Search papers, labs, and topics across Lattice.
This paper identifies the issue of "context rot" in AI-assisted software development, where persistent context provided through configuration files becomes outdated as software evolves. By leveraging existing tools designed for checking consistency between documentation and code, the authors propose a research roadmap to address this problem in the context of AI configuration artifacts. Their preliminary analysis reveals that 23.0% of a sample of 356 repositories contain stale references, highlighting the relevance of traditional documentation consistency methods in mitigating context rot.
Stale context in AI configuration files affects 23% of software repositories, revealing a critical vulnerability in AI-assisted development workflows.
Developers increasingly provide AI coding assistants with persistent context through configuration files such as CLAUDE.md, AGENTS.md, and .cursorrules. These files describe code elements, architecture, and development conventions, forming the context that guides AI tool behavior across sessions. As software evolves, this context can become stale, a phenomenon we call context rot. While AI configuration artifacts are new, the underlying consistency problem connects to decades of software documentation research. Researchers have built tools to check consistency between documentation and code, spanning README files, code comments, API documentation, architecture descriptions, and installation instructions. We argue that this existing toolbox is an immediate starting point for detecting context rot, and we present a research roadmap mapping documentation consistency approaches to corresponding problems in this new setting. As preliminary evidence, applying an existing README/wiki consistency checker to a statistically representative sample of 356 repositories identifies stale code element references in 23.0% of repositories, showing that traditional documentation consistency tools can already surface context rot.
