/attempt I’m interested in the ART-E / ART bounty. I saw that the easy scope is described as an ART-E agent task, and the hard scope is 2-way portability between ART and verifiers. Before I start, can you confirm whether new easy-scope PRs are still being considered, or whether the current open PR already covers that part? If still open, I can keep the deliverable focused: a small reproducible example, deterministic task setup, tests, and clear README/run instructions.

/attempt I’d like to explore the SWE-Swiss bounty. Before going deep, can you confirm what would qualify for the $400 existing-RL-dataset scope? My understanding is that a valid smaller deliverable would package the existing SWE-Swiss dataset into a reproducible RL/verifier training workflow with clear loading, scoring, and run instructions, while the $800 scope would require a full auto-generation pipeline. If that’s right, I can start with the smaller scoped implementation and keep it easy to review.