Codex Open-Sources AI models via Hugging Face Skills integration

AI Impact Summary

OpenAI Codex now integrates with Hugging Face Skills (HF-skills) to enable end-to-end ML experiment workflows, including fine-tuning, RL alignment, live-tracking of training metrics, evaluation, reporting, and export to GGUF for local deployment. This shift can substantially automate model development pipelines, allowing Codex to orchestrate tasks across datasets (e.g., open-r1/codeforces-cots), benchmarks (openai_humaneval), and HUB publishing, potentially reducing manual toil for engineers. Teams should plan for governance around data usage, compute costs, MCP/server configurations, and publishing controls as these automated workflows increasingly influence model training and deployment cycles.