Six signed bundles. Five shapes of result. One record.
Every card below is the same record, smrforge.result.v1 inside an
Evidence Envelope, signed with the SMRFORGE production key. Download the raw JSON, verify it offline
by recomputing the SHA-256 and checking the Ed25519 signature, then re-derive it by running the open
engine on the inputs sealed inside it. The seal proves custody of the computation. It does not prove
the physics is right.
1 · integrity, offline
$ smrforge verify <bundle.json>
Recomputes every SHA-256 in the manifest and checks the Ed25519 signature over the bundle hash. No physics is imported. The in-browser version does the same with no crypto library behind it.
2 · re-derive the number
$ smrforge reproduce <bundle.json>
Re-runs the open engine on the bundle’s own sealed inputs. Deterministic cards re-derive bit-identical. Monte Carlo cards carry a sealed tolerance, a re-run within 3σ.
3 · origin, decided by you
key ed25519:cfb88804…d69b52c
Origin is never a field in the bundle. Your verifier compares the signing key to the pinned production key in the trust root, and decides.
All six bundles below carry schema smrforge.evidence_envelope.v1, signature alg ed25519 over message bundle_sha256, public key cfb88804b92163566d485c8be3757baf74215944ad4852326457fcfd3d69b52c. Raw files are served from gitlab.com/smrforge/smrforge/packs.
The gallery
The seal doesn’t care what the number is.
Each card shows the sealed headline value with its units, the honest grade the engine
wrote into the bundle, and one caveat lifted from the bundle’s own caveats field.
Nothing here is a claim the bundle doesn’t make itself.
scalarfidelity: lab-escalationverdict: advisory
the demo
Criticality, k_eff
physics:openmc · engine openmc-0.15.3-mg · the Break-the-Seal bundle
1.00522k_eff · ±0.000596 (1σ)
sealed inputspreset lwr-smr-2g · 4.95 w/o · 30 cm
reproducere-run within 3σ of k_eff
bundle sha-25640a35716…5dc98
signatureed25519 ✓
Caveat, from the bundle. OpenMC multi-group eigenvalue on a homogeneous slab. Continuous-energy with heterogeneous geometry is the next escalation.
The sealed assembly_power array, all 52 values, drawn on the quarter-core node grid it describes, fuel assemblies only, reflector excluded. Opacity scales with the sealed value, 0.417724 at the corner to 1.612683 at the two outlined peak assemblies. Hover a cell for its value.
lowhigh
n_fuel_assemblies52
k_eff vs reference1.029633 vs 1.029585 · C−E +4.8 pcm at this mesh
converged C−E−3.0 pcm (Richardson, h→0)
reference_power_peaking1.479 (NEACRP-L-138) · mesh-dependent here
reproducebit-identical, the whole array
bundle sha-25685be9c26…9128c
signatureed25519 ✓
Caveat, from the bundle. Code verification against the published 2-D reference, not a physical validation. The peaking at this mesh (mesh_per_10cm 1) is 1.613 and converges toward the published 1.479 as the mesh is refined.
data librarysmrforge-burnup-1g/u235-illustrative · by sha-256
reproducebit-identical
bundle sha-256bd527f12…b01f6
signatureed25519 ✓
Caveat, from the bundle. The depletion solve is exact against closed-form Bateman, but the network and cross sections are synthetic. These isotopics are screening-grade and not a claimable irradiated-fuel assay. Exact arithmetic on unvalidated data.
measured 788 pcm (BEAVRS spec Table 22) · C/E 1.029, computed from the two sealed numbers
Computed 811.2 ± 24.6 pcm as one bar from zero with its 1σ whisker. The dashed line is the measured 788 pcm, which lies inside the whisker. BEAVRS publishes no formal per-worth sigma, so no measured error bar is drawn.
reference statuslab-grade per run · not reference-earned
bundle sha-25625063e70…ddfaba
signatureed25519 ✓
Caveat, from the bundle. A continuous-energy OpenMC bank worth from two verbatim sealed BEAVRS decks. Lab-grade per run. Reference is an earned per-domain validation status, not a per-run label, and this bundle does not claim it.
extrapolation4.95 w/o above the benchmark range [2.4, 2.46]
reproducebit-identical
bundle sha-256d9fde46d…f4db3
signatureed25519 ✓
Caveat, from the bundle. Screening c_k reports n_eff 1.0 for the 7 applicable anchors, so the 95/95 tolerance factor is undefined and there is no qualified statistical USL from this basis. The optimistic bound treats correlated cases as independent. Do not quote it as the USL. A v1 demonstration of the open pipeline, not a licensing product.
physics:openmc-htr10 · engine openmc-0.15.3-ce · ENDF/B-VIII.1 · explicit TRISO and pebble bed
1.016588k_eff · ±0.001023 (1σ)
c_over_e 1.01659, sealed · measured k 1.0000 ± 0.0037 (IAEA TECDOC-1382 Part 2)
benchmark_idHTR-10 first criticality @ IAEA TECDOC-1382 Part 2
geometry16,960 pebbles · 9,689 fuel
generatorsealed by sha-256 · 835b6b09…
reproducere-run within 3σ of k_eff · c_over_e derived
bundle sha-25687d00c46…839a1
signatureed25519 ✓
Caveat, from the bundle. The +1.3 to +1.8% C/E is the documented HTGR continuous-energy over-prediction, attributed in the literature, not bespoke. The claim is consistency with the cited codes, not “we hit 1.0000.” Reference is earned per domain, not a per-run label.
Every grade on this page was written into the bundle by the engine at mint time and is
covered by the signature, so a reviewer who changes screening to reference breaks the seal the
same way as changing the number. A sealed number from a badly specified core is still a faithfully preserved
wrong answer. Nobody should license a reactor on a SMRFORGE number alone.
The seal proves custody of the computation, whatever its shape.
k_eff is the demo, not the limit.
pcm = per cent mille (10−5 Δk/k). w/o = weight percent U-235. C/E = computed over experiment.
Verification (vs analytic and published references) is not validation (vs measured experiments). Nuclear data cited
on this page is openly distributable (ENDF/B-VIII.1).
Your number, sealed the same way
Bring a result a reviewer won’t take on faith.
A scalar, a field, a step, a comparison or a margin. We seal it as one of these records, you verify
it offline and re-run it yourself. Engagements, pilots and Pro at support@smrforge.io.
