How BNY Uses Moderne to Make AI Agents Cheaper and More Accurate
summary
BNY runs Moderne and OpenRewrite to extract deterministic metadata before any LLM call, using regex-based recipes to output CSV + Markdown that agents consume, which cuts token spend and guesswork. They migrate JUnit4→JUnit5 and Spring Boot at scale, find Kafka usage across ~50 repos (99% discoverable by pom.xml), and tap a community catalog of 3,000+ recipes. The pipeline pre-filters code, verifies coverage against dependencies, re-runs tests, and refines recipes in a continuous feedback loop.
AI-written summary · about
chapters (11)
- 0:00Introduction and motivation
- 3:20Best practices for Kafka
- 6:51Problem space and initial findings
- 12:13Creating custom recipes
- 18:28Generating metadata for AI
- 23:21Codifying best practices
- 26:38Refining and verifying results
- 28:26AI integration with Moderne
- 31:41Leveraging MD files and agent tools
- 35:06Key takeaways and future steps
- 39:21Q&A and closing remarks
related talks
45mTrusting AI to Modernize Software at Scale
27mThe State of Large-Scale Code Modernization
54mHybrid Modernization: Combining OpenRewrite's Precision with LLM Intelligence for Spring
37mDeterministic modernization in an age of AI
59mTools & Tips to Modernize Your Java code with AI
27mNavigating Application Modernisation - Leveraging Gen-AI
46mMagic of Automation and Everyday Chores
16mAI-Powered Mainframe Modernization