The engine
Stardust Engine
A browser-native, GPU-accelerated spatial engine — Rust compiled to WebAssembly for the simulation, WebGPU for the rendering, TypeScript for the app. Large, moving worlds at full frame rate, zero install. Written and evolved by LLM agents on a deliberately agent-legible codebase — the engine under the games, and the engine we white-label into enterprise mission-control dashboards.
- 1,000,000actors
- 60Hz sim
- ∞fps · monitor
Capabilities
GPU-accelerated spatial rendering
WebGPU shaders push the rendering of a whole moving scene straight to the client's graphics card, bypassing the DOM and 2D-canvas bottlenecks that choke traditional web maps past a few hundred markers. Instanced rendering draws many distinct assets — ships, drones, trucks, satellites — in a single draw call by streaming per-instance position and status data to the GPU.
WebAssembly simulation core
The simulation is written in Rust and compiled to WebAssembly, running off the main thread at near-native speed over a shared-memory data layout with no per-frame allocation. Millions of position and telemetry updates per second without the garbage-collection stutter that sinks a JavaScript-only simulation. Deterministic by construction — the same inputs replay to the same state.
Speaks Apache Arrow
Data comes in as Apache Arrow — the open columnar standard the modern data ecosystem is built on: Parquet files on disk, Arrow Flight over the wire, and the dataframe and warehouse tools your pipeline already runs. The engine ingests it straight into the live simulation, so an existing telemetry feed or data lake plugs in with no bespoke import layer. It's the interchange standard your engineers already know — not a proprietary format that locks you in.
Zero-install, browser-native
The whole engine is a secure URL. No desktop install, no plugin, no middleware — it loads instantly on any modern laptop, which is exactly what enterprise security and compliance teams want and what heavy native command software can't offer. The same runtime powers a game tab and an operations dashboard.
Games as the proving ground
Every capability ships first inside a real game, at real scale, under a real frame budget — then white-labels into a client engagement: strip the game content, plug in the live telemetry feed, keep the pipeline. The games are the public proof that the engine renders large, moving worlds smoothly in a browser.
Built by agents, by design
The engine is written and maintained by LLM agents operating on a deliberately agent-legible codebase — roughly 35% comment density and a design corpus that runs nearly one prose-token per code-token. The bet: in LLM-driven development the bottleneck isn't typing code, it's loading the right context. Our CLAUDE.md + skills + context-engine apparatus is how a small team ships and evolves an engine at this scope.
Built with
Work with us
Put the engine on your data.
Browser-native digital twins and mission-control dashboards — your live fleet rendered at scale, zero install. A feasibility spike first, then a build your team owns.