cited js_block_ms 2000 us app-logic ceiling that actually governs it. Even at fifty times the native cost the browser stays inside that ceiling, so THE TISSUE SOLVER IS NOT THE CONSTRAINT and optimisation effort belongs elsewhere. Two independent budgets corroborate: the DERIVED 1e6/fps and the CITED frame_budget_ms_90hz row agree, which is its own tooth. The figure is labelled a NATIVE FLOOR -- decisive if it had missed, never proof the browser fits. Sample size is derived from the clock (run doubles until 1 us of error is under one part in a thousand) and it ABSTAINS rather than dividing an unresolved 0 into a per-tick cost of 0. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `RUN-BY:roster` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; ahead of SOFA, Project Chrono, FEBio; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	○	○	●	○
tetrahedralisation-from-surface Measured: `st_add_edge` exists in `runtime/nx_softtissue.nx`, verified at emit. PART. We build our own cage with a lattice step DERIVED from the particle budget (dbs_derive_cell), which is why the ringdown reports cell_mm rather than taking it as an argument. What is absent is tetrahedralising an ARBITRARY watertight surface the way fTetWild and Gmsh do, which is what a general asset pipeline needs. Bar: a manifoldness and sliver oracle that REFUSES a mesh with zero-volume elements before it reaches the solver. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=14 nonval=4` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Bullet; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	●	●	○	●
animal-and-quadruped-biomechanics Measured: `quad_gait_pose_g` exists in `runtime/nx_quadruped.nx`, verified at emit. I PUBLISHED THIS ROW AS "UNSTARTED" ON 2026-08-25 AND IT WAS WRONG -- corrected the same day by an existence audit over this board's own ABSENT rows. nx_quadruped.nx is a real 10014 B rig (18 joints QR_ROOT..QR_TAIL2, quad_build, quad_gait_pose_g with 0=walk 1=trot 2=gallop, quad_add_horns), and TWO gates were compiled and had NEVER RUN: nx_quad_gaits_gate 128583 B and nx_wolf_walk_gate 141131 B, both BUILT-UNPROMOTED because both rolled their own verdict and /api/promote refuses D001 by name. Migrated nx_quad_gaits_gate onto gv_check, promoted, registered, and ran it for the FIRST TIME EVER: GREEN 4/4, emitting a real 52937 B gallop APNG and a 55126 B horned-trot APNG. STILL OPEN and now countable: no cited stride-frequency band in the oracle conf, so the gaits are proven DISTINCT and ANIMATED but not proven CORRECT against a published animal. That is the remaining bar, and it is a data admission, not an organ. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-GATE-ONLY importers=2` — PARTIAL: imported only by validation organs (gates, tests, benches): wire it into a shipping program. not adopted: imported only by validation organs (gates, tests, benches): wire it into a shipping program Pros Implementing source symbol found; runtime acceptance separateCons behind Project Chrono (leads); not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	●	●	◉	●	●
environment-soft-body DARK — `runtime/nx_water_erosion.nx` EXISTS but does not declare `we_shear_transport`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. PART. Terrain deforms through erosion, but soil and vegetation do not respond to LOAD -- no footprint, no compaction, no bending under contact. Chrono [chrono] is BEST here (soft-soil terramechanics is its home ground). Bar: a load-bearing ground model whose deflection under a stated pressure is cited, not tuned. dossierwatching we_shear_transport Pros none measured yetCons behind SOFA, Project Chrono (leads), Bullet, FEBio; open contract; expected source declaration not established	○	●	◉	●	●
hair-strand-to-body-collision DARK — `runtime/nx_softtissue.nx` EXISTS but does not declare `st_strand_contact`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. OPEN, AND DELIBERATELY NARROW BECAUSE HAIR IS ALREADY OWNED TWICE ELSEWHERE AND A THIRD COPY WOULD BE THE DUPLICATE-RULER DEFECT. Measured over the whole compare tree on 2026-08-25: strand DYNAMICS shipped on /compare/graphics as nx_hairdyn_lib hd_ringdown (GREEN 11 of 11, bite-proven, driven as a chain on THIS board's own solver), hair GEOMETRY is that board's gpe_hair_geometry watch, and the strand GROOM that generates the guides is procgen's gr_strand. What no board owns, and what only a soft-body engine can own, is the CONTACT between a strand and the deforming tet cage. The bar is strand-based and real-time in the field: Unreal's groom system renders each individual strand with physically accurate motion at hundreds of thousands of strands [uegroom], and the current research bar is the Augmented Mass-Spring model, which adds edge, bending and torsional degrees of freedom plus a biphasic interaction through two extra one-way springs to simulate thousands of interacting strands in real time [hairams]. READ HONESTLY, THAT PAPER ALSO NAMES OUR GAP AS ITS OWN: its contribution is hair-HAIR interaction, and it cites improved hair-BODY interaction as separate prior work. Bar: a strand chain whose root follows the cage and whose segments resolve against it, with a neg-control proving a strand released inside the body is expelled rather than trapped. dossierwatching st_strand_contact Pros none measured yetCons behind SOFA, Project Chrono, Bullet; open contract; expected source declaration not established	○	●	●	●	○
cloth-to-body-and-multilayer-garment-contact Measured: `st_cloth_contact` exists in `runtime/nx_cloth_contact.nx`, verified at emit. OPEN, AND SCOPED TO THE HALF NOBODY OWNS. Cloth is already contracted on THREE other boards and none of them is this row: procgen owns the SOLVER (gm_xpbd_step, stretch shear and bending compliance over the garment edge list, rung PG29), dcc owns PATTERN SEWING AND DRAPE (gd_drape, rung DC10), and charsim owns cloth and hair riding one solver (dyna_cloth, R8). procgen's row already states the solver must ride THE TET CAGE'S CONTACT -- this row is that contact, plus the two things none of the three name: GARMENT-ON-GARMENT layering, and TEARING. The field bar is explicit that collision comes from the body rig: Unreal's Chaos Cloth is a low-level particle solver whose collision is supplied by the character's Physics Asset [chaoscloth], while the pattern-first tools are Marvelous Designer [marvelous] and CLO [clo3d], both of which simulate a sewn garment rather than a painted one. SOFA is BEST here on constraint-based contact. Bar: two garment layers over a pressed cage with no interpenetration at a derived contact distance, and a tear that propagates along a stated stress threshold rather than at a chosen edge. REPOINTED 2026-09-30 to runtime/nx_cloth_contact.nx, an organ of its own that IMPORTS the tissue solver and the PG29 cloth solver (nx_softtissue.nx is shared and was being extended by another lane the same hour). METHOD: Offset Geometric Contact -- an orthogonal vertex-facet and edge-edge contact set and the per-vertex conservative bound that replaces continuous collision detection [ogc] -- with the contact distance and the layer-to-layer mass split read from Fabric-101's measured thickness and areal density [fabric101]; the newest penetration-free line, a barrier-free augmented Lagrangian over CCD-found pairs [barrierfree], is named and NOT built. The tissue half of the garment-to-tissue mass ratio the open two-way leg needs is NIST ICRP density data, admitted not wired [nist_adipose] [nist_softtissue]. Adoption: `LIB-UNCENSUSED` — PARTIAL: a library absent from the import-graph census (younger than the last eco-graph build, or unreachable from buildroot/runtime). dossiernot adopted: a library absent from the import-graph census (younger than the last eco-graph build, or unreachable from buildroot/runtime) Pros Implementing source symbol found; runtime acceptance separate; ahead of FEBioCons behind SOFA (leads); not adopted yet: a library absent from the import-graph census (younger than the last eco-graph build, or unreachable from buildroot/runtime)	●	◉	●	●	○
free-surface-liquid-wave-field Measured: `fl_step` exists in `runtime/nx_fluid.nx`, verified at emit. CORRECTED FROM AN ABSENCE I WAS ABOUT TO PUBLISH. A capability search for a fluid solver first returned the homonym trap twice over -- viscosity matches 28 times in runtime sources and every one classified is food-science rheology or FDM melt flow, and Navier matches 16 times and every one is the Octane NavierStokes JAVASCRIPT BENCHMARK in the JS VM lane. A glob then found nx_fluid.nx, which is REAL: a height-field shallow-water solver integrating the discrete wave equation on a height and velocity grid with reflecting boundaries, so sum of the laplacian is zero and VOLUME IS EXACTLY CONSERVED by construction, with fl_volume exposing it. It is all-integer like the rest of this board. Its own header states its limit before anyone else could -- FLIP is the particle variant and this is not it. So the honest coding is PART on our side and OUT OF SCOPE for three of the four rival columns, which are solid-mechanics engines; Chrono is the one that ships fluid-structure interaction. This row exists so the next reader does not repeat the search that nearly filed a false ABSENT. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-GATE-ONLY importers=1` — PARTIAL: imported only by validation organs (gates, tests, benches): wire it into a shipping program. not adopted: imported only by validation organs (gates, tests, benches): wire it into a shipping program Pros Implementing source symbol found; runtime acceptance separate; ahead of SOFA, Bullet, FEBioCons not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	●	○	●	○	○
viscous-film-surface-tension-and-clinging DARK — `runtime/nx_fluid.nx` EXISTS but does not declare `fl_film_cling`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. OPEN, AND IT IS THE GAP THE ROW ABOVE MEASURES RATHER THAN GUESSES. A height field is a 2.5D surface: it has no viscosity term, no surface tension, no wetting, and no notion of a film adhering to a moving body, so nothing in this estate can express a liquid that clings to skin, beads, sheets off, or drips. The rival columns are deliberately all coded zero because this axis is outside what all four advertise -- the real bar is a dedicated real-time liquid tool, and LiquiGen states its own scope as water through blood to ketchup and slime with real-time meshing, customizable forces and sparse simulation domains [liquigen]. Houdini FLIP is the offline bar and is named WITHOUT a cite mark on purpose: www.sidefx.com fails TLS in our own fetcher, so no mirror exists and an unfetched reference is a fabrication. Bar: a film whose thickness on the tissue surface is driven by a cited surface-tension and viscosity pair, with a neg-control proving a zero-tension film sheets off completely instead of beading. dossierwatching fl_film_cling Pros none measured yetCons open contract; expected source declaration not established	○	○	○	○	○
bedding-and-support-surface-load-distribution Measured: `st_support_surface` exists in `runtime/nx_support_surface.nx`, verified at emit. OPEN, AND DISTINCT FROM SB9 RATHER THAN A SECOND COPY OF IT -- the distinction is stated because this board already carries a ground row and two load-bearing rows would be the duplicate-ruler defect. SB9 environment-soft-body is OUTDOOR terrain: soil shear transport, footprints and compaction, where Chrono terramechanics is BEST. This row is the INDOOR support surface -- a mattress, cushion or bedding whose deflection redistributes body weight -- and its deliverable is not a dent but a PRESSURE MAP, because the quantity that matters is how load spreads across the contact patch. It composes what already landed rather than needing new physics: st_contact_force gives a per-contact reaction and st_self_contact gives two-body resolution with an inverse-mass split, so a support surface is a second cage with its own compliance rows. Bar: a body at rest on a compliant surface whose summed contact reaction equals its applied load within a stated tolerance, with the indentation depth and the contact-patch area both reported, and a neg-control proving a rigid surface concentrates the same load into a measurably smaller patch. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-UNCENSUSED` — PARTIAL: a library absent from the import-graph census (younger than the last eco-graph build, or unreachable from buildroot/runtime). dossiernot adopted: a library absent from the import-graph census (younger than the last eco-graph build, or unreachable from buildroot/runtime) Pros Implementing source symbol found; runtime acceptance separateCons behind Project Chrono (leads); not adopted yet: a library absent from the import-graph census (younger than the last eco-graph build, or unreachable from buildroot/runtime)	●	●	◉	●	●
engorgement-as-rest-volume-expansion Measured: `st_inflate_rest_volume` exists in `runtime/nx_softtissue.nx`, verified at emit. LANDED 2026-08-25, nx_engorge_gate GREEN 7/7 ON THE SERVED BINARY (built, promoted under expect_sha256, registered, invoked): the rest state grows rather than the skin being pushed, graded against the closed-form s^3 prediction -- commanded +20% linear predicts 1728 permil and measures 1667, correctly UNDER because only the adipose layer (80% of distance constraints) grew; reversible to pristine within 0.011%; layer-selective; a unity command on a pristine cage changes NOTHING to the byte; zero and negative factors REFUSE. How much any real tissue engorges is NOT claimed: no band is mirrored, the factor is the caller's, cite-or-abstain as friction ships. It was EXACTLY THE COUPLING procgen's OWN ROW SAYS IS MISSING -- which is why it is filed here and not there. The PHYSIOLOGY is already shipped and is procgen's: nx_arousal_lib a_perf models drive, lag, perfusion, temperature, heart rate and blood pressure as integer functions shared by simulate and fit so the two cannot drift, and its arousal-on-mesh gate is GREEN 17 of 17. That row states its own blocker in one line -- a pure normal push is not volume-conserving, and THIS BOARD'S conservation gate would refuse it. So the missing piece is not a physiology model and not a renderer: it is driving the tet cage's REST VOLUME from a_perf, so that engorgement is tissue expansion the Neo-Hookean volume constraint carries and the conservation gate can referee, rather than a displacement painted onto a surface. The rival columns are all zero because no general-purpose engine ships an inflation-driven rest-volume channel, which is a scope statement and not a boast. Bar: rest volume driven by a_perf with the conservation gate green at every level of the drive, and a neg-control proving the retired normal-push model FAILS that same gate. Adoption: `LIB-WIRED importers=14 nonval=4` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; ahead of SOFA, Project Chrono, Bullet, FEBio; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	○	○	○	○
muscle-contraction-driven-skin DARK — `runtime/_hdl_build/nx_myo.nx` EXISTS but does not declare `my_contract_fiber`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. OPEN, AND IT IS A DIFFERENT MECHANISM FROM THE FACS ROW TWO BOARDS ALREADY CARRY, WHICH IS WHY IT IS NOT A DUPLICATE. charsim R9 and graphics both watch fa_expr_morph, and that contract is BLENDSHAPES -- expression axes blended over a surface. This row is the other way round: an ACTIVE CONTRACTILE ELEMENT inside the tet solver, where a fibre shortens and the skin above it moves BECAUSE the flesh under it did, so micro-expression emerges from anatomy instead of being authored as a target. Ziva VFX is the production bar and states its own method plainly -- it constructs and simulates muscles, fat, fascia and skin and produces skin deformations informed by the underlying anatomy, and its team took a 2025 SciTech Academy Award [ziva]; MetaHuman is the delivery bar for a complete photoreal human with hair and clothing [metahuman]. THE MOST USEFUL THING IN THAT MIRROR IS ARCHITECTURAL, NOT COSMETIC: Ziva RealTime reaches real time by LEARNING to reconstruct shape from training sets, reproducing in milliseconds what otherwise takes seconds or minutes per frame -- which is the same offline-identify then cheap-at-runtime shape as our reduced-order model, arrived at independently. The difference is the one axis we can actually claim: ours is DERIVED by physics from a measured ringdown, so it carries a conservation guarantee and bit-identical replay, and a learned surrogate structurally cannot state either. Bar: a named fibre whose contraction provably moves the skin surface above it in millimetres by md_measure, with a neg-control proving a fibre at rest moves it zero. dossierwatching my_contract_fiber Pros none measured yetCons behind SOFA, FEBio; open contract; expected source declaration not established	○	●	○	○	●

Rival-claim provenance. Every Yes, Best or Part mark in a rival column is a claim about someone else's product. A cited row carries a reference mark that resolves to a pinned mirror (the refs gate measures that); an uncited row is an observation read and is badged in its evidence until a reference lands. The badge is the receipt, never the proof: a mark proves a mirror exists, not that the mirror supports the code. The rows are counted below.

