Training open models with RL is now possible across multiple coding environments like Claude Code

Core summary

One sentence to understand this update

Ben Burtenshaw announces that open models can now be trained with Reinforcement Learning (RL) within various coding environments such as Claude Code, codex, opencode, or Pi, using unmodified harnesses.

Impact & opportunity

What this could mean

This advancement simplifies the RL training workflow for developers, allowing for greater flexibility and efficiency when experimenting with and fine-tuning open AI models across different platforms.