Babylon and java.util.json

airhacks.fm podcast with adam bien

An airhacks.fm conversation with Paul Sandoz (@paulsandoz) about:

Devoxx conference experiences and Java's evolution over the past decade, energy efficiency studies comparing Java to C/Rust/Ada from 2017, Java performance improvements from Java 8 to Java 25, Code Reflection as manipulation of method bodies versus traditional reflection, tornadovm optimizations for GPU inference achieving 6-10x speedup over CPU, using pointers to keep data on GPUs avoiding transfer overhead, Metal support development for Apple Silicon, relationship between Project Babylon and TornadoVM, HAT project collaboration opportunities, Python's GPU performance through optimized NVIDIA libraries, enterprise challenges with Python in production versus Java's packaging simplicity, BLISS library for NumPy-like operations in Java, DJL.ai for tensor manipulation and Deep Learning, JTaccuino for Jupyter-style notebooks with JavaFX, MCP protocol implementation challenges with poor specification quality, minimal JSON API design philosophy for OpenJDK, cognitive overhead reduction in API design, pattern matching with JSON values, assertion-style API for fail-fast programming, JSON-P versus JSON-B trade-offs in enterprise applications, versioning challenges with data binding approaches, embedded HTTP server use cases for testing and development, JSON-java library as reference implementation, zero-dependency approach becoming more popular, Java 25 instance main methods with automatic java.base imports, zb zero-dependency builder project, marshalling and serialization rethinking in OpenJDK, trusted builds and dependency management in enterprise Java, comparison of Maven/Gradle complexity for simple projects, GPL licensing for OpenJDK code, the java.util.json experiment

Paul Sandoz on twitter: @paulsandoz