High Speed Java Development with LLMs and Agents Fast, Clean, Maintainable
summary
Claims that LLM agents generate production‑ready Java fast and clean are demonstrated concretely: Quarkus + MicroProfile starter projects are cloned in VS Code, agents (Copilot/Clo Code/Ko/devStral/opus45) create feature packages like boundary control entity components (Nordic gods, dwarfs), generate system tests, compile and iterate until tests pass, and follow a project-wide skills.md steering file to avoid hallucinations. Results show high-quality, maintainable code by grounding on stable Java/Jakarta specs and running tests to validate each iteration.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 3:13Java with LLMs and agents
- 8:37Java for AI-native dev
- 12:14Boundary Control Entity pattern
- 16:53Live demo: Nordic gods
- 23:23Dwarves component with Kudo
- 29:03Refining code with steering
- 34:29Java prevents hallucinations
- 38:02Code quality with LLMs
- 44:09MCP vs JAX
- 47:03Maintainable systems with context
- 50:15Skills.md and project rules
- 53:05Closing
related talks
52mLivecoding: Creating Beautiful Java Code With LLM and Agents
39mLLM Agents Gone Wild (And How to Tame Them with Quarkus)
57mReal World Lean Java Practices, Patterns, Hacks and Workarounds
57mPure Java Enterprise AI/LLM Integration (EAI 2.0)
49mCoding Tomorrow
48mAgentic AI with Quarkus, LangChain4j and vLLM
30mPicocli: Put Some Java in Your Terminal!
46mModern Java App Development in the Cloud