Rival claims, counted
Class	Count	Note
rows with rival marks	20	rows carrying a Yes, Best or Part in a rival column
cited	7	a reference mark resolves to a pinned mirror
uncited	13	an observation read, badged until a reference lands
rows, summed	20	RECONCILED: the parts sum to the total 20

Delivery and evidence

Inspect risks, technical debt, rendered observations, experiments and references. Read scope and limitations alongside every result.

Risk register

Risk	Likelihood x impact	Mitigation
Publishing a fidelity or clinical claim before its oracle exists	certain x high	SB4, SB5 and SB6 gate every parity claim; until they land the matrix says so in its own header and the notes refuse the comparison by name.
Adopting a rival engine as a dependency instead of an oracle, which is the exact sovereignty defect this estate exists to remove	possible x high	Operator standing order 2026-08-25: third parties are oracles and benchmarks only. SB5 is the sanctioned pattern -- compare against FEBio, never link it.
A real-time claim with no measured frame cost	certain x medium	SB1 first, and it is the cheapest rung on the board precisely so this cannot be deferred.
A material constant with no readable primary source being written into the oracle conf because it is widely quoted	likely x high	Measured on this very board 2026-08-25: the skin-friction coefficient every graphics paper cites at 0.2 to 0.5 could not be traced to an open-access primary source, and two fetched papers measure something else entirely. The rule that held is that the solver ships the MECHANISM with the constant UNSET and a tooth proving unset behaves as frictionless, so an absent citation cannot masquerade as a physical result.
Software informing therapeutic or clinical use is a regulated device under FDA SaMD and EU MDR	likely x high	Named as a row; route through the legal lane before any pilot. SB2's force readback is the first rung that makes this reachable and therefore the first that needs the review.
The box is saturated and solver sweeps are compute-heavy; the XPBD ringdown already exceeds a 120-second window for four profiles	certain x medium	Identification runs OFFLINE and its output is a handful of constants; the runtime never runs XPBD. Ask the spend gate before sweeps and use the governed build lane, which correctly refused a build during this very session.

Release history and work log

The worked plan, on the board. Every leg appends what it measured, landed, retracted, learned and left queued, per rung, so a crash leaves the next seat a ledger here rather than a transcript to mine. Kinds: measure, land, retract, lesson, queue. Newest last.

