OPEN, CUSTOMER-OWNED SIMULATION

Build simulations faster. Own the toolchain.

We are developing RealityLoop to help engineering teams create and adapt simulations, explore more test scenarios, and verify products virtually before committing to physical tests. Our goal is sufficient fidelity to make those results useful for engineering decisions. That fidelity must be demonstrated against real-world measurements, not assumed from a convincing visualization.

Test more scenarios

Run repeatable experiments and parameter variations, locally or in CI. Compare virtual results with physical measurements to establish where the model can be trusted.

Build and change quickly

Give engineers and AI agents a modular foundation they can adapt, rather than rebuilding each experiment from scratch. Run headlessly across supported environments.

Engineering foundations

Build on Modelica, an established modeling language used across industries, with explicit physical models, reproducible runs and auditable results. The foundations are developed by engineers, not hidden behind a black box.

Expert updates, without dependency

Benefit from improvements by the people developing the foundation—or maintain and extend your own copy. Future updates should be an option, never a condition for using your simulations.

RealityLoop is still experimental. The examples below demonstrate building blocks and research workflows; they are not yet validated, production-ready digital twins.

EXAMPLE GALLERY

Experiments

Independent research demonstrations
MODULAR ROBOTICS & SPACE SYSTEMS

Other experiments

Early-stage research concepts
OWNERSHIP · VERIFICATION · OPTIONAL INTEGRATION

Your simulation. Your source code. Your feedback loop.

RealityLoop is intended as a foundation that engineering teams take into their own repositories, adapt to their products and operate independently. The goal is to bring the fast, repeatable feedback loop of compilers and unit tests to systems that must also be tested against simulated reality.

100% customer-owned simulation

You own the simulation you develop—including its source code, models, scenarios, tests, data and CI/CD setup. You can inspect, change, run and maintain the toolchain yourself, without a Taskyon account or hosted runtime. The underlying building blocks and dependencies retain their own licenses.

A feedback loop for AI agents

Change a design, run a repeatable simulation, evaluate assertions and metrics, then feed the results back to engineers or coding agents—much like compilation and unit tests in software development. Versioned inputs, reproducible artifacts and headless CI runs make this practical. Results are only as trustworthy as the models validated against reality.

Taskyon is an optional extension

RealityLoop already uses the same underlying Taskyon Design Graph execution and reproducibility format independently. The intended optional Taskyon connection builds on those artifacts for integrated design automation, parameter exploration, performance analysis and coordinating or monitoring parallel simulation runs across CI/CD and clusters. You retain ownership and can keep running without Taskyon.

Available today: independent local and headless simulations, Design Graph artifacts, reproducibility checks and replay verification. Not yet integrated: Taskyon UI, remote cluster orchestration, centralized monitoring and authorized P2P publication.

npm run sim -- scenarios/modelica-wheel-breakaway.json runs/wheel