Water Workbench — Atomistic Tier (FSD-W1)

instantiating the engine…

⬡

WATER WORKBENCH REAL ENGINE · ATOMISTIC TIER

—

SCALE —

SIM RATE — —

DEVICE CLASS —

FPS —

— —

—

1.00×

—

MORE CONTROLS

thermostat engaged — heat removed or added is posted to W_ext, so the energy gate still closes (WB-2.1)

Atoms — a scene RESET, never a live mutation (WB-3.1)

Composition is identity, so changing it resets the scene rather than mutating it. The atom slider tops out at this engine's own capacity; a count past it would be delivered as time dilation, which is a worse answer than saying so.

Pressure — the control IS the box (WB-2.2) —

Virial pressure—

Box—

Hand column W_ext—

Boundary

The boundary is here because it is what makes two fences live rather than theoretical. The virial is only a PRESSURE on a wrapping box, so under walls the readout above refuses to call itself one. And a wrapping box has no bottom, so it REFUSES gravity — switch to periodic with the field on and watch the gravity panel fence itself for a reason you can check.

—

Gravity (WB-2.4) — a real, ledgered field —

Field—

Vector (box frame)—

Box tilt—

Potential V_g—

1.00 G

0°

The second slider TILTS THE BOX (WB-2.4c). The field stays pointing world-down; what changes is its direction in the box's own coordinates, which is the entire content of tilting a bucket. At 1 G nothing visible happens — the field is 15 orders below the thermal motion — so turn the strength up to see the water actually run downhill.

One G downward, at every scale. It enters the same external-acceleration array the walls and the hand already write to, so its impulse is booked in the momentum ledger automatically; its energy is a POTENTIAL term (V_g in the ledger), never a work receipt, because a uniform field is conservative and a receipt column would count the same joules twice.

You will see nothing happen, and that is the exhibit. One G is — — far below the thermal energy that is shaking this box apart. The same field is the entire hydrostatic column at a kilometre. What changes across the tiers is not the field's strength but whether the quantity that matters is a per-particle energy or a sum over the column: a hydrogen atom needs 246 km of height to gain one kT, and a kilometre of water still weighs 97 atmospheres because there are 10²⁸ of them.

ENERGY LEDGER (WB-4.3)

—

Kinetic T—

Pair V₂—

Three-body V₃—

Wall—

Hand spring—

Gravity V_g—

W_ext receipt—

Total E—

E − W_ext − origin—

Drift |now|—

Drift peak—

Drift bound—

Energy gate—

The hand's work is a COLUMN, not an exemption: dragging the anchor posts exactly dU into W_ext, so the gate stays closed through a drag. Thermostat heat is posted the same way. Conservation is chart-relative and the hand is in the chart.

Momentum residual—

Momentum bound—

Momentum gate—

CLOSURE-DEFECT LENS (WB-5.3)

—

Holon rows—

Worst defect now—

Worst at formation—

Molecules (census)—

Formations—

Dissolutions—

Closure rejections—

#	kind	members	E_bond	defect now	at formation

How well the coarse view commutes with the dynamics, live, as a number — the holon claim on screen. A candidate whose defect exceeds its budget is REJECTED rather than promoted, and the rejections are counted here.

THE CHART IN THE VIEWPORT (WB-5.1)

—

Pair knots—

R_e—

D_e—

vs 50-digit referee—

pair	route	determinants	price paid	state

No empirical force field, no fitted parameter, no phase-specific term. The curve is solved from closed-form Gaussian integrals at page load and checked against a pinned 50-digit reference. A pair past the in-browser determinant limit is refused BEFORE it is computed, not after.

The PRICE column is measured on this device, this session. It is there because the two doors to a curve do not cost the same: H–H arrives through holon-chem's bespoke H₂ path at about 0.16 ms per knot, and the same curve through the general N-centre route costs roughly ninety times that, for physics that agrees to the digits above. Curves the engine permits but has not been paid for are listed below and are solved only when asked. A page that blocks its main thread for tens of seconds has not obeyed M-CHEAPER-THAN-ITS-PRICE just because the number it eventually prints is right.

THREE-BODY SECTOR

—

State—

Untabulated triples refused
per force pass—

—

FENCE REGISTER (WB-5.2)

—

NOT SERVED BY THIS ENGINE

FENCED

The scale bands are NOT listed here. They were, and it was a second place for the same fact to be stated and to go stale: the ladder above carries each band's owner, exit and certificate, and gates them. One fence, one place.

