AI-native spatial engineering
A million live actors. One browser tab.
A million moving bodies, simulated and rendered live.
- Simulated on the CPU — a million bodies, sixty times a second, on the vector units (Rust · SIMD).
- One shared memory — sim, render, and input threads read the same block; nothing copied between them.
- Drawn on the GPU — instanced through WebGPU, at your display's full refresh rate.
- One format, every job — laid out as Apache Arrow, so the world runs, saves, streams to any data pipeline, and answers questions in plain SQL. No export step.
The games are the live demo; the same engine white-labels into fleet, logistics, and satellite mission-control dashboards.
- 1,000,000actors
- 60Hz sim
- ∞fps · monitor
- Apache Arrow · SIMD · WebGPU
Live demo
Proof runs in a browser tab.
play.aeone.games is the engine running live — the flagship simulating a million live actors in a single browser tab, plus real multiplayer sessions over WebRTC, no install. Not a trailer, not screenshots. Open one and watch the frame budget hold.
LIVE Launch a live session Open a tab. A million actors, simulated live. No install.The stack
One engine — and the AI-native way we ship it.
A browser-native GPU/WASM spatial engine, built and evolved by LLM agents on a codebase engineered to be read by them. The tech, and the reason a small team ships it fast.
Stardust Engine
Browser-native GPU + WASM spatial engine. Rust/WASM simulation, WebGPU rendering, TypeScript — large, moving worlds at full frame rate, zero install. Built by LLM agents; white-labels into enterprise dashboards.
Explore the engine How we buildContext Engine
The bet: the bottleneck in LLM-driven work is loading the right context, not typing code. This is the machinery that makes it pay.
- absorption-spec ▸ Mass + skill transfer on contact 0.95
- moon-lifecycle-spec ▸ Orbit → fire → absorb states 0.91
- stardust-mass-model ▸ Mass is HP; nothing is lost 0.88
The devlog
How the engine works — how the AI builds it.
The shared-memory simulation, the WebGPU render path, and the Apache Arrow data model — down to why the whole world is queryable in plain SQL — plus how AI agents write and evolve the whole thing.
- Jul 13, 2026 One format, four jobs — how Apache Arrow made my simulation queryable in SQL
- Jul 7, 2026 A million entities, zero copies — the shared-memory trick behind a full-rate browser sim
- Jun 25, 2026 AI doesn't have memory. Now it does — here's how we built it.
Work with us
Put the engine on your data.
Browser-native digital twins and mission-control dashboards — your live fleet, drones, vehicles, or satellites rendered at scale, zero install. It ingests Apache Arrow — the open standard your data stack already speaks — so your existing pipeline feeds it directly. A fixed-scope feasibility spike first, then a dashboard your team owns.