⚡ Blackwell Forge

Autonomous node manager for Vast.ai GPU compute infrastructure.

This project exposes the core control loop behind production AI infrastructure: observe → decide → act → wait

🎯 What it is

A minimal, readable implementation of dynamic GPU pricing based on hardware state:

• Temperature-aware pricing (thermal safety through economics)
• Utilization-based adjustments (market-responsive)
• Hardware-first design (GPUs are first-class concerns)

⚠️ What it is NOT

Not an autonomous agent, not a learning system, not a goal optimizer. See NON_GOALS.md for explicit constraints.

→ View on GitHub → Download v1.1-clean

📚 Documentation

BOUNDARIES.md - System role definition
ARCHITECTURE.md - Technical structure
CONTRIBUTING.md - How to contribute

---

Built for dual RTX 5090 compute nodes. Licensed under MIT.