When	Rung	Kind	Entry
2026-09-10	SB1	measure	2026-09-09 investment measurement repair: SB1 and SB7 empty prerequisite fields normalized to canonical '-' through CAS jobs1789009244 and1789009256. All17 contracts and authored efforts retained. Scoped regeneration job1789009389 emitted59561-byte JSON and172424-byte HTML with no failures. JSON generated1789009389 parses17/17 nodes, malformed0, missing_dependencies0, graph_valid1. Public HTTPS job1789009405 returned200 application/json Content-Length59561. Rank remains UNBOUND and recommendations WITHHELD; visual and interaction review unavailable because no browser connected.
2026-09-16	SB18	lesson	OPERATOR 2026-09-16 verbatim intent: we are getting softbody so that it has real body physics to the subtle jiggle of breasts and butt and legs on walking impact that shifts with running, and hips that move; two web reference gifs and a local strut.gif were named. They are referee LENSES and stay outside the estate (the GE36 precedent), never mirrored, never a capture. The mechanism the ask needs is adoption before invention: sb_impulse_mob is built and uncalled, the bands are banked, the springs are wired; SB18 wires the kick to speed and SB19 moves the drive from vertical bob to the pelvis, with gameengine GE75 emitting the pelvis pose.
2026-09-27	SB0	land	LADDER DECLARED 2026-09-27: T-BIC-SOFT (Newton, Warp, IPC, FEBio) and T-FR-SOFT (Fabric-101 wrinkles, millimetre tissue, one solver for hair, cloth and skin); twenty rung roles. The board leaves NO-LADDER. The beach-lounger reference class makes SB15 (a support surface that spreads load) and SB13 (cloth against the body) the rungs graphics GR60 waits on; nx_compare_rank reads both as DARK-CONTRACT (organ exists, symbol undeclared).
2026-10-01	SB11	measure	TRIANGULATED BAR as of 2026-10-01 for penetration-free contact of deformable bodies: three method families hold the bar and none dominates. (1) The IPC lineage keeps CCD-checked non-penetration, and its SIGGRAPH 2026 form drops the log barrier for a second-order augmented Lagrangian that admits every CCD-detected pair at once, reporting up to 103x over GIPC [barrierfree_al_2026]; its thickness-offset ancestor for cloth and rods is C-IPC [cipc_2021]. (2) The GPU-barrier family keeps the barrier and makes it cheap with analytic eigensystems, 3x over the standard barrier and fully on the GPU [gipc_2024]. (3) The offset-geometry family drops CCD for per-vertex conservative displacement bounds and runs in real time, more than two orders of magnitude over IPC-based simulation [ogc_2025]. The newest listing read has no October entry; its newest on-subject entries are PhyVisGen (2609.25653, 2026-09-22, IPC soft-gripper contact for robot data) and TOGEARI (2608.14162v2, 2026-09-11, preconditioning condensed FE systems with IPC contact) [arxivsearch_ipc_20261001] ; witnesses [barrierfree_al_2026] [gipc_2024] [ogc_2025] [arxivsearch_ipc_20261001] ; where they disagree: on the barrier (GIPC keeps it, the 2026 IPC-lineage paper names it the cause of ill-conditioning and time-of-impact locking, OGC replaces it with a two-stage activation over an offset surface); on CCD (the IPC lineage and GIPC run it every iteration, OGC runs none); on what is guaranteed (the IPC lineage claims intersection-free trajectories for everything it simulates, OGC guarantees codimensional objects only, lets volumetric meshes skip the bound and resolve intersections after they occur, and does not protect body-cloth contact when the body is skeleton-driven); and on speed, where 103x over GIPC and more than 100x over IPC-based simulation are measured against different baselines on different scenes, so no shared benchmark ranks the three ; the estate's running lanes split, read from their own sources on 2026-10-01 while their builders were writing them (corrected the same hour from a first draft that judged both lanes from st_self_contact in nx_softtissue.nx alone): SB15 (st_support_surface in buildroot/runtime/nx_support_surface.nx) is a discrete one-sided point-face XPBD projection with no CCD, no displacement bound and no edge-edge test, and its own header says it has neither the IPC barrier nor the 2026 augmented Lagrangian, so it is BELOW all three; SB13 (st_cloth_contact in buildroot/runtime/nx_cloth_contact.nx) implements OGC's per-vertex conservative bound with truncation for vertex-facet and edge-edge contact and holds the tissue's own motion inside its bound, so on mechanism it is AT the real-time member of the triangle, with declared deviations (hard XPBD projection in place of OGC's activation energy and VBD, no parallel-edge filter, feature de-duplication in place of the Gauss-map block test) and a gate verdict this seat has not observed; SB11's measured blindness to deep interleaving still stands for the tissue-to-tissue path; and the contact barrier named as SB15's target is the 2020-2024 member of the triangle that its own lineage replaced in 2026, while the bound SB13 already carries is the guarantee built for a local parallel vertex-level step (OGC integrates it with Vertex Block Descent) and the closest structural fit for SB15 too (an assessment, not a measurement).
2026-10-01	SB15	measure	TRIANGULATED BAR as of 2026-10-01 for body-support interface pressure: SOTA is a subject-specific hyperelastic finite-element body (MRI or CT geometry, Ogden tissue) on a modelled support surface (foam, air cells with internal ideal gas, alternating-pressure cycles) that reports interface stress and internal tissue strain [jner_buttocks_2026], validated against interface pressures from calibrated mapping systems to within 3, 14 and 10 percent of peak pressure [katzgefen_2023]; measurement reports average and peak pressure, contact area and the high-risk area above 40 mmHg from mats of about 3.5 cm2 per cell [hybridtextile_2026]; the graphics side simulates bodies at rest on a pressure mat at scale (206K synthetic pressure images) and judges the simulation by transfer to real mats [bodiesatrest_2020]; the newest listing has no October entry and no on-subject September entry, its newest on-subject entry being diffusion-synthesised in-bed pressure maps from depth with a weight-consistency loss (2512.13757v2, 2026-02-12) [arxivsearch_pressuremap_20261001] ; witnesses [katzgefen_2023] [jner_buttocks_2026] [hybridtextile_2026] [bodiesatrest_2020] [arxivsearch_pressuremap_20261001] ; where they disagree: the Tel Aviv group holds that interface pressure is the wrong end-point because sustained internal stress concentration causes injury, while the measurement study and the graphics simulation treat the map itself as the output; the 2026 Chongqing framework states no measured-map validation in its abstract; the Tel Aviv validation compares peak pressures with maps reported by the surface manufacturer or the literature rather than a map of the modelled subject; the Tehran measurement is one participant at 3.5 cm2 resolution that its own authors call too coarse; and Bodies at Rest validates its simulator only indirectly through pose accuracy on real mats, so no witness publishes a full map-to-map error between a simulated and a measured pressure field ; the estate's running lane SB15 (st_support_surface in buildroot/runtime/nx_support_surface.nx, read 2026-10-01 mid-build; corrected the same hour from a first draft that looked only in nx_softtissue.nx) builds a second XPBD tet cage under the body whose per-vertex reaction shares form the map, with a summed-reaction-equals-load check (its header measures 1016 permil of the load carried on a rigid support once velocity is re-derived after contact, against 847 for the post-step order), indentation depth, patch area, a peak-to-mean ratio and a rigid-surface neg-control, publishing pressure as a shape in solver units with no kilogram and no pascal, and it is BELOW the triangulated best because it cannot express pressure in kPa or mmHg until SB4 lands, compares against no measured map, and reports no internal strain; its load balance and patch area are the right first teeth, the measured peak pressure, contact area and area above 40 mmHg are the comparison it must print once it has units, and a full map-to-map error against a measured mat would be a first among these witnesses (an opportunity, not a claim).
2026-10-01	SB13	measure	TRIANGULATED BAR as of 2026-10-01 for clothing on the body: physics with a guarantee is C-IPC, whose distance offsets strictly keep cloth mid-surfaces at least the applied thickness apart under a C2 strain-limiting barrier [cipc_2021]; real-time physics is OGC, which runs a C-IPC frame reported at 24 s in 0.24 s on CPU and 9.6 ms on GPU with cloth-cloth non-penetration guaranteed [ogc_2025]; learned garments repair rather than prevent intersection, ContourCraft recovering multi-layer outfits with an intersection contour loss and repulsion [contourcraft_2024] and HyperBones driving garments by virtual bones at about 1 ms per frame (2605.20460v4, 2026-09-04, read in [arxivsearch_clothsim_20261001]), and the newest on-subject September entry, found outside that phrase listing, is DiT-Garment, a UV-space diffusion transformer that generates garment deformation over a body in arbitrary motion conditioned on body motion and physical parameters and trained only on synthetic simulations (2609.18510, 2026-09-16) [ditgarment_2026]; measured material is Fabric-101, over 101 fabrics with elastic, friction-recoverable and plastic deformation separated by cyclic loading and fitted by an adjoint elasto-plastic friction simulator (2609.13707, 2026-09-12, accepted to SIGGRAPH Asia 2026) [fabric101_2026]; the listing has no October 2026 entry ; witnesses [cipc_2021] [ogc_2025] [contourcraft_2024] [fabric101_2026] [ditgarment_2026] [arxivsearch_clothsim_20261001] ; where they disagree: guarantee against speed, since C-IPC guarantees thickness and is slow, OGC is real time but states that body-cloth penetration might occur when the body is skeleton-driven, and the learned methods guarantee nothing and resolve intersections afterwards, the generative one returning a sampled distribution of outcomes rather than one determined state; and material, since the contact papers read here state no fabric plasticity or internal friction hysteresis while Fabric-101 measures both in real fabrics and the persistent wrinkles it reproduces are what an elastic-only model cannot keep; tearing is addressed by no witness read here, so it has no triangulated bar yet ; the estate's running lane SB13 (st_cloth_contact in buildroot/runtime/nx_cloth_contact.nx, read 2026-10-01 mid-build; corrected the same hour from a first draft that looked only in nx_softtissue.nx and wrongly called the lane unbounded and its fabric unmeasured) implements OGC's per-vertex conservative bound with truncation around gm_xpbd_step for vertex-facet and edge-edge contact, holds the tissue's own motion inside its bound so neither side can pass through the other, derives the contact distance from half the measured Fabric-101 thickness and splits layer pushes by measured areal density (knowledge/fabric101.conf, seven fabrics from Tables 11 to 13), leaves friction unset until a cloth-on-skin coefficient is cited, and tears at a strain threshold stated in solver units; on contact it is AT the real-time best (OGC) with declared deviations (hard projection in place of OGC's activation energy and VBD, no parallel-edge filter, feature de-duplication in place of the Gauss-map block test), and bounding a SIMULATED body is the case OGC states it leaves open, which becomes an exceed only once its gate (buildroot/runtime/nx_cloth_contact_gate.nx, 28,535 B, verdict not observed by this seat) proves zero cloth-body and cloth-cloth intersections on a moving cage; on material it is BELOW, because only thickness and areal density are measured, not stretch, bending or the friction-recoverable and plastic hysteresis Fabric-101 fits, and Fabric-101's published dataset is the liar-killer for those constants once the solver can take them.
2026-10-01	SB0	measure	OCTOBER ATTESTATION 2026-10-01: nx_barfresh softbody verdict FRESH, 3 of 3 bars ATTESTED. SB-TISSUE-2609 and SB-TISSUE-LOUNGER-2609 by nx_frontier_scan attest (job 1790822299: the arXiv listing read, 8 items, newest 2026-09-30, on-subject newest DiT-Garment 2609.18510 and the stress-free inverse FEM for soft tissue 2609.18109, nothing dated October); SB-SOLVER-2607 by the seat (the NVIDIA SIGGRAPH 2026 post published 2026-07-20 and modified 2026-07-21, nothing newer; the 17-source fan-out found nothing newer than July). RANK by nx_compare_rank (artefact sha256 f33d2936, target 1.0 milestone M2): critical path #1 SB11 contact that survives deep interleaving (st_sdf_contact, priority 225, DARK-CONTRACT), #2 SB7 tetrahedralisation of an arbitrary watertight surface with a sliver oracle (st_tet_from_surface, UNMAPPED), then SB4 Cauchy stress and strain pulled above SB5 the FEBio oracle, SB18 footfall impulse, SB15 support surface; PM filed 2 of the top 3. SB7 is the substrate aesthetictwin AT25 needs for the surgery chain
2026-10-01	SB15	land	REPOINTED: matrix row bedding-and-support-surface-load-distribution now names runtime/nx_support_surface.nx (was runtime/nx_softtissue.nx), contract symbol unchanged st_support_surface. The shared solver gets no line from this rung; the new file imports it and composes st_new, st_add_edge, st_substep, st_eff_damp, st_surface_spacing. Sources pinned on softbody.refs: the seated-buttock rigid-versus-foam FE study [sb15_buttockfat], the IPC barrier bar [sb15_ipc], the SIGGRAPH 2026 barrier-free penetration-free frontier [sb15_barrierfree].
2026-10-01	SB15	lesson	AN EDGE-ONLY LATTICE CANNOT CARRY A BODY: the cited cushion is a v=0 foam, so the first cut carried it as edges without the volume constraint, and under the glute profile it collapsed 86 mm through its own 48 mm pad with inverted tets (min 6V -3565297404; /tmp/nx_support_surface_gate_sb15_1.out, build a04b47f5) -- nothing resists inversion once the incumbent's hard volume constraint is withheld. The support moved to the incompressible gel lattice (the same lattice type as the body). Same run, the ORDER law measured: on a rigid support the post-step order carried 847 per mille of the load and the re-derived order 1016, against 1/(1+k) = 833 predicted by the skin damping row 819/4096.
2026-10-01	SB13	land	RESEARCH LANDED, every source fetched by nx_research_fetch and READ from nx_pdf_text output, never a summariser: Offset Geometric Contact, Chen Hsu Liu Macklin Yang Yuksel, SIGGRAPH 2025 (knowledge/fetched/cmp_softbody_ogc.pdf sha256 6abfd9e1cda6d541...) is the method built -- an orthogonal vertex-facet and edge-edge contact set and, in place of CCD, the per-vertex conservative bound of its section 3.7 b_v = gamma_p min(dmin_v, dEmin_v, dTmin_v) with 0 < gamma_p < 0.5 [ogc]; Fabric-101, Gong et al. 2026, arXiv 2609.13707 (cmp_softbody_fabric101.pdf sha256 4358dab11ca3cc05...) supplies MEASURED thickness and areal density, admitted as data rows in knowledge/fabric101.conf (KN01 cotton jersey 0.7 mm 174.34 gsm, C02 cotton poplin 0.2 mm 108.07 gsm, five more) [fabric101]; the newest penetration-free line, barrier-free augmented Lagrangian elastodynamics over CCD pairs, Zheng Luo Li, SIGGRAPH 2026, arXiv 2512.12151 (cmp_softbody_barrierfree.html sha256 f08bd629e8be61fa...), is named and NOT built [barrierfree]. Three refs rows on softbody.refs, cited inline on the matrix row.
2026-10-01	SB13	queue	GARMENT TENSION INDENTING SOFT TISSUE -- the string pressing into the hip in the operator's reference photographs -- is NOT claimed and is carried by SB13 as its open second leg, with SB15 (a support surface that spreads load) as the board's other load-into-tissue rung and SB4 (stress) for a pressure measure. What exists: every push in st_cloth_contact is already split by inverse mass, and the tissue surface is already held inside its own conservative bound so neither side passes through the other. What is missing, exactly: (1) the tissue side of the push is held at inverse mass 0 because no garment-to-tissue mass ratio is cited -- Fabric-101 supplies the garment's areal density, a mirrored soft-tissue density row with its source is the missing half; (2) the cloth and tissue constraints run STAGGERED (st_run, then the cloth rounds), and a taut band can only indent tissue when both are solved in one iteration loop. Next step: admit the tissue density row, derive the cage particles' contact inverse mass from it, move the tissue iterations inside the bound rounds, and gate a band whose depth into the tissue rises with its tension, with a slack-band neg-control reading zero depth.
2026-10-01	SB13	queue	WHAT OF OGC IS NOT BUILT, named so the hard-projection organ is never passed off as the whole method [ogc]: (1) the two-stage activation ENERGY (quadratic to a stitch at r/2, then logarithmic) and its Vertex Block Descent solve -- contact here is the hard-projection limit inside XPBD, so no contact stiffness is calibrated and no contact force is an energy gradient; (2) the bound is recomputed per ROUND (two per tick), where the paper recomputes it whenever a vertex reaches it inside the solver iterations; (3) the per-vertex polyhedral Gauss map block test is replaced by de-duplicating the contacted face, edge or vertex feature; (4) the smooth parallel edge-edge filter -- parallel pairs are skipped; (5) GPU parallelism. Next step: an energy-form OGC contact on the VBD lane, gated with THIS organ's penetration counts as its oracle.
2026-10-01	SB13	queue	THE SEPTEMBER FRONTIER FOR T-FR-SOFT IS NAMED AND NOT BUILT: barrier-free augmented Lagrangian elastodynamics over every CCD-found pair (2512.12151, SIGGRAPH 2026) [barrierfree] and Fabric-101's elasto-plastic-with-friction cloth that forms persistent wrinkles (2609.13707) [fabric101]. Both need an integer continuous-collision kernel (vertex-triangle and edge-edge time of impact) this estate does not have; OGC's bound was chosen tonight precisely because it needs none. No benchmark shared with any source was run: the numbers on this rung are the estate's own fixture. Next step: the CCD kernel as a shared primitive, then mirror one published scene (OGC's cloth on a rotating sphere) as a fixture and publish the penetration count and per-tick cost beside the paper's.
2026-10-01	SB13	land	st_cloth_contact LANDED in runtime/nx_cloth_contact.nx (LIB, sha256 166969fe51ad2f90, 79687 B), an organ of its own that IMPORTS nx_softtissue.nx and the PG29 cloth solver nx_gm_cloth_lib.nx and writes no second solver: every cloth constraint is still gm_xpbd_step, every tissue constraint still st_substep. Built to Offset Geometric Contact [ogc]: orthogonal vertex-facet and edge-edge contact against offset geometry with one push per contacted face, edge or vertex feature, and the section-3.7 conservative per-vertex bound applied to BOTH sides (cloth and the tissue surface) with gamma_p one half less the integer rounding, recomputed every round over a dense grid sized from the scene. Contact distance and the layer-to-layer mass split are DATA from Fabric-101 [fabric101] (knowledge/fabric101.conf sha256 3ff888586c377840). Tearing: a constraint fails when its strain passes a stated threshold IN SOLVER UNITS (SB4 has not landed), the row leaves the incumbent solver's list and its contact triangles become a hole. Friction: the per-256 positional Coulomb law, abstaining at ST_MU_UNSET. Gate nx_cloth_contact_gate (source sha256 4029946564cb1e25) PROMOTED sha256 f3d4645bb272c41d56e15607916d8311390cea86d582b25cfaec6f3b4da222ba (191569 B), registered, _offc installed. Matrix row cloth-to-body-and-multilayer-garment-contact REPOINTED from runtime/nx_softtissue.nx to runtime/nx_cloth_contact.nx (the shared tissue file was being extended by AT26 the same hour); its watch flips on the next regen, not by edit.
2026-10-01	SB13	measure	nx_cloth_contact_gate GREEN 15/15 on the build that was promoted (f3d4645b): ST_PROF_LARGE_SOFT pressed under SUPINE gravity for 120 substeps (apex 5900 to 5340 cmm), inner layer Fabric-101 KN01 cotton jersey held at both sides and lowered onto it, outer C02 cotton poplin dropped loose, 48 ticks at 60 Hz, 2 bound rounds x 8 iterations. ZERO garment vertices inside the cage on ANY tick (point-in-tet over all 4032 tets); brute-force closest approach inner-to-body 34 cmm against a contact distance of 35 (half of 0.7 mm), inner-to-outer 42 against 45 (0.35 + 0.10 mm), tolerance 4 cmm DERIVED from integer rounding; vertical-ray layer order 0 violations of 50; contacts 8169 body, 58267 layer, 32170 edge-edge; 12246 cloth and 20 tissue truncations; smallest bound 13 cmm; tissue moved at most 30 cmm per tick. NEG-CONTROLS, all BITE-PROVEN: body contact off puts 113 garment vertices inside the cage; the second layer ignoring the first gives separation 0 and 52 order violations; the OGC bound off lets a layer launched at twice the gap TUNNEL (95 violations) where with the bound it does not (0 of 94). TEARING: strain under its own weight 291 permil, under 4 g 632, stated threshold 461 permil (solver units, measured midway): 1 g leaves the sheet INTACT (0 failures at 291), 4 g fails 16 edges over 11 ticks, the first at the most-strained edge, with 12 later failures touching earlier ones -- the tear propagates from where the criterion put it. First version read 10/15: one projection pass per round left a stacked contact (body, inner, outer) Gauss-Seidel-unconverged at 26 and 33 cmm; iterating the contact set like any constraint set closed it.
2026-10-01	SB13	lesson	A CONTACT SET IS A CONSTRAINT SET: a stack of body, inner layer and outer layer projected in one pass leaves the last-processed pair satisfied and the first violated (measured 26 of 35 and 33 of 45 cmm), so contact needs the same iteration budget as the constraints. And the conservative bound is what separates a discrete projection from a penetration-free one: the same layer launched at twice the gap tunnels with the bound off and cannot with it on -- a gate that only drops cloth slowly never sees the difference.
2026-10-01	SB15	measure	nx_support_surface_gate 14/14 GREEN on build 26dd8c1c (job /tmp/nx_support_surface_gate_sb15_3.out): a glute body of 966 free particles at rest on a 48 mm gel pad carrying the solver row nearest the cited cushion-over-fat stiffness (5000 per mille against 4385 cited [sb15_buttockfat]). Summed contact reaction 1013 per mille of the applied load on the gel and 1017 on a rigid board, derived tolerance 29 (half a cmm against the 17.03 cmm gravity displacement per substep). The residual is ATTRIBUTED, not tolerated: the incumbent st_substep leaks 15 and 19 per mille of net internal momentum through its volume-constraint rounding, and reaction minus leak reads 998 on both supports, exactly the 17 over 17.03 the integer gravity step predicts. Gel indentation 650 cmm, closest approach 611 cmm = dhat on both supports by an all-face scan, no gel tet inverted, two runs bit-identical. The board concentrates the same load onto 83 body contacts against 91 on the gel and a body-side peak of 484 against 477 cmm per substep. Same comparison published [sb15_buttockfat]: contact area soft over rigid 1430 to 1887 per mille, peak rigid over soft 3466 to 7450; ours 1096 and 1014.
