Show HN: MaragingLoop: Autonomous Bare-Metal OS Agent
MaragingLoop is an experimental AI agent designed for autonomous bare-metal operating system and kernel development using a ReAct-style loop. It integrates large language models with automated compilation, VirtualBox VM control, and vision-based screenshot analysis to iteratively generate, test, and refine low-level C code. The system operates locally and emphasizes a closed-loop development process that combines code generation with real-world execution feedback.
- ▪MaragingLoop uses an LLM to write, compile, and test OS kernel code in a VirtualBox VM until the system boots as specified.
- ▪The agent captures visual feedback via screenshots and uses it to refine subsequent code iterations, creating a closed-loop development cycle.
- ▪The project was developed locally in 10 days using the Qwen3.6-35B model via llama.cpp, with the core logic contained in a single ~900-line Python script.
- ▪Users must manually configure a VirtualBox VM named 'agentos' and set up boot options to use the generated OS ISO.
- ▪Current limitations include non-functional mouse support and incomplete network driver implementation for NAT with port forwarding.
Opening excerpt (first ~120 words) tap to expand
MaragingLoop Autonomous Bare-Metal OS Agent | 📝 Code → 🖥️ VM Interact → 👁️ Inspect → 🔨 Harden 🔍 Overview MaragingLoop is a ReAct-style AI agent engineered for autonomous bare-metal OS and kernel development. By orchestrating large language models with automated compilation, robust VirtualBox VM lifecycle control, and vision-based screenshot inspection, it iteratively writes, builds, tests, and refines low-level C code until the target system boots and behaves as specified. Unlike text-only coding agents, MaragingLoop treats code generation and VM interaction as equally critical pillars of its development cycle.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at GitHub.