THE THREE CLOCKS (WB-1.4 / WB-2.3)

—

dt (derived)—

ω·dt—

Requested rate—

DELIVERED rate—

Dilation—

Rung—

The DELIVERED rate is simulated time actually integrated per wall second, differenced from the engine's own clock — not the rate that was requested. When the device cannot hold the request the governor dilates time and says so; it never degrades accuracy to keep a number looking good (WB-6.2).

SCENE

—

Atoms—

Bonded pairs—

Clusters—

Temperature—

Simulated time—

Substeps—

THE TWO BOXES (FSD-W2)

—

Zoom ratio—

WORLD box — the physics—

SCENE box — the view—

View centre—

Drawing—

Outside—

Recent crossings—

Four world boxes carry the physics; one scene box carries the view, and they are never the same knob. The HAND acts on the world box — that is the pressure control. The ZOOM is a ratio: the scene box is the world box divided by it, so 3× zoom on a large world and 3× on a small one are different scene-box sizes. Zoom never touches the physics.

Nothing is destroyed by a crossing. A holon leaving the scene box leaves the VIEW; the world box keeps simulating it, which is exactly what makes the pressure and temperature below whole-only and zoom-invariant. Zoom back out and the same holons return at their current evolved state — the water kept moving while you were not watching it. A cache that froze them would return you to a world that stopped when you looked away.

THE LEDGER BY CHANNEL (WB-9.1) — SIX THINGS TWO WHOLES DO TO EACH OTHER

—

The force law is organised by SECTOR — the near pair table, the far tail, the three-body tables, the clusters, the field, the seam. The physics is organised by CHANNEL: six things two closed wholes do to each other, each with a rate derived from standard theory rather than fitted. These are two different partitions of one energy and they map many-to-many, which is why the value column below is mostly empty.

"Not served" is the honest answer, and it is not zero. A ledger row's number can hold several channels at once, and then it is a BOUND on each of them and the value of none. The engine hands the page a number only where a row carries one channel wholly and alone; everywhere else it hands back a refusal, and the cell prints the refusal. A zero there would read as "this channel contributes nothing", which is a measurement nobody made.

The columns are the account's six kinds, and their order is the grounding stack: Premises founds Model, Model derives Facts, Facts modulate Confidence — then Priorities (what we want first) and Manner (how it is written), which are about the instrument rather than about the world. A channel's history walks its row left to right, because a Confidence cannot be set on Facts not yet measured. Each filled cell is a campaign in this repository and carries the line of the record it came from; an EMPTY cell is empty, not a dash and not a sentence written to fill the width.

The two sixes are not the same six. There are five kinds, not six: Identity fills two rows because it is the only kind with an edge, and an edge has two sides — refusal is which electron is which, sharing is electrons lent across while both stay whole. The equal count is a coincidence and no cell here claims a pairing.

channel	reach, this scene	value, this scene	Premises	Model	Facts	Confidence	Priorities	Manner

THE SCALE LADDER (FSD-W3 §11.2) — 1 km TO THE NUCLEUS

—

A kilometre of water is about 3×10³¹ molecules. Nobody simulates that and this page does not pretend to. What the zoom axis shows instead is the ladder itself, from the 1 km cube down to the nucleus of a single picked atom: every band is here, and each one either runs its certified physics or carries a DEBT with the build that is paying it off. The fences are this page's work queue and the rungs going live one by one is what the queue emptying looks like.

Three kinds of warrant, and they do not substitute for each other. The three coarse bands flip on a node-G closure certificate — a coarse view of the dynamics beneath them, certified by the census. The two fine bands hold no certificate and are not entitled to one: there is no coarse view at an atom, there is the solve, so they flip on their EXPORTS being in this artifact, named one by one in each band's debt line. The floor below the nucleus is fenced on physics nobody has yet.

—

Acuity is the allocator. A band's population is what this view can actually distinguish: nothing while the view is far wider than the band's own scale, ONE when the view has zoomed to that scale — the pinned seed — and cubic growth after. That is why a cube of 3×10³¹ molecules is cheap to stand in front of: by the time finer structure is visible at all, almost nothing else at that scale is in view, and everything outside it stays coarse under the de-allocation law. No representative slabs, no bulk fine-simulation.

The zoom de-allocates; there is no handoff machinery and none is owed. A holon that is not load-bearing releases its members' fine degrees of freedom, and load-bearing is not a heuristic — it is the measured per-row closure defect shown live on the molecular band. A row reading ~0 is autonomous by measurement and its composite carries it exactly on closure boundaries; a row being buffeted, grabbed, or coupled into the visible region keeps its fine allocation. Zooming out is the same accounting event in reverse, and formation and dissolution are accounting-only: energy before equals energy after, bit for bit.