2026-10-01	SB15	land	st_support_surface SHIPPED in runtime/nx_support_surface.nx (lib, 43471 B, imports nx_softtissue.nx and adds no line to it): a body-vertex against support-face point-face contact with the barycentric inverse-mass split, recorded as a support-side pressure map (each face vertex share, what a sensor mat reads) and an alias-free body-side map, the summed reaction against the solver's own gravity increment, indentation, patch, an all-face interpenetration oracle, an inversion census and the solver leak. Velocity is re-derived after contact (constraints, then velocity, then damping) so the damper holds none of a resting body's weight. Gate nx_support_surface_gate PROMOTED sha 26dd8c1c 121674 B, registered, _offc twin installed, scratch row reaped and its binary retired reversibly. In-gate controls bite: the post-step order carries 848 per mille (fires), a run without the contact pass reads -11551 cmm penetration, refusals move nothing.
2026-10-01	SB15	queue	BARRIER CONTACT, NOT BUILT TONIGHT, AND EXACTLY WHAT IT ADDS. The bar is IPC [sb15_ipc]: a C2 log barrier on (0, dhat) inside a projected-Newton implicit step with a CCD-truncated line search; the September 2026 frontier keeps the CCD and the second-order solve and swaps the barrier for an augmented Lagrangian over every CCD-detected pair [sb15_barrierfree]. st_support_surface has neither: a hard ONE-SIDED projection (body vertex against support face) after an explicit XPBD substep. Three tests expose the gap and are the next done-rule: (1) a RIDGE (the lounger frame edge) pressed into the body, where support vertices slip between body vertices because nothing tests support vertex against body face or edge against edge, so an all-pairs scan in BOTH directions must read zero; (2) a cushion FOLDED over its own edge, which the plan-projection locate (a height field over the rest grid) cannot represent; (3) RIM FLICKER, rim contacts that change state between consecutive windows, which a C2 barrier force drives toward zero and a hard projection does not. Next step: derive the body cage surface triangulation from its Kuhn boundary faces, add vertex-face both ways plus edge-edge CCD, and solve a barrier energy per contact island.
2026-10-01	SB15	queue	THE PUBLISHED SHAPE IS NOT YET REACHED, AND THE MISSING PARTS ARE NAMED. Ours reads contact area 1096 and peak 1014 per mille (rigid against compliant) where the source reads 1430 to 1887 and 3466 to 7450 [sb15_buttockfat]. In order: (a) the source peak sits under the ischial tuberosity, a bone pressing through 6 to 30 mm of fat with 300 N of upper body on the sacrum, while our dome is all soft tissue under a free flat base; (b) our support is an incompressible gel where the source foam has v = 0, because the edge-only foam collapsed and st_substep solves volume hard -- a compressible cushion needs a COMPLIANT volume row with an inversion barrier at J to 0 in nx_softtissue.nx through its owner lane; (c) the cited stiffness ratio 4385 lands on the nearest profile row 5000 through an UNCALIBRATED lattice map -- a uniaxial compression tooth must measure each lattice's effective modulus, and a per-cage compliance row must replace profile rows (st_eff_alpha_dist reads profile rows only). Done-rule for the next pass: a bony prominence inside the glute cage on a calibrated foam, the two ratios inside the published band.
2026-10-01	SB15	queue	THE SHARED SOLVER LEAKS MOMENTUM AT REST, MEASURED BY THIS RUNG: st_substep's volume correction rounds each vertex's sub-cmm share separately, so near convergence the shares do not sum to zero and a resting glute body carries 15 to 19 per mille of its load as net internal momentum (sps_grav_disp against sps_disp_z, gate values gel_solver_leak_permil and rigid_solver_leak_permil). Owner lane nx_softtissue.nx: round the tet's summed correction and hand the remainder to the largest share (or carry sub-cmm fixed point) so internal forces sum to zero; then nx_support_surface_gate's reaction reads 998 per mille with no attribution needed, and every other contact readback on this board loses the same bias.
2026-10-01	SB13	measure	THE PROMOTED BINARY RE-RUN FROM THE LIVE PATH (nx_job_run nx_cloth_contact_gate, nishihost/nx_cloth_contact_gate.elf sha256 f3d4645b): GREEN 15/15 with every emitted value identical to the staged run (output 4261 B both times, inside_max_no_body 113 both times) -- the integer solver is deterministic. Scratch registry row nx_cloth_contact_gate_sb13_staged REAPED and its staged file retired reversibly to knowledge/retired/; a cap naming nx_cloth_contact_gate minted (private response) so the gate is callable, not REGISTERED-UNAUTHORISED. Fleet referee nx_compare_refs_gate names none of the three new refs rows (ogc, fabric101, barrierfree) in any finding; its 40/45 RED is the pre-existing redirect and moved ratchets.
2026-10-01	SB13	queue	SHIPPED, NOT COMPLETE: gate GREEN 15/15 is the mechanical receipt only. Completion needs the referee panel and the operator's ACCEPT row; no render of the garment over the body was produced or judged tonight, so how the cloth LOOKS sitting on the body is UNMEASURED. Next step: render the gate's final drape (KN01 band over the pressed cage, C02 over it) through the beach-twin render lane and put it before the referee panel; persistent wrinkles (Fabric-101's elasto-plastic-friction model) are the visual half the frontier target T-FR-SOFT still asks for.
2026-10-01	SB13	land	THE TISSUE HALF OF THE MASS RATIO IS NOW DATA, so the two-way leg is no longer blocked on a citation: NIST's ICRP entries read from the stored bytes -- ADIPOSE TISSUE (ICRP) Density (g/cm3) = 9.20000E-01 [nist_adipose] and TISSUE, SOFT (ICRP) 1.00000E+00 [nist_softtissue] -- admitted as knowledge/tissue_density.conf rows (mg per cm3). Derived, not typed: an adipose lattice particle of the 10 mm cage carries 0.920 g against a KN01 cotton jersey vertex at 15 mm spacing of 0.0392 g (174.34 gsm times 225 mm2), a garment-to-tissue mass ratio of 1 to 23.5. NOT wired tonight: a staggered push at that ratio would move the tissue by an iteration-dependent sliver per round, which is a mechanism, not the monolithic cloth-plus-tissue solve the indentation needs; shipping it as garment tension indenting skin would be a lesser method wearing the claim.
2026-10-01	SB15	queue	THE FE PATH TO A kPa COMPARISON, NAMED BY FUNCTION. AT26's hyperelastic tier landed in nx_softtissue.nx the same evening (gate nx_softtissue_hyper_gate 28/28) and takes FEBio-convention Ogden rows verbatim (st_hyper_material_ogden), the convention the source states for fat (mu1 1166.7 Pa, alpha 16.2) and muscle (1907.4 Pa, 4.6) [sb15_buttockfat]. Composition: a fat-over-muscle block (st_hyper_block, st_hyper_assign_box), a rigid sphere for the ischial tuberosity (st_hyper_set_implant with contact candidates from st_hyper_contact_plane and the augmented Lagrangian on via st_hyper_set_al, the formulation family of the September 2026 frontier [sb15_barrierfree]), the seat as a Dirichlet plane (rigid) or a neo-Hookean layer at E = 15.3 kPa with K set for v = 0 (foam), quasi-static st_hyper_solve, and the seat-face pressure map read from element internal forces at the seat nodes (st_hyper_element) in pascals. Done-rule: peak contact pressure on the rigid seat inside the source's 0.052 to 0.149 MPa across fat 6 to 30 mm and inside 0.015 to 0.020 MPa on the foam, internal tissue stress read by st_hyper_cauchy beside it (the end-point the 2026-10-01 bar row says injury follows).
2026-10-01	SB15	lesson	A FACE-SIDE SENSOR COUNT IS ALIASED, AND IT ONCE READ THE GEL WITH A SMALLER PATCH THAN THE BOARD. One body contact lights one support-face vertex when it sits on it and three when it sits mid-face, so the face-side patch read the gel 125 against the board 127 while the body side, a regular lattice where each contact counts once, read 91 against 83 and its peak 477 against 484 (build bb8fa3b9, /tmp/nx_support_surface_gate_sb15_2.out). The patch and peak teeth read the body side; the face-side map stays published as the sensor-mat deliverable. A sensel-integrated face map (each cell's share over its own area) or a participation-ratio effective area would remove the aliasing on the face side too, and is the cheap next refinement.
2026-10-01	SB13	queue	FRICTION OF A GARMENT ON SKIN AND ON ITSELF IS WIRED AND ABSTAINING: st_cloth_contact carries the estate's per-256 positional Coulomb law on every vertex-facet push, but no cloth-on-skin or cloth-on-cloth coefficient is cited (Fabric-101's friction is internal yarn hysteresis, not a contact coefficient), so the scene defaults to ST_MU_UNSET and the friction path has NEVER RUN on a real coefficient. Next step: mirror a measured textile-on-skin and textile-on-textile friction study, admit its coefficients beside knowledge/fabric101.conf, and add a tooth where the outer layer on a tilted inner layer stays put above the cited coefficient's angle and slides below it, with mu UNSET as the sliding neg-control.
2026-10-01	SB13	measure	INDEPENDENT RUNNER AGREES: /api/gate_run on the promoted nx_cloth_contact_gate returned exit_code=0 verdict=GREEN (receipt _jobs/job_1790826259.out, drained inline after outrunning the 14750 ms reply window); thirteenth evidence row appended to knowledge/compare/softbody.gates. nx_ladder softbody still LADDERED (20 rungs placed, SB13 contender for T-FR-SOFT).
2026-10-01	SB13	queue	CLOTH AGAINST ITSELF IS WIRED BUT NOT SEPARATELY GATED: same-layer pairs that share no vertex already carry both the conservative bound and the push (the rung's self-contact half), but tonight's fixtures never fold a layer onto itself, so a self-separation tooth would be vacuous and none was claimed. Next step: let ctc_scene_new take initial positions so a sheet can be built PRE-FOLDED with its rest state measured folded, drape it over the pressed cage, and gate its self-separation against its own thickness with CT_FL_NO_SELF as the neg-control that must interpenetrate.
2026-10-01	SB15	land	RE-PROMOTED against the shared solver as AT26 left it: AT26 landed the hyperelastic tier into nx_softtissue.nx (78002 to 136681 B) after this rung's first promote, so the gate was rebuilt against the new closure (debug flavour from the incumbent twin, 140032 B, sha 0428c8f0), run 14/14 GREEN (/tmp/nx_support_surface_gate_sb15_4.out) with every value AND the determinism hash -6676300930262145470 bit-identical to the run on the old closure, content ruler lost_from_live=0, and promoted over 26dd8c1c with expect_sha256; _offc twin refreshed, scratch registration reaped, scratch binary retired reversibly. So the shared-file edit is measured behaviour-neutral for st_substep, not assumed.
2026-10-01	SB15	measure	COMPLIANCE SWEEP, PRE-DECLARED AND PARTLY REFUTED, NOT SHIPPED (/tmp/nx_support_surface_gate_sb15_5.out, build 142256 B, 16/17 RED). A third support carried the body's own fat row, so the three supports formed the sweep the source itself runs [sb15_buttockfat]. Body contacts ARE monotone in compliance: board 83, cited-row gel 91, tissue-matched gel 99 (area ratio 1192 per mille against the board, indentation 650 then 1131 cmm). The body-side peak is NOT: 484, 477, 483 -- the pre-declared tooth demanded it fall and it did not. The tissue-matched run balances exactly (reaction 1000 per mille, solver leak 2) and keeps the gap and every tet, but its summed body speed at the window is 3560093 against 1677468 on the cited gel, so it is not demonstrably at rest. The tooth was NOT loosened to pass: the gate source was reverted to the promoted 14/14 version, and the sweep is queued with the instrument it lacks.
2026-10-01	SB15	queue	THE SWEEP NEEDS A MEASURED REST, NOT A FIXED SETTLE. Three damping periods settle the cited-row gel but leave the tissue-matched gel moving (summed body speed 2.1x), and a peak read off a moving body is not a resting peak. Next step: settle by MEASUREMENT -- run windows of one damping period until two consecutive windows agree on the reaction within the derived tolerance AND the summed body speed stops falling, under a stated budget, and print the windows used; then re-run the three-support sweep and re-declare the same tooth (contacts rise and the body-side peak falls with compliance). If the peak still does not fall at rest, that is a finding about a bonded incompressible pad, whose compliance comes from lateral flow, and it is published as such rather than tuned away.
2026-10-01	SB15	land	BUILD FOSSIL RESTORED, DECLARED RATHER THAN LEFT AS A LOADED GUN: the unshipped sweep build (142256 B) had overwritten buildroot/_build/nx_support_surface_gate.sov.elf, which a drift census would read as BEHIND the promoted gate and offer to promote, and it is RED. The gate source was reverted to 15760 B (zero sweep remnants by txtscan), the confirming rebuild was refused by admission six times at load 17 to 20, so the fossil was restored from the promoted bytes (nx_filecopy verified=1; nx_filehash cmp identical 0428c8f0 140032 B). Open confirmation: one rebuild under admission, expected as a build-cache hit on key 64dcb60b, which stored exactly 0428c8f0.
2026-10-01	SB15	measure	CONFIRMED, the open item of the row above closed: the reverted gate source keys the build cache at 64dcb60b, the exact key that stored 0428c8f0, and the admitted rebuild was a cache hit verified against its sidecar; nx_filehash cmp reads _build and the promoted gate identical (0428c8f0, 140032 B). So the source on disk, the build fossil and the served gate are one artifact, proven rather than declared.
2026-10-01	SB13	land	CLOTH AGAINST ITSELF IS NOW GATED, superseding the 1790826581 queue row: st_cloth_rest_from_current (a garment folded, gathered or sewn takes that shape as its rest state) added to runtime/nx_cloth_contact.nx (sha256 a655441bddfb2351, 80329 B); nx_cloth_contact_gate grown to a FOLDED KN01 layer draped over the pressed cage (source sha256 85b3057a394f8493, 32681 B) and RE-PROMOTED sha256 ceaaa23dbc4f2c887f296058a06e0c3c693891d10b9ae2728649ae3eba255e20 (223272 B, debug flavour as the builder chose from the incumbent), after nx_contentdiff live-vs-candidate GREEN lost_from_live=0 ONE-WAY-GAIN; _offc reinstalled; scratch rows nx_cloth_contact_gate_sb13b_staged reaped and its file retired reversibly.
2026-10-01	SB13	measure	nx_cloth_contact_gate GREEN 18/18 on the re-promoted build (run /tmp/nx_cloth_contact_gate_sb13b_1.out, every earlier value reproduced exactly): a KN01 jersey layer folded in half with its top half two self-contact distances (140 cmm) above the bottom, draped 48 ticks over the cage pressed under supine gravity -- 28036 self contacts and 5684 body contacts fired, ZERO garment vertices inside the cage, brute-force self-separation of the folded halves 66 cmm against the fabric's own measured thickness of 70 (tolerance 4, so at the edge of the derived bound and named as such); NEG-CONTROL BITE-PROVEN: with self-contact off the halves pass through each other to a self-separation of 0.
2026-10-01	SB13	measure	INDEPENDENT RUNNER ON THE RE-PROMOTED BUILD: /api/gate_run nx_cloth_contact_gate (ceaaa23d) was promoted to the job lane and reaped state=DONE rc=0 verdict=GREEN (_jobs/job_1790827948.claim, 5030 B output, the same size as the staged run), source_newer=0; the evidence row in knowledge/compare/softbody.gates now reads 18 of 18 with both receipts. Every scratch registry row from this lane is reaped and both scratch stage files are retired reversibly to knowledge/retired/.
2026-10-01	SB15	measure	BACKFILL-HUMAN (d) a simulation validated against a measured interface-pressure map: THREE independent groups now counted, each read in full text: Tel Aviv (Katz and Gefen 2023, existing row katzgefen_2023); Lyon LBMC with Beihang (PIPER adult comfort 2023, refs sb15_piper_lyon_beihang_2023, mirror knowledge/fetched/cmp_softbody_bfh_piper_comfort_2023.xml sha256 3d2b93ba): a subject-specific FE model against the same subject's mats at four seat-pan angles, seat-pan contact area, peak and mean pressure off by 5.8, -6.4 and 1.2 percent, the four-region pressure shares off by 5.2, -1.4, -13.9 and -3.1 percent, backrest peak off by 62.1 percent; Hunan University (Yan 2025, refs sb15_yan_hunan_2025, mirror knowledge/fetched/cmp_softbody_bfh_yan2025_seatangle.xml sha256 193a3cdc): four 50th-percentile males on capacitive mats, back peak +20 percent, hip and leg contact area -2 percent, pressure-point counts per 4 kPa bin compared. Independent count 3. The critic's finding STANDS: none of the three publishes a cell-by-cell map-to-map error; the closest are a four-region share (PIPER) and a 4 kPa histogram (Yan), so a map-to-map metric is still an exceed candidate. Examined and excluded: an ejection-seat cushion study (PMC13057320) that measured pressure but left coupled human-cushion FE to future work (mirror knowledge/fetched/cmp_softbody_bfh_ejectseat_2026.xml); search listing knowledge/fetched/cmp_softbody_bfh_epmc_fepressure_search.json
2026-10-01	SB13	measure	BACKFILL-HUMAN (e) measured fabric material and tearing: the critic's abstract-only note on Fabric-101 is STALE, this board's row fabric101 was read in full through nx_pdf_text on 2026-09-30 (knowledge/fetched/cmp_softbody_fabric101.pdf). A second independent measured-fabric group now: SEDDI with Universidad Rey Juan Carlos and Universidad Carlos III, Madrid (Rodriguez-Pardo 2023 Eurographics, refs sb13_rodriguezpardo_seddi_2023, mirror knowledge/fetched/cmp_softbody_bfh_drapelike_2304_06704.pdf sha256 1a118500, full text read through nx_pdf_text: real fabrics with ground-truth stretch, bending and density parameters, a drape metric with 86.68 percent user agreement). Measured-fabric independent count 2, SHORT of 3. Tearing, was 0 witnesses, now 1 weak one: a tear-test PROCESS study (Huang 2024 PLoS One, Zhejiang standards and inspection institutes, refs sb13_huang_zhejiang_2024_teartest, sha256 12682c12) that measures the test, not a fracture energy or a tear-propagation law; an arXiv listing of tearing AND cloth or fabric or textile or garment, 30 newest (knowledge/fetched/cmp_softbody_bfh_arxiv_tearing_listing.xml), held 0 on-subject cloth-tearing papers, every hit a homonym (graphene tearing, fabrication). Not obtained: Pfaff, Narain, de Joya and O'Brien 2014 Adaptive Tearing and Cracking of Thin Sheets (graphics.berkeley.edu): plain http refused status 0, https failed in the sovereign TLS cert pipeline rc -207; unblock is cert-chain support for that host or a library copy
2026-10-01	SB15	land	BACKFILL-HUMAN refs rows landed this lane in softbody.refs, each mirror pinned by nx_filehash, cited here so the refs referee's inline-citation tooth sees them: [sb15_piper_lyon_beihang_2023] [sb15_yan_hunan_2025] for SB15, and for SB13 [sb13_rodriguezpardo_seddi_2023] [sb13_huang_zhejiang_2024_teartest]
2026-10-04	SB11	measure	METHOD READ THIS MONTH for contact that survives deep interleaving: Shortest Path to Boundary for Self-Intersecting Meshes, He Chen, Elie Diaz and Cem Yuksel, ACM TOG 42(4), SIGGRAPH 2023, arXiv 2305.09778 [spb_2023], read 2026-10-03 from nx_pdf_text of the stored 21.2 MB PDF (knowledge/fetched/cmp_softbody_spb2023.pdf, sha256 dbbb164e). The exact shortest VALID path from an interior point to a boundary that may overlap itself: Theorem 1 (the path is the shortest valid line segment), Definition 4 (valid iff an element traversal carries it), Theorem 2 (one candidate per boundary face, its Euclidean closest point, tested in increasing distance), sec. 3.3 (topological ray traversal that branches through edges and vertices within a face tolerance, a visited list, a candidate stack, no premature termination) and sec. 5.4 (infeasible-region culling, identical results). It is the RECOVERY half of the 2026-10-01 triangulated bar: the barrier-free augmented Lagrangian [barrierfree_al_2026], GIPC [gipc_2024] and Offset Geometric Contact [ogc_2025] all keep an intersection-free trajectory from an intersection-free start, and OGC itself says volumetric meshes may resolve intersections after they occur -- a tangle that already exists needs this query.
2026-10-04	SB11	lesson	THE BLIND CASE ON THIS RUNG'S OWN ROW WAS GRAZING, NOT A TANGLE: two settled LARGE_SOFT cages placed touching and pushed 600 cmm past it left an INDEPENDENT point-in-tetrahedron census at 0 both ways while SB10 still saw 9 pairs -- the surface particles slip between the other cage's staircase cells. A fixture that never reached the condition proves nothing, so nx_sdf_contact_gate pushes in half-spacing steps until that census first reads non-zero (2994 cmm in practice, two mesh spacings past the crossing), and there SB10 leaves the pair interleaved after all 24 of its passes (census 122 before, 124 after). A ONE-SHOT PUSH IS AN INVERSION GENERATOR: pushing a 1.3 mm penetration in one step plus a dx/dt velocity injection inverted a tetrahedron of the free cage within one round; the fix is a trust region (no particle moves more than half the smallest rest altitude of its own tetrahedra in one pass, the per-vertex bound of Offset Geometric Contact read as an inversion bound) and an inelastic normal impulse that removes only approaching relative velocity, so penetration depth never becomes kinetic energy. THE SOLVER'S SURFACE FLAG IS NOT THE MESH BOUNDARY: it omits the pinned chest-wall layer (352 boundary vertices of one LARGE_SOFT cage carry srf 0), and a contact keyed on it left wall particles inside the other cage; the subject is every vertex of a boundary face, read from topology.
2026-10-04	SB11	measure	THE LOADED CASE, MEASURED AND NOT CLAIMED: two LARGE_SOFT domes each pinned to its own wall, pressed tip to tip 2994 cmm past touching (122 particles interleaved). With one capped contact pass per full 8-iteration step the deepest penetration a round met went 2518, 1938 (round 8), 950 (round 32) and then PLATEAUED near 965 cmm for 224 more rounds; forcing it clear with nine extra passes cleared the census but inverted a tetrahedron. With SMALL STEPS (the same budget as 8 substeps of one iteration, a contact pass after each) it went 2477, 822 (round 2), 329 (round 4) and held 312 to 402 cmm through round 64 with no inversion, 20 particles still inside. Reading: under a sustained load the tissue recoils about 300 cmm every one-iteration substep, more than one inversion-safe pass can move, so a POST-STEP projection always leaves a residual -- the source's Algorithm 5 projects collisions INSIDE the solver iterations, which needs a hook in st_substep (queued). The done-rule fixture is therefore the free pair, where nothing forces the bodies back together.
2026-10-04	SB11	queue	WHAT OF THE SOURCE IS NOT BUILT, named so the routine is never passed off as the whole method [spb_2023]: (1) the in-solver projection of Algorithm 5 -- the contact runs AFTER each st_substep, so a sustained load leaves the measured residual; the remedy is a contact hook inside st_substep's iteration loop (nx_softtissue.nx is shared with AT26, so it lands through that lib's owner, never as a second solver); (2) the edge and vertex infeasible-region culls (only the face-interior cull is built; the traversal still decides those candidates, so results are identical and only cost differs); (3) inverted elements and the backward ray of sec. 3.5 -- a traversal into an inverted tetrahedron is counted and that branch dies; (4) the CONTACT response for a cage folding into ITSELF needs the incident-element exclusion -- the query already answers a self-overlapping mesh (the gate's folded C); (5) the BVH candidate order -- every boundary face's closest point is computed per query, exact but O(faces); (6) the CCD-first hybrid the paper recommends, which here is SB10's proximity contact for the shallow case. Next step: (1), measured on the loaded fixture, accepting when the census reads zero after the last substep of a pressed pair.

