AI workflow
in buildAI helps set up cases, read references, critique assumptions, generate tests, and diagnose solver behavior. Benchmarks carry the physics claims.
Anvil Sim is engineering simulation software for FEA, CFD, thermal, electronics, and geometry workflows. The product loop is simple: research, setup, solver run, benchmark comparison, visual replay, reviewer packet, and the next validation step.
AI helps set up cases, read references, critique assumptions, generate tests, and diagnose solver behavior. Benchmarks carry the physics claims.
Source geometry, units, source-face binding, mesh quality, and remesh stability move through explicit intake gates.
Structural FEA, modal, thermal, FRF, electronics, and CFD work progresses through narrow kernels, analytic checks, reference solvers, and public benchmark rungs.
Heat-sink, enclosure airflow, and future AI chip package cases attach power maps, material assumptions, temperature QoIs, and reference data as the evidence grows.
Each public claim should have a benchmark row with reference source, QoI, tolerance where available, status, and next step.
The Mac app is the review surface: visual replay, solver fields, benchmark deltas, provenance, evidence packets, and validation steps.
A reviewer packet binds the claim, input assumptions, solver output, reference comparison, visual replay, current validation focus, and next evidence step.
Each line shows how today's public evidence grows into broader product capability.
| Horizon | Focus | Evidence today | Validation focus |
|---|---|---|---|
| Current | Organized public proof surface | Boomerang flight-path replay, benchmark wall, investor route, whitepaper, and domain-grouped demos. | Keep every demo connected to a status row and validation note as public language expands. |
| Next | Benchmark closure depth | FRF, static FEA, NAFEMS core rows, flame reference calibration, CFD targets, and geometry/mesh validation focus rows. | Target and roadmap rows move through executable closure before they become stronger proof claims. |
| Then | Native review workflow | Evidence packets and visual replay move into a repeatable Mac review workflow for demos, benchmarks, and customer-style packets. | The native UI must keep visual readiness separate from solver and benchmark readiness. |
The full wall keeps every row explicit: reference source, QoI, status, latest evidence, validation focus, and next benchmark step.
SCENARIO EVIDENCE
QoI: Trajectory enters the 0.50 m return window while airborne before first ground contact.
next: Attach measured-flight or public aerodynamic reference data before expanding the flight-validation claim.
REFERENCE PASS
QoI: Peak 8,150 Hz vs CalculiX 8,178.154 Hz; 0.3443% peak-frequency delta and 0.0233 normalized curve RMS.
next: Add another geometry and a physical/reference sweep before expanding customer-style FRF readiness.
REFERENCE PASS
QoI: Stress error 1.4588e-2 at 20x4x2 and 5.1531e-3 at 40x8x4.
next: Extend to source-bound geometry, mesh-convergence receipts, and an external solver comparison.
REFERENCE ROWS
QoI: Selected LE rows are within declared tolerances; additional families remain in validation.
next: Separate each passing LE row from mixed modal/shell families and keep tolerances visible per case.