Coursera Spring Sale
40% Off Coursera Plus Annual!
Grab it
Explore the evolution of continuous integration when artificial intelligence components enter software development in this 15-minute conference talk. Learn how traditional deterministic testing approaches must adapt to accommodate the inherently stochastic nature of AI systems, moving beyond binary pass/fail outcomes to embrace uncertainty as a fundamental characteristic. Discover practical strategies for building comprehensive test suites for AI components, including the concept of "stochastic unit tests" that verify system behavior without requiring absolute determinism. Examine how to shift from red/green testing results to nuanced "shades of gray" evaluations, replace traditional build gates with trend analysis and performance monitoring, and maintain engineering confidence despite statistical variance. Gain insights into thoughtful approaches for incorporating AI evals into existing CI pipelines while preserving code quality standards. Drawing from real-world experience building Zed, a next-generation native IDE in Rust that runs at 120 frames per second, understand how teams can successfully navigate the paradigm shift from eliminating non-determinism to embracing probabilistic outcomes in AI-powered software systems.