On these two registers. Rows are declared in the domain's plan file and carry the debt id, which is the join key back to the sovereign debt plane — that plane, not this page, is the authority on state. Reconciling them automatically (the regen reading the plane and refreshing these rows) is a named, owed rung; until it lands, treat an id here as a pointer to look up, not a status to trust.

Honest verdict. Their constitutive fidelity is higher offline and that is stated plainly on every row it applies to. Ours is the only one of the five that validates its resulting MOTION against a measured human, publishes the derived figures on the page they drive, and REFUSES the build when code and citation drift apart. It is also the only one that runs in a browser tab with bit-identical replay. Stress tensors, self-contact and force readback are open, named, and block every clinical claim until they land.

Person · product · place — not yet measured for this domain

Every compare carries this layer. Declare knowledge/compare/softbody.ppp (rows surface|nishi or c1..c4|label|url|connect naming OUR live surface and each rival's front door), run nx_ppp_probe domain softbody, and this section fills itself on the next beat: the same ruler on both sides — privacy and CX (third-party hosts, tracker classes, cookies, security headers), design and longevity (design hygiene, computed WCAG contrast, render-blocking resources, unsized media, script weight, theme and motion queries), findability (landmarks, skip link, on-site search, breadcrumb, headings, internal links).

Mined from the field — review and competitive intelligence proposed to this board

Rows written by nx_intelmine_propose from nx_reviewmine_lib signals (Steam reviews under the exceeds / meets / mixed / does-not-meet rubric, the installed-title census) and routed here by capability_map.conf. DEFECT names a rival failing that a rung here answers; DEMANDED a capability buyers ask for and do not get; WE-DO-BETTER an exceed the reviews corroborate; THEY-DO-WELL a bar this board must meet; SHIPS a rival capability to match. A proposal is a lead, never a rung — it closes only when a rung with a gate lands, and nothing in this band changes a measured cell.

Signal	Kind	Term	Title	Proposed rung	Evidence
WE-DO-BETTER	feature	physics	DanceXR 1905510	Physics believability: DanceXR -- measured against the field's physics with a referee panel	praised=1 complained=1 mixed=0 demanded=0

proposals 1|malformed rows 0 (counted, never rendered: a row that is not prop| with 10 fields)|read-capped 0

The field — discovered, not chosen

Rows written by nx_field_discover from softbody.seeds: the industry's own lists (Wikipedia wikitext, GitHub topics, awesome lists) read mechanically, every candidate counted across seeds. The matrix columns above are a SEAT'S pick; this band is the population they were picked from, and the stats line measures one against the other. A rival here is a lead, never a verdict — it earns a column when its capabilities are read and pinned.

Discovery receipt. One row per counter the run wrote, in the receipt's own order; the rival rows follow in their own table.

Field discovery, counted
Class	Count	Note
seeds	4
scanned	4
fetched	4
reused	0
failed	0
named	4
candidates	4
mentions	4
capped	0

Rank	Rival	Seeds	Mentions	First seed	Kind	Link
1	SOFA	1	1	col0-sofa	named	https://www.sofa-framework.org/
2	Project Chrono	1	1	col1-project	named	https://projectchrono.org/
3	Bullet	1	1	col2-bullet	named	https://github.com/bulletphysics/bullet3
4	FEBio	1	1	col3-febio	named	https://febio.org/

Exclusive and shared capabilities

One capability is graded on one board; every other board references it. The fleet duplicate census (nx_rivalcensus dupes, on the fleetdupes clock row) reads every board's matrix and names each capability graded as a row on two or more boards, with a home derived from evidence: rung when exactly one board's plan contracts the symbol (that board is starred), gates when no plan contracts it and exactly one board's executed gate map runs the organ's gate (that board is plussed), ambiguous when several boards carry the deciding evidence (a decision for their seats), and none when no evidence decides it, with the reason beside it: unmapped (no board of the duplicate declares a gates map; declare the maps) or unproven (a map exists and runs none of the organ's gates; gate the organ on its proving board, or declare it a shared library by policy). A home elsewhere means this board should carry an @xrow reference to the home board's row rather than a graded row of its own; a home here means the other boards should. The table is this board's slice of one estate-wide census: the shape is shared, the rows are the board's.

