Researchers propose a method for automatically generating both fixes and bug reproduction tests in software repair. This approach, called agentic program repair with cogeneration, allows developers to increase their confidence in AI-generated patches by including bug reproduction tests alongside the fix. The study evaluates this method on 120 human-reported bugs at Google and shows that it can generate bug reproduction tests for at least as many bugs as a dedicated test agent without compromising the generation rate of plausible fixes. --- Why it matters: This matters to researchers in AI because it demonstrates a practical application of agentic program repair, which has the potential to reduce engineering effort in software maintenance. The ability to automatically generate both fixes and bug reproduction tests could also improve the efficiency of software development pipelines. Source: https://arxiv.org/abs/2601.19066