Vinci
Physics foundation model for continuous design verification · Vinci4D
Vinci builds a physics foundation model that provides continuous physics reasoning throughout the hardware design process, embedding physical intelligence from day one rather than as end-stage validation. Their system computes physics predictions in real-time and emphasizes trust through traceable, non-hallucinating results that don't require customer data for training. Like Neuralectic, they target engineers working on complex physical systems where every calculation has downstream manufacturing consequences.
Why: Very strong match on verification-first approach and trust mechanisms. Differs slightly in wedge—focuses on continuous physics throughout design rather than specifically augmenting existing artifacts. Both emphasize traceable, physics-based validation over statistical confidence.
ColibriV
Solver-verified engineering design agent · ColibriV
ColibriV Design AI is an engineering agent that explores design spaces overnight, running real physics simulations and delivering ranked designs with full evidence trails. Built initially to design their own hydrogen engine, it emphasizes solver verification at every step. The system targets physical hardware teams needing verified designs with traceable evidence, using real physics simulation as the verification backbone.
Why: Excellent match on verification-as-core and evidence trails. Strong on search/refinement pattern with physics verification guiding exploration. Wedge is slightly different—focuses on design space exploration rather than augmenting existing engineering artifacts.
Simulyx
Deterministic-check-gated FEA agent · Simulyx
Simulyx is a simulation agent that refuses to execute solve commands until deterministic checks pass, keeping engineers in the loop for approval at every step. It runs FEA from CAD to verified report using battle-tested open-source solvers on the user's own machines. The system emphasizes deterministic verification as a gate before computation, with engineers approving proposals in text or 3D view.
Why: Very strong on deterministic verification as a gate and CAD-as-entry-point wedge. Explicitly blocks execution until checks pass, similar to Neuralectic's verification-guided approach. Slightly narrower scope (FEA specifically) than multi-component system integration.
Deep Design Systems
AI decision infrastructure with simulation and evidence traceability · Deep Design Systems
Deep Design Systems builds AI decision infrastructure for complex problems in early industrial product development, connecting simulation, evidence and traceability to evaluate alternatives before decisions become expensive to change. They focus on early-stage decisions where teams have the least information but are making choices that determine program cost and performance. The emphasis is on traceable evidence chains through simulation.
Why: Strong match on traceability and evidence-based decision infrastructure. Focuses on early-stage decisions with simulation backing. Similar buyer persona (engineers on complex industrial products). Less explicit about deterministic verification as RL signal.
Standard Physics (Archimedes)
Physics-verified design agent with solver-in-loop · Standard Physics
Archimedes uses AI agents that write designs as code, test every one in a real physics simulator, and iteratively improve the best candidates. The system explicitly states when results don't hold up in verification. It runs swarms of agents through design-test-refine loops with real physics simulation as the verification layer, targeting physical hardware design problems.
Why: Strong on verification-guided refinement and physics-in-loop architecture. Uses real simulator to test and improve designs iteratively. Wedge differs—generates designs from scratch rather than augmenting existing engineering artifacts. Similar search-and-refine pattern.
Lever
Physics-aware executable model for early hardware decisions · Lever
Lever is a physics-aware engineering system for proposal and concept development, turning requirements, assumptions, architecture and physical constraints into an executable model. It targets the earliest phase when teams have the least information but are making decisions that determine program cost and performance. The system makes early hardware decisions executable by connecting requirements to physical constraints.
Why: Good match on wedge—uses requirements and architecture as entry point. Focuses on early-stage decisions with physics constraints. Less explicit about deterministic verification or RL fine-tuning architecture. Strong on requirements-to-constraints traceability.
Dr.Q
In-house FEM solver with deterministic AI agent · Dr.Q
Dr.Q built their own FEM solver natively driven by an AI agent, emphasizing deterministic, self-checking simulation in the browser. The system is designed for the agent era with traceability down to every assumption, allowing design teams to run routine cases themselves. They position as 'not a copilot on legacy CAE' but simulation built for deterministic agent control.
Why: Strong on deterministic, self-checking architecture and traceability to assumptions. Built custom solver for agent control. Wedge differs—targets routine cases for broader team rather than augmenting expert engineer workflows. Good match on trust mechanism.
Vecteur
Spacecraft design verification system · Vecteur
Vecteur focuses on spacecraft design verification, addressing the problem that a single system parameter can take a year to verify across specialists and errors may only surface in orbit. The founder built it so engineers can check a design before it costs the program. It targets the multi-specialist verification problem in space systems where physics limits and interface compatibility are critical.
Why: Excellent match on verification-as-core-product and high-stakes design decisions where errors are catastrophic. Strong on multi-component system integration. Less detail on AI/agent architecture or RL fine-tuning approach. Domain-specific (space) but similar problem shape.