This board grades 32 capabilities: 26 exclusive to it and 6 shared with other boards. Shared rows on this page: 6 = home here 0 + home elsewhere 1 + ambiguous 3 + undecided 2 + refused 0 (partition 6); undecided = unmapped 0 + unproven 2 + unevidenced 0; homonyms 0; since the previous census: new 0, resolved 0. Census FRESH age_s=6864.

Capability	Organ	Boards grading it	Home	Rule	Evidence	Standing on this board
sd_step	`nx_softdyn.nx`	graphics, procgen, softbody, dcc, gameengine	—	none	unproven	undecided
md_measure	`nx_mmdev_lib.nx`	procgen*+, softbody+, dcc, anatomy+	procgen	rung	plan	home elsewhere adopt: `@xrow procgen\|md_measure\|home by rung (fleet duplicate census)`
cg_bulge	`nx_softtissue_conserve_gate.nx`	procgen+, softbody+	—	ambiguous	gates	ambiguous
st_substep	`nx_softtissue.nx`	softbody+, hapticcare+, dcc	—	ambiguous	gates	ambiguous
ce_ship	`nx_craft_emit.nx`	softbody, gameengine	—	none	unproven	undecided
st_contact_force	`nx_softtissue.nx`	softbody+, hapticcare+	—	ambiguous	plan	ambiguous

Gaps from the record — what the estate does that no board carries

The record census (nx_goalmap record) reads the invoked-tool population and every plan queue row and files each organ or directive that NO matrix, plan or gates row names. A row here is a callout the boards missed: adjudicate it onto a board or declare it infrastructure. Census state STALE (age 1395969 s), sources read 4 of 7 declared — a BLIND census is a FLOOR: unread sources can only add rows.

kind	name	board	source	evidence

rows shown 0|this board's directives 0|estate-wide un-boarded organs 452 (listed in full on /compare/ecosystem)|census rows 639|malformed 0 (counted, never rendered)

References

Beyond a link list. Every reference below resolves twice — the publisher's copy and, where banked, the estate's own non-rottable library mirror with a content pin — and carries its evidence class plus the exact claim on this page it grounds. Keyed marks like [key] in the matrix notes jump here. A dash means honestly absent, never assumed.

