SafeRE: Building a production-quality regex library with agents (eaftan.github.io)

The launch of SafeRE marks an innovative step in the AI/ML field as it introduces a production-quality, linear-time regular expression library for Java, developed primarily by AI agents. The project's aim is to streamline regex processing while avoiding common pitfalls associated with traditional backtracking engines, such as regular expression denial of service (ReDoS). By utilizing agents to handle the bulk of the coding process, the creator, who previously worked on the Java team at Google, seeks to demonstrate the potential of AI in tackling complex software engineering challenges without heavy human intervention. While the technical foundation is rooted in existing libraries like RE2, the journey to ensure correctness has proven to be intricate, requiring extensive validation techniques and testing methodologies. SafeRE aims to be a drop-in replacement for Java's standard regex library but must navigate the nuances of behavior and implementation expected by developers. Key testing strategies have included leveraging open-source test suites, substituting SafeRE in large Java projects, and employing exhaustive regex sweeps. This project not only explores the practical applications of AI agents in software development but also raises important questions about system correctness and collaboration between human engineers and AI-driven code generation. The library is open-source and available for public testing on GitHub.