Set up the case
Inputs, units, assumptions, references, and intended QoIs become one runnable case.
AI-first engineering simulation software
Anvil Sim.
An AI-native multiphysics simulation platform that can set up, run, debug, compare, and explain engineering simulations across thermal, structural, fluid, electromagnetic, and coupled domains.
EVIDENCE READYsetupruncomparereview
review workflow
The result is not done until it can be inspected.
Anvil shortens the path from engineering question to reviewable result. The public site now mirrors the product shape: inputs, solver run, comparison target, evidence artifact, and the next validation step stay close together.
simulation reviewvisible checks
setupgeometry, units, material card, boundary condition
runsolver log, residual trend, convergence note
comparereference source, QoI, tolerance, delta
decidemargin, limitation, next validation step
Run the physics
Solver kernels produce fields, trajectories, modal responses, thermal maps, and logs.
Compare the result
The reference, QoI, tolerance, status, and artifact stay visible in the same row.
Review the decision
The output becomes an engineering packet that can be inspected, questioned, and reused.
current flagship visual
Boomerang flight-path replay
A 3D flight-path replay makes the boomerang trajectory legible first, then links it to return-window diagnostics and the evidence packet. The flight-path video makes the motion legible, while the diagnostics packet keeps the solver evidence and next validation steps connected.
Analysis companion
The diagnostics panel keeps the engineering review visible: return-window error, height over time, receipt summary, and next validation steps.
Physical claim
Geometry, source faces, and material assumptions
Mesh quality, solver logs, and convergence
Benchmark/reference comparison and .anvilprov route
evidence: V4 return-window evidence packet
validation: Configured v4 return-window solver gate, with measured-flight and aerodynamic correlation as the next validation path.
Visual demos are the front door. Benchmarks are the roadmap.
Anvil should look impressive and stay reviewable. The product direction is to make each flagship demo clickable down to the physics claim, inputs, solver evidence, benchmark gate, validation focus, and next evidence step.
Boomerang trajectory replay
EVIDENCE READYA configured blade-element v4 solver trace enters the 0.50 m return window while airborne before first ground contact. The site shows the visual replay and the evidence route together.
evidence: V4 return-window evidence packet
next: Measured-flight and wind-tunnel correlation.
FEA, modal, and FRF trust wall
REFERENCE PASSESStatic, modal, and forced-response evidence reels show Anvil's current structural evidence spine: benchmark rows, native capture receipts, and .anvilprov-backed replay paths.
evidence: FRF harmonic-response evidence packet
next: Broader certification, fatigue, weld, and preload cases.
Thermal and electronics simulation
VALIDATION TRACKThermal-SIMP and electronics-airflow previews point toward heat-sink, enclosure, and AI chip package workflows where temperature QoIs, power assumptions, and validation data must stay visible.
evidence: Thermal-SIMP and electronics thermal-airflow public clips
next: Chip-package temperature tolerances and reference data.
CFD validation ladder
IN BUILDThe CFD work is being pushed through canonical rungs: manufactured/simple flows, conservation checks, mesh sensitivity, then public benchmark cases before any flagship claim.
evidence: Benchmark ladder and solver-gauntlet receipts
next: Reacting-flow reference cases and published deltas.
Geometry and mesh provenance
IN BUILDSource CAD, topology hashes, source-face selection, load/constraint binding, mesh quality, and remesh-stable evidence are becoming first-class evidence inputs.
evidence: Evidence-packet schema and provenance receipts
next: Source-face evidence carried through meshing and solver setup.
Robotics decision workflow
VALIDATION TRACKRobot bracket and robotics-style workflows are being shaped into design-review packets: claim, geometry, loads, material assumptions, solver output, validation steps, and next evidence.
evidence: Robotics evidence-packet queue and validation focus ledger
next: Source-face and reference-solver comparisons for robotics cases.
Native Evidence Console
IN BUILDThe Mac app is becoming the review room: evidence-packet library, visual replay, benchmark deltas, provenance, reviewer packet routes, and validation steps in one place.
evidence: Native Mac Evidence Console build
next: Visual readiness is kept separate from proof readiness.
benchmark wall preview
Reference, QoI, validation focus, evidence.
The public trust wall makes the current evidence level easy to scan: reference passes, target rows, validation focus, and latest artifacts.
| Benchmark | Reference | QoI / tolerance | Status | Validation focus |
|---|---|---|---|---|
| Boomerang return-window gate Boomerang flight path | Scenario-local acceptance gate with a public expansion path | Trajectory enters the 0.50 m return window while airborne before first ground contact. | SCENARIO EVIDENCE | Current public evidence is scenario-local return-window evidence; measured-flight and wind-tunnel correlation are the next expansion path. |
| FRF cantilever harmonic response Structural dynamics | Analytic fixed-free beam and same-input CalculiX B32 | Peak 8,150 Hz vs CalculiX 8,178.154 Hz; 0.3443% peak-frequency delta and 0.0233 normalized curve RMS. | REFERENCE PASS | Current public evidence is cantilever harmonic response against analytic and CalculiX references. |
| Static cantilever Hex8-EAS stress Static FEA | Bernoulli-Euler same-location Gauss-point stress reference | Stress error 1.4588e-2 at 20x4x2 and 5.1531e-3 at 40x8x4. | REFERENCE PASS | Current public evidence is integration-point stress evidence with mesh refinement. |
| NAFEMS linear-elastic core rows NAFEMS structural | NAFEMS published linear-elastic targets | Selected LE rows are within declared tolerances; additional families remain in validation. | REFERENCE ROWS | Current public evidence separates passing linear-elastic rows from modal and shell expansion rows. |
| NAFEMS FV4 offset-mass cantilever Modal dynamics | NAFEMS FV4 first-six-mode reference case | Target first-six-mode accuracy; direct shifted-solve routes are finite, but worst-mode error remains about 91% at the n10 rung. | VALIDATION TRACK | Mode-shape, model-form, and dense-solver comparison are the current validation focus. |
why it matters
More than a prettier contour plot: a reviewable engineering artifact.
The customer-grade surface is the Evidence Console: a skeptical engineer should be able to inspect the visual, open the evidence packet, check source geometry and assumptions, read solver/reference deltas, and see exactly how the evidence package expands next.
Open the organized demo board.
The site leads with the Boomerang flight-path replay, pairs it with solver diagnostics, and keeps the rest of the proof portfolio available by domain for engineers, customers, and reviewers.