Three debts this panel owes, shown rather than hidden. First: no threshold separates "~0" from "not ~0" yet, so the panel reports the defect DISTRIBUTION and not a verdict on which rows would release. That number belongs to the grain law's own stated bound and is the engine's to derive, not this page's to guess. Second: this page performs NO de-allocation — none is implemented — so what you see is the criterion being read, never a budget being reclaimed. Third: the fenced bands above are not yet rows in FENCES.md, which the fence law requires; they carry their owner and exit here, and their registration is owed.

What this build does not do (WB-10.6). It does not flip the three coarse bands — that is node G's certification and no page work substitutes for it. It does not draw the interior of a nucleus. And the gauge vacuum below the nucleus is not on this ladder at all: the operator sequenced it after this build, and a band on the page for work nobody has scheduled is a promise wearing a fence's clothes.

THE FLUID ELEMENT, LIVE — FLUID-1's ORIENTATION LATTICE (WB-10)

—

This is the band's own instrument, running — not a picture of it. Every particle on the lattice carries a direction; a bond forms when the geometry CT-2's measured angular map favours holds; a bonded pair is a closure that moves as one and the rent clause releases it. Blue is free, orange is bonded, opacity is how full the cell is, and the lines are the live bonds drawn along the donor's own arm. The lattice is triangular and the array is square, so the picture is sheared by the engine's own Euclidean embedding — nothing here knows what a hexagon is.

What is live, and what is quoted. Live: the structure and the conserved integers below. Quoted, on the band above with the line of the record it came from: FLUID-1's transport numbers. Those were read at L = 256 over millions of member-steps with a fitted decay per wavevector, and this cell does not re-derive them — a Schmidt number fitted from a few thousand browser frames would be a measurement nobody made.

—

The control rebuilds the scene, it does not flip a switch. FLUID-1's no-bond control is a run with bonds forbidden from the first step — and with bonds forbidden the occupation is Lattice::advance_with_colour's bit for bit, which a lattice carrying bonds formed before a mid-run flip would not be. The door enforces that; this page only asks for it.

THE LEDGER — every one an integer, audited at every step

Mass—

Momentum, first axial component—

Momentum, second axial component—

Orientation census, total—

Tracer bits—

Bonds held now—

Bonds formed, cumulative—

Released by the rent clause—

Released by the exclusion (blocked)—

Collisions fired—

Steps taken—

Every integer exact?—

The bond balance is the one that is not obvious. Formed minus released by rent minus released by the exclusion equals held, at every step, exactly — and the two kinds of release are counted apart rather than summed, because FLUID-1's fifth finding is that the exclusion is 36 % of them at the chart and 100 % of them in the cold control. On an FHP-6 torus there is no rest state, so a bonded pair refused its joint move has to break: the lattice, not the rent, is the ceiling on the bond.

THE BOND GRAPH — the band's own question

Bonds per particle—

Largest component—

Spanning is a loop around the torus, not a reach across the box. A component spans when walking its bonds around a cycle returns to the starting cell displaced by a non-zero multiple of L — which is what a percolating network means and, unlike "touches both edges", does not change when the configuration is translated.

THE RULE, READ — every angle carries the record it came from

—

Nothing in this rule was typed. The six amplitudes and the retention are CT-1's and CT-2's own numbers, shipped as a byte-identical copy of the campaign's record, digest-pinned by this page and diffed against the conformance tree by this page's gate. They are PUSHED through the engine's doors one at a time, and the engine refuses to build a lattice until all seven have arrived. The rent is read BACK from the engine after the retention goes in: this page pushed a measured fraction and the engine computed the chart.

THE LADDER'S READOUTS (FSD-W3 §11.2)

—

The ladder above carries each band's state, owner, exit and citations. This card carries the numbers those bands serve. LIVE means the digits came out of this artifact this frame. DECLARED means a measured input the Hamiltonian never computes, read from a committed table and labelled as such (WB-1.7) — including when it arrives through an export, because a function returning a measured mass does not make the mass computed. COMPUTED means arithmetic on this page over inputs that are themselves live or declared. FENCED — PENDING means the export that serves the row is not in this artifact yet; the row names it and shows no digits.

THE PICK — WB-1.6

Pinned atom—

In a molecule, or free—

THE ATOM BAND — electronic structure, WB-10.2

Electronic energy—

Electrons in the solve—

Solve residual—

Solve exit—

THE NUCLEUS BAND — WB-10.1, declared inputs per WB-1.7

Z—

Isotope—

Mass—

Nuclear spin—

Charge radius—

Thermal de Broglie wavelength—

A declared door that has nothing to declare returns a sentinel, and these rows fence on it rather than drawing it: the integer doors return u32::MAX, which reaches this page as −1, and the real-valued doors return zero. Two of the ten elements here have a nuclear spin of exactly zero, so the test is the sentinel itself and never falsiness. The three rows fence independently, because the engine's mass table covers more elements than its nucleus table — at Z = 11 the mass is served while the spin and charge radius are not. That fence is not reachable from this page's own species set: every element in the shipped palette has a declared nucleus, so the guard is correct and cannot fire here, and this page's gate re-measures that on every run.

