Horizon Labs Report
Summary
Horizon Labs presents an eight-week case study of running Horizon, an autonomous agent, with full access to a blank Hugging Face account to measure real-world adoption and model development. The report details the creation of 38 models, extensive usage across GitHub repositories, and the economics of training and hosting these models, while examining guardrails, security considerations, and potential misuse. It also discusses limitations, workflow, and data transparency, including an appendix with raw data and tool calls.