[sofa] SOFA Framework -- open-source framework for interactive real-time simulation, targeted at medical and soft-tissue mechanics publisher · read in our library knowledge/fetched/cmp_softbody_sofa.html · pin hc28d11ec6accec23055c395ce343f57d3a8c606aa9c9e5127b63857ffb9f5293 · accessed 2026-08-25 · vendor-docGrounds: xpbd-tet-cage-layered-compliance
[chrono] Project Chrono -- open-source multi-physics simulation engine covering multibody dynamics, FEA and deformable terrain publisher · read in our library knowledge/fetched/cmp_softbody_chrono.html · pin h90921eecf7cd5f29ab48fd1d7e40ebde4dfb9d019ce16819d95a2a3ad350ef47 · accessed 2026-08-25 · vendor-docGrounds: environment-soft-body
[bullet] Bullet Physics SDK -- real-time collision detection and multi-physics for games, robotics and VR publisher · read in our library knowledge/fetched/cmp_softbody_bullet.html · pin h52245668d2c934aecc050f13d2a365194447e07c8c8ac4f6e883b52fc898b165 · accessed 2026-08-25 · source-readGrounds: single-process-real-time-in-browser
[febio] FEBio -- finite element solver purpose-built for biomechanics and the analytical reference other solvers are checked against publisher · read in our library knowledge/fetched/cmp_softbody_febio.html · pin had8b6691342716904e07ceebd6976d8579a2ce6e8416cd540a57a48cfd999ca7 · accessed 2026-08-25 · vendor-docGrounds: febio-analytical-oracle
[skinfriction] Zhu, Y. et al. Characterization of Skin Friction Coefficient, and Relationship to Stratum Corneum Hydration in a Normal Chinese Population. Skin Pharmacol Physiol (PMC2997446). publisher · read in our library knowledge/fetched/cmp_softbody_skinfriction.html · pin h3c36520e1c4c7df03ba63ca61ea7c93621f8d3abd7f86adc97616c9bd831518e · accessed 2026-08-25 · published-paperGrounds: skin-friction-coefficient-sourcing
[skinmicro] Masen, M. et al. Skin Microstructure is a Key Contributor to Its Friction Behaviour. Tribol Lett (PMC6961497). publisher · read in our library knowledge/fetched/cmp_softbody_skinmicro.html · pin h1b0be509e179cca9a0670911dd37df1f3041cb081c718f3f0a80aada7223f4b7 · accessed 2026-08-25 · published-paperGrounds: skin-friction-coefficient-sourcing
[hairams] He, Amador Herrera et al. Digital Salon -- An AI and Physics-Driven Tool for 3D Hair Grooming and Simulation. arXiv 2507.07387, Yale University. Adapts the Augmented Mass-Spring Model of Herrera et al. 2024, modelling standard edge, bending and torsional degrees of freedom while incorporating an augmented biphasic interaction through two additional one-way springs, embedded within a hybrid Eulerian and Lagrangian framework to resolve hair-hair interactions, simulating thousands of interacting strands in real time. publisher · read in our library knowledge/fetched/cmp_softbody_hairmassspring.html · pin h3465ee22cda8931b329212fa816d387cdfa5147e4aef995215f25aee164e440f · accessed 2026-08-25 · published-paperGrounds: hair-strand-to-body-collision
[ziva] DNEG Ziva. Ziva VFX constructs and simulates muscles, fat, fascia and skin for digital characters and produces precise skin mesh deformations informed by their underlying anatomy, and its team received a 2025 Scientific and Technical Achievements Academy Award. Ziva RealTime analyses training sets that demonstrate how a character should look as it moves and learns to reconstruct that shape at realtime speeds, reproducing in milliseconds effects that traditionally require seconds or even minutes per frame. Ziva Face Technology converts scans and models into digital Face Puppets. publisher · read in our library knowledge/fetched/cmp_softbody_ziva.html · pin hd14d80cdb8b593d22d557474a2009e0bc74ec5bbbcfbe16d05a6a1206fc0be6c · accessed 2026-08-25 · vendor-docGrounds: muscle-contraction-driven-skin
[metahuman] Epic Games MetaHuman. Create and animate a bespoke photoreal digital human for any 3D project, complete with hair and clothing, in minutes, starting from a preset, with MetaHuman Animator for real-time performance, delivered inside Unreal Engine. The mirror is a marketing front door and is cited ONLY for that positioning -- it does not describe vasodilation, subsurface scattering or micro-expression internals and no such claim is sourced to it here. publisher · read in our library knowledge/fetched/cmp_softbody_metahuman.html · pin he316f3a924da268e4ab9e7cf1378661948775161609478118013e224919d0b8f · accessed 2026-09-05 · vendor-docGrounds: muscle-contraction-driven-skin
[uegroom] Epic Games. Hair Rendering and Simulation, Unreal Engine 5.8 documentation. The groom rendering and simulation system uses a strand-based workflow to render each individual strand of hair with physically accurate motion, enabling artists to simulate and render hundreds of thousands of strands. publisher · read in our library knowledge/fetched/cmp_softbody_groom.html · pin h536c86460e0cde8d254aa9a334690adaa37f5296862924c0b6d74b52da9fbfda · accessed 2026-08-25 · vendor-docGrounds: hair-strand-to-body-collision
[chaoscloth] Epic Games. Clothing Tool in Unreal Engine. Unreal Engine uses the Chaos Cloth solver, a low-level clothing solver responsible for the particle simulation that runs clothing; clothing is painted onto a material element of a skeletal mesh with the Cloth Paint Tool, and the character's Physics Asset supplies the collision the cloth simulation uses. publisher · read in our library knowledge/fetched/cmp_softbody_chaoscloth.html · pin h5aae6901e21688b8e88e0c880c9a5186e855614fe4a7bf99b180c5b1b5a8dd33 · accessed 2026-08-25 · vendor-docGrounds: cloth-to-body-and-multilayer-garment-contact
[liquigen] JangaFX LiquiGen. A real-time liquid simulation tool for creating anything from water and blood to ketchup and slime, with real-time meshing, customizable forces, sparse simulation domains and a real-time path tracer. publisher · read in our library knowledge/fetched/cmp_softbody_liquigen.html · pin hf8ad73c9caec09208159423caf38dffedf77a8803c3478096c71cdbb311eeaf4 · accessed 2026-08-25 · vendor-docGrounds: viscous-film-surface-tension-and-clinging
[marvelous] Marvelous Designer. A digital cloth creation tool for 3D artists positioned by its vendor as providing the most accurate and fastest cloth simulation, built on garment pattern making with a Solidify property assignable to parts of a garment. publisher · read in our library knowledge/fetched/cmp_softbody_marvelous.html · pin h30e9779e5716db6fbdf8cf2ea79aff8ab650b3f0170fac15945ec6ed4e717b02 · accessed 2026-08-25 · vendor-docGrounds: cloth-to-body-and-multilayer-garment-contact
[clo3d] CLO Virtual Fashion. CLO is a 3D fashion design software program creating virtual true-to-life garment visualization with simulation technology, covering 3D fashion design, garment simulation, digital patternmaking and virtual garment visualization, founded 2009. publisher · read in our library knowledge/fetched/cmp_softbody_clo3d.html · pin h7b52b70f4c2018272ff2671d861f29503a1498b0a1102e439695ab702239d861 · accessed 2026-08-25 · vendor-docGrounds: cloth-to-body-and-multilayer-garment-contact
[arxiv_softbody_202609] arXiv API listing, (soft tissue OR cloth simulation OR garment OR soft body) AND simulation, newest first, 20 entries, read 2026-09-16: persistent wrinkle formation with Fabric-101 2609.13707, occupancy-guided soft-tissue resection 2609.16186 publisher · read in our library knowledge/fetched/cmp_softbody_arxiv_202609.xml · pin h2aa5eccc1578ea8a629657f5b25d80292435e2f42c8aee792f61825afae0829d · accessed 2026-09-16 · mirrored-indexGrounds: SB-TISSUE-2609
[nvidia_siggraph2026_blog] NVIDIA's SIGGRAPH 2026 announcement, mirrored 2026-09-27: the Newton physics engine gains new solvers for snow, sand and elastic solids; MotionBricks, GPC and ArtiFixer beside it publisher · read in our library knowledge/fetched/cmp_graphics_nvidia_siggraph2026.html · pin hfe0809948d5a5de467f3e88fdd7da457e01f254131164537a6c733365d0e1a45 · accessed 2026-09-27 · vendor-docGrounds: SB-SOLVER-2607
[arxiv_co_siga2026] arXiv API venue-tag scan, comment phrase SIGGRAPH Asia 2026, 55 entries, newest 2026-09-24, read 2026-09-27: S4R scaling for rigid-body interpenetration resolution in procedurally assembled scenes (2609.20524, 2026-09-17), a neural hierarchical-matrix preconditioner for real-time GPU solves (2605.13343) publisher · read in our library knowledge/fetched/cmp_graphics_arxiv_co_siga2026.xml · pin h51d49183aafe7aa1b0dbf264bc0c029b4716c07ea617c03a35fa2b58286599df · accessed 2026-09-27 · mirrored-indexGrounds: SB-TISSUE-2609
[arxiv_co_siggraph2026] arXiv API venue-tag scan, comment phrase SIGGRAPH 2026, 85 entries, newest 2026-08-31, read 2026-09-27: YASPS symbolic framework for extensible high-performance IPC simulation (2605.23088), robust and efficient penetration-free elastodynamics without barriers (2512.12151), fast sparse matrix permutation for mesh-based direct solvers (2602.00898) publisher · read in our library knowledge/fetched/cmp_graphics_arxiv_co_siggraph2026.xml · pin ha5eae48d6f540067918910978e53b083ba123cf01265d4b600c1ad1e52c40211 · accessed 2026-09-27 · mirrored-indexGrounds: SB-SOLVER-2607
[barrierfree_al_2026] Zheng, Juntian; Luo, Zhaofeng; Li, Minchen. Robust and Efficient Penetration-Free Elastodynamics without Barriers. arXiv 2512.12151v3 (revised 2026-05-28), ACM Transactions on Graphics 2026 with presentation at SIGGRAPH 2026, DOI 10.1145/3811035 as tagged in the arXiv listing. Read from the mirrored abstract: a barrier-free second-order augmented-Lagrangian solver that claims to match IPC robustness, adds every CCD-detected contact pair at once to avoid time-of-impact locking, updates Lagrange multipliers instead of raising penalty stiffness so the system stays well conditioned, and reports up to 103x speedup over GIPC on contact-rich benchmarks with a GPU-optimized simulator. publisher · read in our library knowledge/fetched/cmp_softbody_barrierfree2026.html · pin hf08bd629e8be61fab6513d83a5983f87573b6adad8930dff9faf062e055109dd · accessed 2026-10-01 · published-paperGrounds: self-contact-deep-interleaving
[gipc_2024] Huang, Kemeng; Chitalu, Floyd; Lin, Huancheng; Komura, Taku. GIPC: Fast and stable Gauss-Newton optimization of IPC barrier energy. arXiv 2308.09400v4, ACM Transactions on Graphics, related DOI 10.1145/3643028. Read from the mirrored abstract: keeps the IPC log barrier but rewrites it from a simplicial geometric measure of contact so analytic eigensystems (including the nearly-parallel edge-edge case) feed a projected-Newton solver, 3x faster than the standard IPC barrier formulation from the eigensystems alone, and an entirely GPU-based implementation of IPC. publisher · read in our library knowledge/fetched/cmp_softbody_gipc2024.html · pin h8044b7a2a44b388f6a1b40db7262c290704e3b8ac0b168edd323a53e30f85a3d · accessed 2026-10-01 · published-paperGrounds: self-contact-deep-interleaving
[ogc_2025] Chen, Anka He; Hsu, Jerry; Liu, Ziheng; Macklin, Miles; Yang, Yin; Yuksel, Cem (University of Utah and NVIDIA per the paper title block). Offset Geometric Contact. ACM Transactions on Graphics 44(4), August 2025, DOI 10.1145/3731205. Read from an nx_pdf_text extraction of the mirrored 25.9 MB PDF: each face is offset along its normal so contact forces stay orthogonal and the contact radius can be large, and per-vertex conservative displacement bounds replace continuous collision detection (the paper frames this as a trust region), giving more than two orders of magnitude over IPC-based simulation at a near-constant iteration budget and integrating with Vertex Block Descent; the guarantee is for codimensional objects, volumetric meshes may skip the bound and resolve intersections after they occur, and with a skeleton-driven body only cloth-cloth penetration is prevented while body-cloth penetration might occur; a C-IPC frame reported at 24 s runs in 0.24 s on CPU and 9.6 ms on GPU. publisher · read in our library knowledge/fetched/cmp_softbody_ogc2025.pdf · pin h6abfd9e1cda6d5413de348d8455ab4779473d0714f527f54f472e40de5598f98 · accessed 2026-10-01 · published-paperGrounds: self-contact-deep-interleaving
[arxivsearch_ipc_20261001] arXiv search listing, abstract phrase incremental potential contact, newest first by submission date, all 31 results read 2026-10-01 (the export.arxiv.org API answered 429 on four spaced attempts, so the arxiv.org search page is the mirrored listing): newest entry PhyVisGen, an IPC arm-gripper coupling for soft-contact robot manipulation data generation (2609.25653, submitted 2026-09-22, under review, co-author Kemeng Huang of GIPC); newest solver-method entry TOGEARI, interaction-space preconditioning for condensed finite-element systems with IPC contact on a 325,260-coordinate finger and cup problem (2608.14162v2, 2026-09-11); no October 2026 entry. publisher · read in our library knowledge/fetched/cmp_softbody_arxivsearch_ipc_202610.html · pin hd50033d2faac3a73a5e833d5fe750b27e6185c384b1d901c336bb17215c3785e · accessed 2026-10-01 · mirrored-indexGrounds: self-contact-deep-interleaving
[katzgefen_2023] Katz, Tomer; Gefen, Amit (Department of Biomedical Engineering, Tel Aviv University). The biomechanical efficacy of a hybrid support surface in protecting supine patients from sacral pressure ulcers. International Wound Journal 2023, PMC10502263, mirrored as the Europe PMC full-text XML because the pmc.ncbi.nlm.nih.gov fetch returned a reCAPTCHA challenge page. Read from the mirrored XML: a finite-element model of the supine buttocks on a hybrid foam-in-air-cell surface in static and alternating modes, validated by comparing model-predicted hydrostatic pressures of the skin elements to sacral interface pressures from calibrated pressure-mapping systems reported by the surface manufacturer or in the literature, the peak pressures differing by 3, 14 and 10 percent for the static mode, cycle A and cycle B; the authors state that measuring interface pressure alone does not address the cause of pressure injury, which is sustained soft-tissue stress concentration. publisher · read in our library knowledge/fetched/cmp_softbody_katzgefen2023.xml · pin h4b0bd0a49f4ba7dd7fc0374afb25455d22e561bcd46c4ebc65b7510019f42518 · accessed 2026-10-01 · published-paperGrounds: bedding-and-support-surface-load-distribution
[jner_buttocks_2026] Yang, Liqing and 13 co-authors, corresponding author Chen, Lin (Key Laboratory of Biorheological Science and Technology, Chongqing University, with Chongqing Polytechnic University of Electronic Technology, the CAS Suzhou Institute of Biomedical Engineering and Technology, Nanjing Tech University and Beihang University). A multi-scale finite element framework for evaluating buttocks-support surface interactions under static and dynamic loading. Journal of NeuroEngineering and Rehabilitation, 2026-08-31. Read from the mirrored structured abstract (the page carries no full text): an MRI and CT subject-specific buttocks model with Ogden hyperelastic tissue and foam and calibrated Mooney-Rivlin air-chamber membranes with internal ideal gas, under static and alternating 24/12 kPa loading, reporting stiffness (foam 7.69 N/mm, spherical air cells 5.35 N/mm), the share of simulated interface stress above 16 kPa and region-specific ischial and sacrococcygeal strain; the abstract states no validation against a measured pressure map. publisher · read in our library knowledge/fetched/cmp_softbody_jner2026_buttocks.html · pin h6ae95ed0fa81b5ec6a2328650892aea23d1292cd8dcca175601633d5412c5db7 · accessed 2026-10-01 · published-paperGrounds: bedding-and-support-surface-load-distribution
[hybridtextile_2026] Badin Dahesh, Melika; Asayesh, Azita; Asgharian Jeddi, Ali Asghar (Textile Engineering Department, Amirkabir University of Technology, Tehran). Linking compression mechanics and pressure redistribution in hybrid support surfaces for pressure ulcer prevention. Scientific Reports 16, 26693, 2026-05-21, DOI 10.1038/s41598-026-51596-7, PMC13507053. Read from the mirrored full text: body interface pressure measured with a PT-Mat-F10L2000 sensor pad at 3.5 cm2 per cell in supine and lateral postures, triplicates averaged after a 30 s stabilisation, reporting average and peak pressure, total contact area and high-risk area above 40 mmHg; higher compression work and elastic modulus went with lower peak interface pressure and greater yield strain with larger contact area; one healthy participant at pediatric-scale loading, and the authors name the sensor cell size as too coarse for fine pressure gradients. publisher · read in our library knowledge/fetched/cmp_softbody_hybridtextile2026.html · pin hab414df45365972252023e71db4250e98a14072cbb35d902cbc972db87cc9c9a · accessed 2026-10-01 · published-paperGrounds: bedding-and-support-surface-load-distribution
[bodiesatrest_2020] Clever, Henry M.; Erickson, Zackory; Kapusta, Ariel; Turk, Greg; Liu, Karen; Kemp, Charles C. Bodies at Rest: 3D Human Pose and Shape Estimation From a Pressure Image Using Synthetic Data. CVPR 2020, pages 6215-6224, CVF open-access page. Read from the mirrored page: a physics-based method that simulates human bodies at rest in a bed on a pressure-sensing mat, the PressurePose set of 206K synthetic pressure images with 3D poses and shapes, and PressureNet with a pressure-map reconstruction network, trained only on synthetic data and performing well on real participants' pressure images, with performance dropping substantially when that reconstruction network is ablated. publisher · read in our library knowledge/fetched/cmp_softbody_bodiesatrest.html · pin h4796b2a171f6704973ab4c8349afebcf6686b93573d1884f9648a7a60065d044 · accessed 2026-10-01 · published-paperGrounds: bedding-and-support-surface-load-distribution
[arxivsearch_pressuremap_20261001] arXiv search listing, abstract phrase pressure map, newest first by submission date, first 50 results read 2026-10-01 (the export API answered 429, so the arxiv.org search page is the mirrored listing): every September 2026 entry is a tactile or sensor-material paper or a haemodynamic homonym; the newest on-subject entry is 2512.13757v2 (2026-02-12; Manavar, Meyer, Wassmuth, Hammer, Schneider), diffusion-based synthesis of in-bed pressure maps from a depth image with an informed latent space and a weight-optimisation loss for physical plausibility; no October 2026 entry. publisher · read in our library knowledge/fetched/cmp_softbody_arxivsearch_pressuremap_202610.html · pin h923f6e5180435d349cc030661387fdf273ef91a69aa0fc79e0092bdbda570229 · accessed 2026-10-01 · mirrored-indexGrounds: bedding-and-support-surface-load-distribution