THE BIT-IDENTITY GATE — WB-10.2

holon_law_probe() in this wasm—

the pinned native value—

Bit identity—

One determinant kernel runs every space (§11.1), so the page's solve and the native referee's are the same arithmetic — which makes EQUAL TO THE BIT a thing that can be displayed rather than a thing that has to be trusted. The comparison is on the 64-bit patterns and carries no tolerance: a near-equality here would pass on two numbers that differ, which is the property the lane kernel exists to make checkable.

THE CUBE BAND'S COLUMN — ρ g h, §11.2

Scene mass density—

ρ g h at the view's own depth—

ρ g h at the cube band's 1 km—

Arithmetic on measured constants, and nothing more. §11.2 grants the cube band this one readout while its DYNAMICS stay fenced: the density is summed from the engine's own box and the species table's declared masses, and g comes from the engine's own constant. It is the gravity exhibit (WB-2.4a) at the top of the ladder — what changes across the tiers is not the field's strength but whether the quantity that matters is a per-particle energy or a sum over the column.

It is not a claim about water. This scene is a handful of atoms in a box a few nanometres across, and its density is its own. A kilometre of liquid water gives about 9.8 MPa, or 97 atmospheres — that figure is measured in tests/gravity.rs and belongs to WB-2.4a, not to this readout. The same arithmetic on two different substances is the whole of the exhibit.

THE WATER STORY (WB-9.6)

RECORD

This is the only panel on the page whose numbers did not come out of the engine you are watching. They were measured by the closure census on banked trajectories, and each one carries the artifact and line it came from — checked by this page's own gate, so a citation cannot outlive the number it cites.

What is settled: water formed and HELD. Two separate blocks of one oxygen and two hydrogens clear a window that was frozen before the instrument measuring it existed — 893.8 fs and 923.9 fs, both CERTIFIED-STRICT, both on moving carriers, both named by the census.

Both arose with the four-body term absent, and in the second case absent by MEASUREMENT: the solver's own counter reads exactly zero, which is functional proof it never fired where a symbol check could prove nothing, the symbol being inlined away. That run is a true one-variable control — same seed, same commit, same binary, the term switched off — and its provenance is pinned above.

What was RULED, and the sentence that stood here is gone. This panel said the causal question was open with its control unadjudicated. It has been adjudicated: the claim that three-body physics alone stops at hydroxyl and the four-body correction carries the system to water is FALSE on this seed. Both arms certify the same three atoms, strict, with a control rate of 0.000 against 111 candidates each — §12.3's branch (b), staked before any of the data existed.

The ruling rests on the CONTROL, not on the comparison. The arm with the four-body term switched off — solver counter reading exactly zero, momentum at 3.84e-5 of bound — certifies water on its own. Nothing in that conclusion needs the treated arm at all, which is what makes it a refutation rather than a disagreement between two runs.

What is still NOT ruled: what the term DOES. The treated molecule held longer, 2599.8 fs against 923.9 — and that difference may not be credited to the term. The treated run alone leaves its declared 2D plane, reaching 11.49 bohr against a 12.0 half-depth from the frame just after the term first fires, while seventeen other trajectories hold their third coordinate bit-exactly for 20,000 frames. The one-variable design was defeated not by an uncontrolled variable but by the treatment producing one, and a bent triatomic has more room in the space that opened. That question is open; this one is closed.

Two qualifications any reading has to carry. A formula census at the final frame is not a closure certification — the rows above are the census's own verdicts. And every certification here rests on an O–O pair curve that is not converged: worst residual 4.81e-6 hartree on the controlled pair of arms against the code's own threshold of 1e-9, and 2.68e-6 on the fenced arm, which was measured at a different commit again. Identical between the two arms being compared, so it introduces no second variable there — but it qualifies any absolute claim, and it belongs in the verdict rather than in a log.

DETERMINISM (WB-5.4)

Run A digest—

Run B digest—

Verdict—

Each run—

Two runs of the same seeded scene, reset and stepped identically, digested over the RAW f64 bits of every coordinate. A one-ulp difference changes the digest. The claim is per device class, and the class is declared in the manifest.

THE ENGINE DID NOT LOAD

PROVENANCE MANIFEST (WB-3.4)

The artifact this page instantiated

Physics integrity statement (WB-5.1)

What this page does NOT yet demonstrate