Skip to content

NobleJoy

Argon Game Studio (NobleJoy) is an AI-assisted 3D game studio and robotics pipeline that connects a native Rust and Bevy 0.14 driving engine with a Python, FastAPI, and MuJoCo physics backend.

Code Video Cerebral Valley

Core trick

A multimodal QA agent reads a live viewport screenshot and patches colliders and MuJoCo motor torque.

Tags

creative/musicrobotics/hardware

Description

Argon Game Studio (NobleJoy) is an AI-assisted 3D game studio and robotics pipeline that connects a native Rust and Bevy 0.14 driving engine with a Python, FastAPI, and MuJoCo physics backend. Normally, when you use AI to generate 3D game levels or robot simulations, things break right away—colliders are missing so cars phase through walls, overlapping meshes cause flickering Z-fighting on the road, obstacles spawn in the middle of the lane, and the physics aren't balanced enough to build the robot in real life. We built Argon Game Studio so you can actually drive around a level in real time using WASD while prompting the backend to restyle the world on the fly (like turning a cyberpunk street into a medieval kingdom), swap PBR car skins, and morph your vehicle between a 4-wheel rover, a bipedal mech, a spider crawler, and a tank without resetting your position. When you run into broken level geometry or missing colliders, clicking our "Observe & Fix" button runs a WGPU compute shader check alongside a multimodal QA agent that looks at a live screenshot of the viewport and patches the colliders, moves roadblocks out of the way, and tunes the MuJoCo motor torque automatically. Once your rover works in the simulation and passes center-of-mass and servo torque checks, you can export the actual 3D-printable M3-calibrated STL files, URDF, bill of materials, and ESP32-S3 MicroPython code to build it physically.
jorgeajimenez/NobleJoy★ 00 forksPythonCreated Oct 25, 2026 · last push Oct 26, 2026

README excerpt

# Argon Game Studio (NobleJoy) — Autonomous AI Game Studio, MuJoCo Rover Physics & Native Rust Engine

**Target:** Gemini 4 Argon SF Hackathon  
**Track:** Track 2 — More Dimensions  
**Prize Target:** Best Use of Managed Agents  

---

## Team Work Split & Independent Project Guides

We have structured the monorepo so **Gretchen** (`backend/`) and **Jorge** (`game_client/`) can develop and test completely independently or connected together:

1. **[`backend/README.md`](./backend/README.md) — Gretchen's Autonomous Studio & MuJoCo Physics Service**
   - Gemini 4 Argon Managed Agents (`antigravity-preview-09-2026`), `gemini-nano-banana-2.1` vehicle skin generator, `/workspace` `.tar` artifact puller, MuJoCo 4-wheel rover simulation (`mujoco_sim.py`), and physical manufacturing exporter (M3 STLs + `rover.urdf` + `BOM.csv` + MicroPython firmware).
   - Includes standalone test script (`uv run python test_studio_pipeline.py`) and `curl` commands so Gretchen can test everything without compiling Rust.
2. **[`game_client/README.md`](./game_client/README.md) — Jorge's Native Rust / Bevy 0.14 3D Driving Engine**
   - Hardware-accelerated 3D rendering on Apple Silicon (`wgpu` Metal), 4-wheel articulated rover + 2-DOF actuator arm, live PBR skin loader, viewport PNG screenshot capture, and HUD telemetry.
   - Includes **Standalone Offline Preview Mode** so Jorge can run `cargo run` and drive/test immediately even when the Python backend is offline.

---

## Architecture Overview

![Argo

Team