Type: url
Original: https://arxiv.org/pdf/2604.09805
Status: generated
Summary
“Building an Internal Coding Agent at Zup: Lessons and Open Questions” (Pinto et al., Zup Innovation) is an experience report on CodeGen, an internal CLI-based coding agent built around a manually implemented agentic loop, a FastAPI backend and an orchestration engine called Maestro. It reports 13 design decisions and argues that tool design — such as string-replacement edits instead of full-file rewrites and a read-before-edit policy — and layered, cross-tool safety guardrails improved reliability more than prompt engineering, while progressive human oversight (approval mode first, then autonomous mode) drove organic adoption. It closes with six open questions on tool manifest design, reasoning boundaries, cross-tool safety policy, adaptive trust, long-term memory and quality assurance for agent-generated code.