A detailed account of modernizing a 20-year-old Java 1.5 codebase using AI assistance, Docker containerization, and incremental refactoring. The author describes four phases: forensic analysis using targeted AI prompts to uncover structural rot, containment via a Docker 'time capsule' that recreated the 2008 environment without changing source code, lifting the artifact to Java 8 and Gradle 7.6, and finally refactoring to JUnit 5 and type-safe generics. Key lessons include: AI defaults to optimism and hallucinates modern solutions for legacy code; grounding AI in evidence and specific compiler output makes it far more useful; and preserving a stable, honest test baseline before making any changes is critical. The 'Tourist Prompt' anti-pattern is contrasted with an 'Archaeologist' mindset that prioritizes understanding over quick fixes.
