An airhacks.fm conversation with Burr Sutter (@burrsutter) about: discussion about integrating LLMs into enterprise Java applications, challenges with non-deterministic LLM outputs in deterministic code environments, limitations of chat interfaces for power users in enterprise settings, preference for form-based applications with prompts running behind the scenes, using LLMs to understand unstructured data while providing structured interfaces, maintaining existing CRUD systems while using LLMs for unstructured data like emails and support tickets, practical examples of using LLMs to generate code from business requirements, creating assistants with system messages and short user prompts, potential for embeddings to replace text prompts in the future, developer journey in learning LLM integration including prompts, tools, RAG, and agentic workflows, benefits of specialized agents over one general agent, using LLMs for code generation with limitations for complex use cases, hybrid approaches combining LLMs with human oversight, using LLMs for email routing and support case classification, potential for extracting knowledge from enterprise data sources like Confluence and SharePoint, quality assurance with LLM judges, discussion of small language models versus large ones, model distillation and fine-tuning for specific enterprise use cases, cost considerations for model training versus using off-the-shelf models with better tool invocation, prediction that models will become more efficient and run on commodity hardware in the future, focus on post-training inference and reliable results

Burr Sutter on twitter: @burrsutter

An airhacks.fm conversation with Volker Simonis (@volker_simonis) about: early computing experiences with Schneider CPC (Amstrad in UK) with Z80 CPU, CP/M operating system as an add-on that provided a real file system, programming in Basic and Turbo Pascal on early computers, discussion about gaming versus programming interests, using a 9-pin needle printer for school work, programming on pocket computers with BASIC in school, memories of Digital Research's CP/M and DR-DOS competing with MS-DOS, HiMEM memory management in early operating systems, programming in Logo language with turtle graphics and fractals, fascination with Lindenmayer systems (L-systems) for simulating biological growth patterns, interest in biology and carnivorous plants, transition to PCs with floppy disk drives, using SGI Iris workstations at university with IRIX operating system, early experiences with Linux installed from floppy disks, challenges of configuring X Window System, programming graphics on interlaced monitors, early work with HP using Tickle/Tk and python around 1993, first experiences with Java around version 0.8/0.9, attraction to Java's platform-independent networking and graphics capabilities, using Blackdown Java for Linux created by Johan Vos, freelance work creating Java applets for accessing databases of technical standards, PhD work creating software for analyzing parallel text corpora in multiple languages, developing internationalization and XML capabilities in Java Swing applications, career at Sun Microsystems porting MaxDB to Solaris, transition to SAP to work on JVM development, Adabas and MaxDB, reflections on ABAP programming language at SAP and its database-centric nature

Volker Simonis on twitter: @volker_simonis

An airhacks.fm conversation with Alvaro Hernandez (@ahachete) about: discussion about stackgres as a complete database cloud solution for PostgreSQL, kubernetes as an abstraction layer over infrastructure providing a programmable API, Stackgres offering high availability with primary and replica nodes using patroni, integrated connection pooling with PgBouncer, kubernetes operators and Custom Resource Definitions (CRDs) as a powerful way to extend Kubernetes, day two operations automated through CRDs including benchmarks and version upgrades, Stackgres supporting sharding with Citus for horizontal scaling similar to DynamoDB, Change Data Capture capabilities using embedded debezium, failover mechanisms taking typically 30 seconds with DNS updates, synchronous vs asynchronous replication options affecting data loss during failover, Stackgres being implemented in Java using quarkus, ContainerD as a programmable container runtime that can be used without Kubernetes, Stackgres offering multiple interfaces including CRDs, REST API, and a web console, considerations for running databases on Kubernetes vs cloud-managed services, the advantages of containerization for infrastructure, the challenges of multi-leader setups in PostgreSQL requiring conflict resolution, the value of Kubernetes for on-premises deployments vs cloud environments

Alvaro Hernandez on twitter: @ahachete

An airhacks.fm conversation with Kevlin Henney (@KevlinHenney) about: first computer was a Sinclair ZX81 with 1K of memory, programming in Basic and later Z80 assembly language, creating simulations like volcano explosions and n-body problems as a teenager, transitioning to the ZX Spectrum and other early home computers, studying physics at university but becoming more interested in programming, learning fortran, getting his first programming job at a small software house in Bristol where his boss had a wall of books on programming languages and paradigms, becoming self-taught through reading these books, developing an interest in AI and philosophy of mind which led to pursuing a master's degree in parallel computer systems, creating a virtual machine for a Lisp-based actor model, learning about the occam programming language based on Communicating sequential processes (CSP) for transputers, discovering Object-oriented programming and being fascinated by modularity and encapsulation, encountering Java in the mid-90s as a free downloadable language with platform independence, appreciating Java's familiar C-like syntax while offering object orientation without low-level concerns, using Java primarily for training and consultancy work rather than application development, discussing the evolution of Java features like inner classes (Java 1.1) and interfaces, explaining his unique perspective on interfaces coming from distributed systems experience with IDLs, reflecting on his work with various distributed Java technologies like RMI, Jini and JavaSpaces, continuing his career as an independent consultant, trainer and speaker with strong involvement in the patterns community

