The Era of AAP: Ai Augmented Programming using only Java by Stephan Janssen

Discover how Java is becoming AI-ready with local LLM inference, integrated development tools, and frameworks that enable private, efficient AI-augmented programming.

Key takeaways
  • Java is becoming a first-class citizen in AI programming with new frameworks like JLama and Llama3.java enabling local LLM inference

  • AAP (AI Augmented Programming) tools can now run large language models locally with up to 131,000 token context windows, eliminating the need for cloud services

  • DevOps Genie plugin for IntelliJ provides open-source AI coding assistance while keeping code private and running models locally

  • GPU support through projects like Babylon and TornadoVM will enable faster matrix multiplications and LLM inference directly in Java

  • Model sharding allows distributing large models across multiple machines, making it feasible to run billion-parameter models on consumer hardware

  • Token caching and streaming responses (20 tokens/second) improve the user experience when working with LLMs

  • RAG (Retrieval Augmented Generation) capabilities are evolving from basic to advanced implementations using semantic search and graph databases

  • Panama Vector API improvements enable faster matrix multiplications natively in Java, especially on ARM processors

  • Local LLM deployment eliminates privacy concerns around sending proprietary code to cloud services

  • The combination of local models, Java frameworks, and IDE integration creates a complete AI-assisted development environment that can increase developer productivity