Autonomous agent runs an AI company for 7 weeks: 38 models, 50k downloads on ~$1,030
maksym_andr · x · 2026-11-18
An experiment let Horizon, an agent based on Opus 5.5, autonomously run an AI company for seven weeks with a blank Hugging Face account and compute. Running 24/7 in Claude Code on a single Team Premium subscription with 500 GPU-hours/week in credits, it burned 600+ GPU-hours, created 38 models used in 100+ GitHub repos with 50,000+ downloads, and processed 1.9B tokens (96% prompt-cache hits, $1,030 at API prices). Observers call it an impressive preview of current autonomous agent capabilities worth scaling.
More from coding & agent
- FastMCP 4.2 ships Skills over MCP — drag a skill folder, get an MCP server instantly — HanchungLee · 2026-11-18
- Open-Source Local Agent Observer Locks Your Phone to Stop Doomscrolling — Roy3838 · 2026-11-18
- Amazon paper: weak LLM critics can sink search loops, ideal gain only 0.005 AUC — rohanpaul_ai · 2026-11-18
- OpenAI engineers secretly use Astra for Law to write PRDs and design docs — sherwinwu · 2026-11-18
- START-HERE.md: Mapping Your Vault So Agents Find Files Without Burning Tokens — tomcrawshaw01 · 2026-11-18
- AnyMachine turns a single painting into a 160-part walking robot rig built by an agent in Blender — bennash · 2026-11-18