Livecoding: Creating Beautiful Java Code With LLM and Agents
summary
LLMs produce idiomatic Java using Quarkus/MicroProfile templates and a Boundary‑Control‑Entity structure, generating Java 25 code with records, JAX‑RS endpoints, in‑memory stores and end‑to‑end system tests that compile and run on stage. Tools shown include Claude/Opus/Mistral/Copilot agents and Kiro‑style spec agents, and the demo builds a “Cinedome” movies+tickets service that works immediately. Claims: LLM Java output hits 87% success rate vs ~60% for JS/Python and can cut inference cost 30–80%.
AI-written summary · about
chapters (8)
- 0:00Introduction and context
- 2:34Java and LLM performance
- 3:56Boundary-control-entity pattern
- 7:24Leveraging LLMs for development
- 14:00Standardization and portability
- 20:00Live coding: Cinedome project
- 27:10Refining skills and architecture
- 36:01Advanced agentic development
related talks
53mHigh Speed Java Development with LLMs and Agents Fast, Clean, Maintainable
57mPure Java Enterprise AI/LLM Integration (EAI 2.0)
39mLLM Agents Gone Wild (And How to Tame Them with Quarkus)
01h 02mJava + LLMs (Build LLM Apps in Java with JakartaEE)
49mCoding Tomorrow
48mAgentic AI with Quarkus, LangChain4j and vLLM
45mComplementing LLM with agentic AI: a Quarkus story on AI
57mReal World Lean Java Practices, Patterns, Hacks and Workarounds