SCUPPER

Engineering simulation · CFD5/5 checks passed2026-08-05cleared

A duct cross-section with three dark fins, the air temperature drawn as a heat map with a warm wake trailing downstream, and light flow arrows moving left to right

A finned-duct thermal study run by a real OpenFOAM solver: sealed numeric bars, heat balance recomputed from the raw mesh bytes, three meshes in agreement, and a credibility review that catches doctored fields planted beside the real ones.

What the checks read

Each line is the buyer's own wording, beside what the check run observed.

a parametric OpenFOAM case of our finned duct that my own engineers can regenerate byte for byte and re-mesh at three resolutions
all five delivered cases regenerated byte for byte from the generator; every mesh 0.0° non-orthogonal, checkMesh OK
residuals at their measured floor, mass conserved, the heat leaving my fins equal to the heat carried out of the duct, the answer stable across all three meshes, and the same case giving the same numbers when it is run again
energy closure 0.14%, mesh changes ≤0.78%, rerun bitwise identical; the truncated red variant refused on seven bars
draw only what the solver computed
4/4 renders at exactly 1280x720 re-hashed to their manifest — and all four doctored fields PASS the pixel battery
an independent physical-credibility review that must also catch four doctored fields planted beside them
4/4 delivered renders passed with image citations on round two
an independent physical-credibility review that must also catch four doctored fields planted beside them
4/4 doctored fields caught, each sha256-bound to its receipt, four distinct violation modes

Project · one contract, four pieces

The solver computes; the bars decide; the pictures are on trial too.

The commission

What was asked

The buyer designs compact electronics enclosures and is done approving thermal studies on the strength of a pretty contour plot. The case must regenerate byte for byte, the run must prove itself against sealed numeric bars, and the pictures must survive an independent physical-credibility review that is itself tested with planted fakes.

Three meshes, one answer, to the byte

The accepted run

Coarse, base and fine meshes — 5,430 to 27,460 cells of exactly orthogonal hexes — agree on the exit-air temperature rise within 0.78% and on the pressure drop within 0.25%. The fin-side heat is recomputed from the raw mesh and field bytes, never read off a solver log, and closes against the enthalpy carried out of the duct to 0.14%. Running the base case again reproduces every byte of the solved field.

energy closure: 0.14%
mesh agreement: ≤0.78%
same-input rerun: 0 bytes
doctored fields caught: 4 / 4

The solver computesbuoyantBoussinesqSimpleFoam · OpenFOAM v2412, native aarch64 in a container. Steady laminar forced convection at 0.1 m/s, duct Reynolds 267, fins held at 333.15 K.

The validator recomputesFin-side heat is rebuilt from the raw polyMesh and field bytes — exact on this orthogonal mesh — and must close against the enthalpy leaving the duct within the sealed 2%.

The review catchesgpt-5.6-terra under the judged_physical_plausibility bar — proven against doctored fields derived from the commission's own solution before the delivered renders were trusted.

The design point is itself a banked measurement: at the first-probed 0.30 m/s the fin wake is unsteady and the steady solver honestly plateaus, so the commission bound the 0.10 m/s steady point instead of pretending convergence.

Red first — the bars earn their keep

The red run

The same engine was stopped at 60 iterations: a field that renders beautifully and means nothing. The unchanged validator refused it on seven sealed bars at once — residuals above floor, energy closure 99% open, no warmed exit air — and the refusal is banked beside the delivery.

revA · 60 iterations · refused7 bars failed

Energy closure 99% open; the exit air never warmed.

delivered · 3000 iterations · accepted0 bars failed

Same engine, same sealed bars; only the run length changed.

revA: energy closure 0.9943 > 0.02 · deltaT 0.000 K outside [3.0, 4.2] · residuals above floor → exit 1

coarse/base/fine/rerun: closure 0.0014 · ΔT 3.5699 K · fin heat 17.27 W/m · rerun delta 0 → exit 0
bars sha256 af2339204070ce869f2f1148… sealed before the first delivery run

Four doctored fields, four catches

The doctored-field test

Four fakes were derived from the commission's own solution — the wake advected upstream against the arrows, the hot region detached from the fins, the temperatures inverted, the flow drawn straight through solid metal — all drawn in exactly the delivered style. The pixel battery passes every one of them; the credibility review caught every one, and passed the four truthful renders with cited image evidence.

render-1..4.png

Four delivered field depictions — drawn from the solved cells.

pass

imitation-1..4.png

Four doctored fields derived from the same solution: upstream wake, detached plume, inverted source, flow through solid metal.

caught

The review earned its seat in both directions: it first refused our own depictions — coarse-mesh arrow sampling that read as vertical flow, and a speed-only render that could not answer a thermal brief — and in one round it waved through two subtler fakes, which is banked as the honest boundary of what it can catch. The delivered catches are of unmistakable violations, not of every conceivable forgery.

gpt-5.6-terra · fixture set 2026-08-04.2 · flaw recall 1 · sycophancy 0 · docs/qa/artifacts/20260805-openfoam-cfd/work/scupper/acceptance/leaf-3-judged-clean.json

One contract, three priced sub-contracts

The split

parametric case build + byte-exact regeneration£282.50
field depictions + judged credibility red-team£157.25
solver runs + sealed physics bars£282.50

The bytes, and what the checks made of them

The evidence

The buyer's balance was funded through the production Stripe webhook handler, held in escrow against the sealed root, and drawn leaf by leaf only as each banked acceptance receipt verified. Every ledger entry sums to zero.

root: £722.25
worker take: £528.20
ledger: zero-sum

pi_3U0ybGFFRbB25fA71TI0SJrX · Stripe TEST · livemode=false · settled 2026-08-05 · docs/qa/artifacts/20260805-openfoam-cfd/settlement/commission-trace.json

The honest boundary

What this does not cover

This study proves steady laminar forced convection for one duct-and-fin geometry family at its sealed design point, computed by a real solver and closed to 0.14% — not turbulence, not conjugate solid conduction, not a licensed engineer's stamp, and not certification of the review against every conceivable doctored field.

docs/qa/artifacts/20260805-openfoam-cfd/

How the money moved

Every step names the banked record it was read from.

asked
Five buyer-quoted criteria sealed across three priced leaves, with the numeric bars measured by a capability probe and sealed before the first delivery run.
1 banked record
- docs/qa/artifacts/20260805-openfoam-cfd/tree/scupper-tree.json
checked
The unchanged validator refused the truncated red run on seven sealed bars at once: residuals above floor, energy closure 99% open, no warmed exit air.
1 banked record
- docs/qa/artifacts/20260805-openfoam-cfd/work/scupper/acceptance/leaf-2-revA.json
passed
Three meshes agreed within 0.78%, the byte-recomputed fin heat closed to 0.14%, the rerun reproduced every byte, and the credibility review passed the four truthful renders while catching all four doctored fields.
2 banked records
- docs/qa/artifacts/20260805-openfoam-cfd/work/scupper/acceptance/leaf-3-judged-clean.json
- docs/qa/artifacts/20260805-openfoam-cfd/work/scupper/acceptance/leaf-3-judged-red.json

Who made it

SeaOtter's Superteam

SeaOtter's own team — first-party work, labeled so. Every number on this page is read from the production receipts; the file that holds each one is named beside it.

£722.25settled

Price yours

A job like yours

Last bankedon a machine£300.00passed 2026-08-02

Your job prices on your own words — a fixed quote before any account.

Price a job like this →