Reproduce “Choose Your Own Adventure: Non-Linear AI-Assisted Programming with EvoGraph” 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
Vassilios Exarhakos, Jinghui Cheng, Jin L.C. Guo
EvoGraph, an IDE plugin, introduces a non-linear, graph-based interface for AI-assisted programming, addressing the challenges of managing iterative development and code provenance. A user study found that developers experienced reduced cognitive load and found it significantly easier to explore alternative solutions and trace code changes compared to linear chat interfaces.
Key insights
- A graph-based interface for managing AI-assisted programming sessions significantly reduces cognitive load and encourages safer, more comprehensive exploration of alternative solutions.
- Integrating micro-level versioning and explicit code provenance directly into the development workflow supports developer comprehension and critical reflection on AI-generated code.
- Providing the AI assistant with a structured overview of the full development history, including past prompts and code states, improves the relevance and continuity of AI responses across extended interactions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.