Everything reality can do, in one engine
Poly Sim OS collapses the walls between physics, biology, chemistry, and math into a single AI-powered canvas. This is the complete picture of what the Everything Engine unlocks.
5
Domains, one canvas
5B+
Particles per cloud run
60 fps
Real-time WebGPU render
0
Installs required
The Domains
Five disciplines that finally speak the same language
Each domain ships a deep library of typed, unit-aware nodes. The magic is that any node can connect to any other — across every field.
Physics
Rigid bodies, fields, and fluids resolved to the limit of your hardware.
- Navier–Stokes fluids
- N-body gravitation
- Rigid & soft bodies
- EM field solvers
Biology
From single-cell kinetics to population dynamics, modeled as living graphs.
- Enzyme kinetics
- Gene regulatory nets
- Lipid nanoparticles
- Tissue diffusion
Chemistry
Reaction networks and transport coupled directly to physics and biology.
- Reaction-diffusion
- Stiff kinetics
- Phase transitions
- Catalysis models
Mathematics
The solver layer — pure math nodes that drive every other domain.
- ODE / PDE solvers
- Stochastic processes
- Optimization
- Spectral methods
Solvers & Energy
Adaptive integrators that auto-tune to the stiffness of your system.
- Runge–Kutta RK45
- Implicit BDF
- Finite element
- Symplectic integrators
…and the spaces between them
Biophysics. Electrochemistry. Computational fluid biology. The most important science happens at the seams — and the seams are where Poly Sim OS lives.
Groundbreaking Capabilities
Six things no other tool puts in one place
Multiphysics Coupling
Wire a fluid-dynamics node straight into a nanoparticle node, driven by a math solver. Domains that never spoke before now share the same canvas, units, and time-step.
AI Graph Synthesis
Describe a system in plain English and a model fine-tuned on arXiv and PubMed assembles validated nodes, variables, and formulas — never inventing physics, only assembling proven building blocks.
Near-Native WebGPU
A WebGPU + WebAssembly engine renders high-fidelity 2D/3D output in real time, at desktop-class speeds, with zero installs and nothing to maintain.
Typed Node Contracts
Every port carries a type and a physical unit. A deterministic validator checks dimensions, cycles, and solver fit before a single step runs — so graphs are correct by construction.
Open Community Fabric
Publish, share, and fork complex node clusters. A grad student's quantum-superposition cluster becomes a building block for a pharma lab the same afternoon.
Live Data Inspector
Real-time graphs, heat maps, and raw matrices stream as the simulation runs — instant error-checking instead of post-hoc guesswork.
In Practice
What you'll be able to simulate
A taste of the node clusters that turn a sentence into a running model.
Double pendulum
prompt ▸ Model a double-pendulum under chaotic gravitational interference
solver ▸ RK45 adaptive · rtol 1e-6 · symplectic option
Lipid nanoparticle
prompt ▸ Simulate lipid nanoparticle delivery through laminar blood flow
solver ▸ Implicit ODE (BDF) · Crank–Nicolson transport
Krebs cycle
prompt ▸ Build a kinetic model of the Krebs cycle with enzyme feedback
solver ▸ Stiff BDF (Gear) · adaptive order 1–5
Navier–Stokes
prompt ▸ Visualize turbulent Navier-Stokes flow around an airfoil
solver ▸ FEM · k–ε turbulence · SUPG stabilization
Numerical Accuracy
Provable convergence, not vibes
Every integrator ships with a known order of accuracy. Switch methods below and watch the global error collapse as — measured live against an analytic solution.
Global error scales as for a method of order . Halving the step size should cut the error by .
error(h) — log scale
| Method | Order p | Err @ h=1/64 | f-evals/step |
|---|---|---|---|
| Forward Euler | 1 | 7.7e-3 | 1 |
| Midpoint RK2 | 2 | 4.0e-5 | 2 |
| Runge–Kutta RK4 | 4 | 9.1e-10 | 4 |
| Dormand–Prince RK45 | 5 | 2.3e-12 | 6 |
| Particles | Backend | Steps/s | Frame |
|---|---|---|---|
| 1e4 | WebGPU local | 1,920 | 0.5 ms |
| 1e6 | WebGPU local | 144 | 6.9 ms |
| 1e8 | Cloud A100×4 | 61 | 16 ms |
| 5e9 | Cloud A100×64 | 12 | 83 ms |
* Figures from reference kernels on Apple M2 (local) and NVIDIA A100 80GB (cloud); indicative, hardware-dependent.
The Compute Fabric
Free on your hardware. Limitless in the cloud.
Standard runs render locally on your CPU/GPU through WebGPU — costing nothing. When a run outgrows the browser, Poly Sim OS escalates seamlessly to serverless GPU clusters, metered by Compute Tokens and never trusted to the client.
Local execution
WebGPU + Wasm · near-native speed
Cloud escalation
Serverless AWS / GCP clusters
Metered & secure
Server-side Compute Tokens
The Roadmap
The Unified Canvas
Visual node graph, live render viewport, and the AI Co-Pilot vision — the foundation everything else snaps onto.
Live AI Synthesis
Wire the Co-Pilot to a model so prompts generate real, validated node graphs on the fly.
Cloud Compute Fabric
Seamless escalation to serverless GPU clusters, metered by Compute Tokens, for billion-particle runs.
Air-Gapped Enterprise
Dedicated bare-metal clusters and on-prem deployment for pharmaceutical and aerospace R&D.
The most ambitious simulation engine ever built for the browser
Explore the workspace today. The cloud, the live AI, and the enterprise fabric land as we grow.