What CI looks like at a 100-person team (www.mendral.com)

PostHog recently announced the launch of Mendral, an AI agent designed to address continuous integration (CI) challenges within their large-scale engineering team, which manages an extensive public monorepo. Last week, PostHog's CI processed 575,894 jobs, executed over 33 million tests, and generated an impressive 1.18 billion log lines. Despite a high pass rate of 99.98%, the sheer volume of tests led to significant noise due to flaky tests, prompting the development of a tool that can efficiently diagnose failures, quarantine inconsistent tests, and automate the process of opening pull requests for fixes. Mendral's technical capabilities include ingesting vast amounts of log data at high speed, correlating test failures with specific code changes, and intelligently notifying team members based on relevance to mitigate distractions. By analyzing patterns over time, the agent proactively enhances CI health while maintaining transparency through its public repository engagement. The introduction of Mendral signifies a crucial evolution in how teams can manage CI at scale, especially as AI coding tools increase the frequency of code changes. This not only streamlines the development process but also ensures that engineers remain productive amid growing testing demands, making Mendral a promising solution for the AI/ML community facing similar challenges.