News Summary

• Autofoundry is a CLI companion to Karpathy's autoresearch.
• Point it at a shell script, pick your GPU configuration, and it handles the rest: provisioning instances, distributing experiment runs, streaming results live, and producing a final metrics report.
• Supported on RunPod and Lambda Labs.
• Run any ML experiment script across GPUs on multiple cloud providers with a single command.
• On first run, walk you through configuring provider API keys, SSH key path, default GPU type, minimum download bandwidth (default 5000 Mbps — filters out slow Vast.ai hosts), and HuggingFace token.