[cipc_2021] Li, Minchen; Kaufman, Danny M.; Jiang, Chenfanfu. Codimensional Incremental Potential Contact. arXiv 2012.04457v3, ACM Transactions on Graphics 40(4), SIGGRAPH 2021, related DOI 10.1145/3450626.3459767. Read from the mirrored abstract: IPC extended to codimension 0 to 3 DOFs in arbitrary combination with frictional contact, a C2 constitutive barrier that enforces cloth strain limiting as an energy while preserving rest state, and distance offsets that strictly guarantee shell mid-surfaces and rod mid-lines never move closer than the applied thickness, with a continuous collision detection that iteratively refines a lower bound converging to time of impact for the accuracy thickness demands. publisher · read in our library knowledge/fetched/cmp_softbody_cipc2021.html · pin h92131a3cda8fc3476e52f1b96804d15291748e5f53c566d4e5bfa12aca80b19d · accessed 2026-10-01 · published-paperGrounds: cloth-to-body-and-multilayer-garment-contact
[contourcraft_2024] Grigorev, Artur; Becherini, Giorgio; Black, Michael J.; Hilliges, Otmar; Thomaszewski, Bernhard. ContourCraft: Learning to Resolve Intersections in Neural Multi-Garment Simulations. arXiv 2405.09522v2, accepted for publication by SIGGRAPH 2024 conference track. Read from the mirrored abstract: a graph-neural-network cloth simulator with an intersection contour loss and a collision-avoiding repulsion objective that recovers from intersections introduced by missed collisions, self-penetrating bodies or errors in manually designed multi-layer outfits instead of relying on intersection-free inputs, shown on diverse multi-layer outfits under dynamic human motion. publisher · read in our library knowledge/fetched/cmp_softbody_contourcraft2024.html · pin h81ff349c54607b99e096ede9112db97389332a24bde68253b3f4c817a1240b88 · accessed 2026-10-01 · published-paperGrounds: cloth-to-body-and-multilayer-garment-contact
[fabric101_2026] Gong, Deshan; Mao, Ningtao; Yang, Xiaoyuan; Lu, Xinyu; Wang, He; Komura, Taku. Learnable Persistent Wrinkle Formation in Cloth Simulation. arXiv 2609.13707, submitted 2026-09-12, comment states accepted to SIGGRAPH Asia 2026. Read from the mirrored abstract: Fabric-101, over 101 common fabrics following textile standards with cyclic loading-unloading measurements that separate self-recoverable elastic, friction-driven recoverable and unrecoverable plastic deformation, and a differentiable elasto-plastic cloth simulator with friction whose fabric parameters are learned by the adjoint method from the measured hysteresis curves to reproduce persistent wrinkles. publisher · read in our library knowledge/fetched/cmp_softbody_fabric101.html · pin hd918b108ca8124672b91195cb1716ab23d72306cbf6a9a3ab7c502a01b3c71f3 · accessed 2026-10-01 · published-paperGrounds: cloth-to-body-and-multilayer-garment-contact
[arxivsearch_clothsim_20261001] arXiv search listing, abstract phrase cloth simulation, newest first by submission date, all 49 results read 2026-10-01 (the export API answered 429, so the arxiv.org search page is the mirrored listing): newest entry a simulator-agnostic gripper interface for robot cloth manipulation (2026-09-28); newest on-subject entries Fabric-101 (2609.13707, 2026-09-12) and HyperBones (2605.20460v4, 2026-09-04; Srivastava, Chen, Goldade, Herholz, Jiang, Lin, Yang, Sarafianos, Stuyck, Sharma, Larionov), a reduced-space neural garment simulator driven by virtual bones with corrections over linear blend skinning and a UV-space wrinkle network, trained physics-self-supervised without an offline simulator, over 30x faster than autoregressive neural simulators at about 1 ms per frame on a consumer GPU; no October 2026 entry. publisher · read in our library knowledge/fetched/cmp_softbody_arxivsearch_clothsim_202610.html · pin hbdd19866d434a9e40375b895233d1cb398cb1b07dcd4f1bd96691b9ae6aa904f · accessed 2026-10-01 · mirrored-indexGrounds: cloth-to-body-and-multilayer-garment-contact
[sb15_buttockfat] Wang K, Chen Y, Huang S, Wang L, Niu W. Subcutaneous Fat Thickness Remarkably Influences Contact Pressure and Load Distribution of Buttock in Seated Posture. J Healthc Eng 2021, article 4496416 (PMC8654536). A finite element buttock seated on a rigid (2 GPa) and on a 15.3 kPa foam cushion (linear elastic, v = 0), fat a one-term Ogden solid mu1 = 1166.7 Pa with v = 0.495: on the rigid cushion peak contact pressure 0.052 to 0.149 MPa and contact area 0.062 to 0.100, on the soft cushion peak 0.015 to 0.020 MPa and area 0.117 to 0.143, across four fat thicknesses (area unit printed as mm2, read as m2 against the 0.094 m2 baseline the same paper prints). publisher · read in our library knowledge/fetched/cmp_softbody_sb15_buttockfat.html · pin hd178af587c85dd550bd71a1cd14e3c26f41743c559b6320a38a808fce4f04b05 · accessed 2026-09-30 · published-paperGrounds: bedding-and-support-surface-load-distribution
[sb15_barrierfree] Zheng J, Luo Z, Li M. Robust and Efficient Penetration-Free Elastodynamics without Barriers. ACM Transactions on Graphics 2026 (SIGGRAPH 2026), arXiv 2512.12151. Matches the robustness of Incremental Potential Contact while removing its two bottlenecks, the ill-conditioned logarithmic barrier and time-of-impact locking, with a second-order augmented Lagrangian solver that incorporates every contact pair detected by continuous collision detection; up to 103x over GIPC. publisher · read in our library knowledge/fetched/cmp_softbody_sb15_barrierfree.html · pin h2277631852199923060151b6bafe77af2898754a9d4cfbb0724b263ee0434b45 · accessed 2026-09-30 · published-paperGrounds: bedding-and-support-surface-load-distribution
[sb15_ipc] Li M, Ferguson Z, Schneider T, Langlois T, Zorin D, Panozzo D, Jiang C, Kaufman DM. Incremental Potential Contact: Intersection- and Inversion-free Large Deformation Dynamics. ACM Trans Graph (SIGGRAPH) 39(4), 2020, the paper's project page with its abstract and citation. IPC variationally solves implicitly time-stepped nonlinear elastodynamics and maintains an intersection- and inversion-free trajectory regardless of material parameters, time step sizes, impact velocities and severity of deformation. publisher · read in our library knowledge/fetched/cmp_softbody_sb15_ipcsite.html · pin h120abacec5456865554f6f8cfa5107c5ba8e701373d70d3c36db5e8c90336971 · accessed 2026-09-30 · published-paperGrounds: bedding-and-support-surface-load-distribution
[ditgarment_2026] Dumoulin, Antoine; Boissieux, Laurence; Regateiro, Joao; Hellier, Pierre; Wuhrer, Stefanie. DiT-Garment: Garment Dynamics with Diffusion Transformers. arXiv 2609.18510, submitted 2026-09-16 (cs.CV), preprint with no venue stated on the page. Read from the mirrored abstract: dynamic 3D clothing over human body models in arbitrary motion from a 2D diffusion transformer that learns 3D deformation in UV space, conditioned on a UV position map of the template garment, on body motion and on physical parameters; generative, so it learns a distribution of possible outcomes; trained only on synthetic simulations of automatically generated designs and evaluated on captured and artist-made garments. publisher · read in our library knowledge/fetched/cmp_softbody_ditgarment2026.html · pin hc03910945769dfacf01a6c7cded410ef639cd959956ebe12f3a88753b1fcae40 · accessed 2026-10-01 · published-paperGrounds: cloth-to-body-and-multilayer-garment-contact
[ogc] Chen, A. H., Hsu, J., Liu, Z., Macklin, M., Yang, Y., Yuksel, C. Offset Geometric Contact. ACM Transactions on Graphics 44(4), SIGGRAPH 2025. Guaranteed penetration-free simulation of codimensional objects from an offset geometry whose normal contact force stays orthogonal to the contact surface, so a large contact radius drags nothing sideways; vertex-facet and edge-edge contact sets; and, in place of continuous collision detection, the conservative per-vertex bound of Wu et al. 2020, b_v = gamma_p times the smallest of the vertex-facet, edge-edge and facet-vertex distances around v with 0 < gamma_p < 0.5, a vertex past its bound truncated and the bound recomputed. Read from nx_pdf_text output of the stored PDF on 2026-09-30. publisher · read in our library knowledge/fetched/cmp_softbody_ogc.pdf · pin h6abfd9e1cda6d5413de348d8455ab4779473d0714f527f54f472e40de5598f98 · accessed 2026-09-30 · published-paperGrounds: cloth-to-body-and-multilayer-garment-contact
[fabric101] Gong, D., Mao, N., Yang, X., Lu, X., Wang, H., Komura, T. Learnable Persistent Wrinkle Formation in Cloth Simulation. ACM TOG 2026, arXiv 2609.13707. Fabric-101, over 101 common fabrics tested in a professional textile lab under cyclic loading, separating self-recoverable, recoverable and unrecoverable deformation, with a differentiable elasto-plastic simulator with friction; its Tables 11 to 13 give each fabric's thickness in mm and areal density in gsm, the rows knowledge/fabric101.conf admits (KN01 cotton jersey 0.7 mm 174.34 gsm, C02 cotton poplin 0.2 mm 108.07 gsm). Its friction is internal yarn friction, not a cloth-on-skin coefficient. Read from nx_pdf_text output of the stored PDF on 2026-09-30. publisher · read in our library knowledge/fetched/cmp_softbody_fabric101.pdf · pin h4358dab11ca3cc05d889ae3806f97e0792408748bc1c9bc59a1aa004f50e23a8 · accessed 2026-09-30 · published-paperGrounds: cloth-to-body-and-multilayer-garment-contact
[barrierfree] Zheng, J., Luo, Z., Li, M. Robust and Efficient Penetration-Free Elastodynamics without Barriers. ACM TOG 2026, SIGGRAPH 2026, arXiv 2512.12151 v3. A barrier-free second-order constrained optimization with a custom augmented Lagrangian solver that immediately incorporates every contact pair found by CCD, avoiding IPC's logarithmic barriers and time-of-impact locking, reported up to 103x faster than GIPC on contact-rich benchmarks. Read from the stored abstract page on 2026-09-30; the newest penetration-free line, NOT the method built in SB13. publisher · read in our library knowledge/fetched/cmp_softbody_barrierfree.html · pin hf08bd629e8be61fab6513d83a5983f87573b6adad8930dff9faf062e055109dd · accessed 2026-09-30 · published-paperGrounds: cloth-to-body-and-multilayer-garment-contact
[nist_adipose] NIST ESTAR, ASTAR and PSTAR material compositions: ADIPOSE TISSUE (ICRP), Density (g/cm3) = 9.20000E-01, mean excitation energy 63.2 eV, with its elemental composition by weight. Admitted as the density row knowledge/tissue_density.conf adipose_icrp, the tissue half of the garment-to-tissue mass ratio SB13's two-way leg needs. publisher · read in our library knowledge/fetched/cmp_softbody_nist_adipose_icrp.html · pin h45d4c1b525ce25a51af70a84179db3ac00811489ed6674e3ee99583abd7055c9 · accessed 2026-09-30 · datasetGrounds: cloth-to-body-and-multilayer-garment-contact
[nist_softtissue] NIST ESTAR, ASTAR and PSTAR material compositions: TISSUE, SOFT (ICRP), Density (g/cm3) = 1.00000E+00, mean excitation energy 72.3 eV. Admitted as the density row knowledge/tissue_density.conf soft_icrp. publisher · read in our library knowledge/fetched/cmp_softbody_nist_softtissue_icrp.html · pin h8aa1a5803d4b0b942ae224e0b7e9efbcfae99718d8981123f2b1043eb0a83961 · accessed 2026-09-30 · datasetGrounds: cloth-to-body-and-multilayer-garment-contact
[sb15_piper_lyon_beihang_2023] Liu S, Beillas P, Ding L and Wang X (Beihang University, Beijing; Universite de Lyon, Universite Gustave Eiffel, LBMC UMR_T 9406, Lyon), PIPER adult comfort: an open-source full body human body model for seating comfort assessment and its validation under static loading conditions, Front Bioeng Biotechnol 2023 11:1170768 (PMC10267746, CC BY), FULL TEXT read 2026-10-01 from the Europe PMC JATS XML: a subject-specific FE model compared with the same subject's measured seat, back and foot forces and pressure mats at four seat-pan angles; contact force error under 22.3 N horizontal and 15.5 N vertical against a 785 N body weight; seat-pan contact area, peak and mean pressure off by 5.8, -6.4 and 1.2 percent on average; the pressure share of four seat-pan regions off by 5.2, -1.4, -13.9 and -3.1 percent; backrest contact area, peak and mean pressure off by 20.2, 62.1 and 3.4 percent; measured pressures corrected per trial after Zemp 2016; no cell-by-cell map error is reported publisher · read in our library knowledge/fetched/cmp_softbody_bfh_piper_comfort_2023.xml · pin h3d2b93baf190f3a810c6dacf9b2c77fb5e439a778787388149d5b22dd82c1352 · accessed 2026-10-01 · published-paperGrounds: bedding-and-support-surface-load-distribution
[sb15_yan_hunan_2025] Yan L, Yu L, Li N, Shangguan L and Luo M (State Key Laboratory of Advanced Design and Manufacturing for Vehicle Body, Hunan University, Changsha; Xiangya 3rd Hospital, Central South University), Research on static seating comfort of the Chinese population under different seat angle design parameters, Front Bioeng Biotechnol 2025 13:1592166 (PMC12202639, CC BY), FULL TEXT read 2026-10-01 from the Europe PMC JATS XML: a human-seat FE model verified against capacitive pressure-mat tests of four 50th-percentile male volunteers; hip and leg mean and peak pressure consistent with the test, back peak pressure 20 percent above it, hip and leg contact area 2 percent below, and the count of pressure points per 4 kPa bin compared for hips, legs and back; contour maps compared by eye; no cell-by-cell map error is reported publisher · read in our library knowledge/fetched/cmp_softbody_bfh_yan2025_seatangle.xml · pin h193a3cdca6429e61af9e67e41adba02b8841ed6e80da42477dd881d993830c45 · accessed 2026-10-01 · published-paperGrounds: bedding-and-support-surface-load-distribution
[sb13_rodriguezpardo_seddi_2023] Rodriguez-Pardo C, Prieto-Martin M, Casas D and Garces E (SEDDI, Madrid; Universidad Rey Juan Carlos; Universidad Carlos III de Madrid), How Will It Drape Like? Capturing Fabric Mechanics from Depth Images, Eurographics 2023 (Computer Graphics Forum), arXiv 2304.06704, FULL TEXT read 2026-10-01 through nx_pdf_text of the stored PDF: real fabric samples with known ground-truth mechanical parameters in a simulator space of stretch (warp, weft, bias), bending (warp, weft, bias) and density, estimated from two depth images of a hanging and a stretch scene; a drape similarity metric checked by a user study with 86.68 percent agreement between participants; cites Kawabata KES and the Feng 2022 drape tester as its measurement lineage. A second measured-fabric source from a different group than Fabric-101 publisher · read in our library knowledge/fetched/cmp_softbody_bfh_drapelike_2304_06704.pdf · pin h1a1185001d03702be06dfa3f0b4fafb21d33972228351a3caad74ce8743f0bc1 · accessed 2026-10-01 · published-paperGrounds: cloth-to-body-and-multilayer-garment-contact
[sb13_huang_zhejiang_2024_teartest] Huang J, Jiao Q, Zhang Y, Xu G, Wang L and Yue D (Zhejiang Institute of Standardization; Zhejiang Light Industrial Products Inspection and Research Institute; Zhejiang University of Science and Technology, Hangzhou), Fabric tearing performance state perception and classification driven by multi-source data, PLoS One 2024 (PMC11020389, CC BY), read 2026-10-01 from the Europe PMC JATS XML: laboratory tear-strength testing of textiles monitored through electrical parameters, environment and operator behaviour, with working states classified by a decision-tree SVM above 98.73 percent accuracy. A TEARING TEST-PROCESS source only: it measures the test, not a fracture energy or a tear-propagation law a simulator could consume publisher · read in our library knowledge/fetched/cmp_softbody_bfh_huang2024_fabric_tearing.xml · pin h12682c12a8cb3755a6981f74fd18661bff111e11f99d3d32f470bf001f88b1c1 · accessed 2026-10-01 · published-paperGrounds: cloth-to-body-and-multilayer-garment-contact
[spb_2023] Chen, He; Diaz, Elie; Yuksel, Cem. Shortest Path to Boundary for Self-Intersecting Meshes. ACM Transactions on Graphics 42(4), SIGGRAPH 2023, arXiv 2305.09778 (submitted 2023-05-16). Read 2026-10-03 from nx_pdf_text of the stored 21.2 MB PDF, never a summariser: Theorem 1, the shortest path from an interior point to the boundary is the shortest line segment to a boundary point that is a valid path; Definition 4, a segment is valid iff an element traversal (tetrahedra consecutively sharing a face) carries it; Theorem 2, a closest boundary point on face f is the Euclidean closest point of f, so one candidate per boundary face suffices, tested in increasing distance; sec. 3.3, a topological ray traversal with a face tolerance that branches through edges and vertices, a visited list and a candidate stack, premature termination called entirely unacceptable; sec. 5.4, infeasible-region culling 10 to 30 times faster on some frames with identical results; Table 1, average operations Q, Tr and Tet printed as 65, 12.0 and 13.6 for the pre-intersected noodle (40K vertices, 110K tetrahedra, 15.2K DCD collisions) and 89, 12.0 and 14.0 for the pre-intersected squishy ball (219K vertices, 704K tetrahedra, 45.8K DCD collisions), the caption calling them BVH queries, traversals and tetrahedra visited on average per time step. publisher · read in our library knowledge/fetched/cmp_softbody_spb2023.pdf · pin hdbbb164eb1c48b6e60fe2a8f17a3849324dcf6f8687e8909fd3cca2268c4ea00 · accessed 2026-10-03 · published-paperGrounds: self-contact-deep-interleaving
[arxivsearch_selfint_20261003] arXiv search listing, abstract phrase self-intersection AND simulation, newest first by submission date, all 50 results read 2026-10-03: the three September 2026 entries are homonyms (merging supermassive black hole binaries 2609.xxxxx in astro-ph.HE, and two singularity-free guiding vector field papers 2609.19726 and 2609.07269 in cs.RO); the on-subject entries are older -- Shortest Path to Boundary for Self-Intersecting Meshes (2305.09778, May 2023) is itself in the listing, beside Efficient Incremental Potential Contact for Actuated Face Simulation and a robust grid-based meshing algorithm for embedding self-intersecting surfaces; no October 2026 entry and no newer method for recovering a deformable volume that already overlaps itself. publisher · read in our library knowledge/fetched/cmp_softbody_arxivsearch_selfintersection_202610.html · pin h027afb5cadde833cf5bf2fac078c8a1010dd5128aedfc72a63e2536f92fa2e62 · accessed 2026-10-03 · mirrored-indexGrounds: self-contact-deep-interleaving

generated by nx_swcompare_matrix (sovereign NishiLang organ) from knowledge/compare/softbody.matrix · source checks show implementation presence; runtime and user-outcome evidence are reported separately · JavaScript supports page controls