Reproduce “The Work Behind Delegation: A Framework for Supervising AI Coding Agents” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yeon Su Park, Nadia Arvi, Hae Ri Lee, Sehoon Lim, Qianou Ma, Juho Kim
A qualitative study of experienced developers introduces a seven-stage framework for describing how people supervise AI coding agents, from planning and monitoring to reviewing, correcting, and saving reusable guidance. Analysis of observed work and 102 online discussion threads shows that delegation shifts rather than removes human effort: developers continue to manage alignment with intent, code quality, and their own understanding of changes.
Key insights
- The seven stages are Plan, Monitor, Wait, Review, Teach, Manual Fix, and Update Assets; workflows move iteratively among them rather than following a fixed sequence.
- Supervision repeatedly addresses three concerns: keeping work aligned with intent, enforcing standards beyond basic functionality, and understanding code developers did not write themselves.
- Developers manage this work by investing in plans, distributing checks across agents and automation, and recording repeated guidance in reusable project assets.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.