Graph based interface helps developers oversee AI coding agents
Vibe-GUIDE: A Graph-based User Interface in IDEs for Oversight in Vibe Coding
Human-Computer Interaction
Summary
When developers use AI to write code, it can become hard to keep track of changes and understand the project. The authors created Vibe-GUIDE, a tool that shows the project as a live graph organized by functions to help developers see and control what the AI changes. In tests, developers using Vibe-GUIDE understood the project better and stayed more involved in guiding the AI than those who just chatted with the AI. This suggests that visual, structured representations can help people manage AI-assisted coding more effectively.
What this means in practice
- •For software development teams: Use graph based interfaces like Vibe-GUIDE to help developers maintain understanding and oversight when AI agents generate code changes during a project.
- •For ai assisted development tool builders: Incorporate persistent structural visualizations into coding tools to support developer involvement and comprehension alongside natural-language AI interaction.$Commercial implications: Enables creation of advanced development platforms with better transparency and control for developers over AI code generation.
Authors
Chifang Chou, Sam Yu-Te Lee, Rudrajit Choudhuri, Kwan-Liu Ma
Abstract
In agentic coding, developers shift from implementing changes themselves to specifying intent, evaluating the agent's work, and making approval decisions. However, delegating implementation can introduce cognitive debt that erodes project comprehension over time, constraining developers' ability to provide oversight. In this work, we investigate the role of persistent shared representations in supporting project comprehension and oversight of coding agents. We present Vibe-GUIDE, an agentic coding interface built around a structural, live, manipulable, and adaptive graph representation organized by functional modules. We evaluated our interface in a randomized between-subjects study comparing how 16 developers completed three cumulative coding tasks using our interface or a Chat-only baseline. We found that Vibe-GUIDE can support project comprehension and sustained task performance while keeping developers cognitively involved in oversight. These findings show how persistent shared representations can complement natural-language interaction and help developers maintain the understanding needed to oversee agent-generated changes as projects evolve.