Download and read the papers before the MiniPLoP

This paper describes the Continuous Delivery for Artificial Intelligence (CD4AI) pattern, a strategy to balance deployment velocity and rigorous validation in AI-enabled systems. The pattern establishes a three-stage lifecycle for these systems: (1) testing: a Continuous Integra- tion / Delivery (CI/CD) step combining automated tests and probabilis- tic evaluations, allowing new versions into production if they succeed at a minimum rate (e.g., 95% success rate); (2) monitoring: a monitoring step that collects real-world interactions for later analysis, flagging po- tential anomalies via automated heuristics; and (3) curation: a feedback loop step to gather candidate interactions that can be converted into new tests, incrementing the regression suite. By leveraging production observability as a source of truth for test generation, this pattern enables teams to maintain high deployment velocity while automatically evolv- ing their test suites to match the complexity of real-world usage. This ensures both the testability and robustness of the system over time.