Kevlin Henney on twitter: @KevlinHenney

An airhacks.fm conversation with Francesco Nigro (@forked_franz) about: Netty committer and performance engineer at Red Hat, discussion of Netty's history, focus on low-level core components like buffers and allocators in Netty, relationship between Vert.x and Netty where Vert.x provides a more opinionated and user-friendly abstraction over Netty, explanation of reactive back pressure implementation in Vert.x, performance advantages of Vert.x over Netty due to batching and reactive design, detailed explanation of IO_uring as a Linux-specific asynchronous I/O mechanism, comparison between event loop architecture and Project Loom for scalability, limitations of Loom when working with IO_uring due to design incompatibilities, discovery of a major Java type system scalability issue related to instance-of checks against interfaces, explanation of how this issue affected Hibernate performance, deep investigation using assembly-level analysis to identify the root cause, collaboration with Andrew Haley to fix the 20-year-old JDK issue, performance improvements of 2-3x after fixing the issue, discussion of CPU cache coherency problems in NUMA architectures, explanation of how container environments like kubernetes can worsen performance issues due to CPU scheduling, insights into how modern CPUs handle branch prediction and speculation, impact of branch misprediction on performance especially with memory access patterns, discussion of memory bandwidth limitations in AI/ML workloads, advantages of unified memory architectures like Apple M-series chips for AI inference

Francesco Nigro on twitter: @forked_franz

An airhacks.fm conversation with Burr Sutter (@burrsutter) about: first computer: IBM PS/2 386SX funded by grandparents' Kona coffee sales, early passion for programming and problem-solving, self-taught C programming, database engine development as a student, transition from theater aspirations to computer science, work with Progress 4GL and Silverstream, shift to .net development, joining JBoss and Red Hat through acquisition, Mark Fleury's impactful "free don't suck" presentation, evolution of Java application servers and middleware technologies, enterprise service bus and SOA, impact of docker and kubernetes on the industry, Red Hat's adaptation to cloud-native technologies, development of quarkus, current interest in language models and GenAI, Java's longevity and adaptability, Quarkus' fast startup time and compatibility with legacy Java EE applications, work on Kubernetes and Quarkus, the importance of Java's "write once, run anywhere" principle, Java's performance compared to other languages

Burr Sutter on twitter: @burrsutter

An airhacks.fm conversation with Christian Humer (@grashalm_) about: early programming experiences with DOS text Adventures and Captain Comic, transition from graphics design to computer science, work on Java Server Pages (JSPs) and point-of-sale systems, development of Swing GUI for touchscreens, introduction to GraalVM and Truffle framework, ActionScript, Adobe Flash and Adobe Flex, explanation of Futamura projections and partial evaluation in Truffle, discussion on the challenges of implementing dynamic language runtimes, de-optimization in JIT compilers, Nashorn JavaScript engine vs. GraalJS, language interoperability in GraalVM, reuse of libraries across different programming languages, embedding of JavaScript and React in Java applications, comparison with PyPy in the python ecosystem, current work on bytecode DSL for generating bytecode interpreters, the importance of math in computer science and its relation to programming concepts

Christian Humer on twitter: @grashalm_

An airhacks.fm conversation with Richard Bair (@RichardBair) about: early programming experiences with Basic and building computers, his first production app at 17 for his father's auto repair shop, starting computer science degree at 16 and completing it at 43, joining Sun Microsystems' Swing team, working on JavaFX from its inception as F3 through its evolution, becoming lead of JavaFX team and chief architect for client Java, moving to Oracle's IoT team, current role as VP of Engineering at Hedera blockchain company, explanation of Hedera's Hashgraph algorithm solving the Byzantine Generals Problem, implementation of Hedera's technology in Java, open-source nature of Hedera's codebase, resources for learning about and developing with Hedera including documentation, Hedera Improvement Proposals - hips, and developer hedera playground, discussion of blockchain technology and its potential impact on open protocols and decentralized networks, comparison of blockchain to distributed databases, explanation of consensus mechanisms in distributed ledgers, tokenization of real-world assets

Richard Bair on twitter: @RichardBair

An airhacks.fm conversation with Alvaro Hernandez (@ahachete) about: discussion on Postgres JDBC driver contributions, SCRAM authentication library implementation, importance of connection pooling for Postgres performance, tuning Postgres configuration, PgBouncer as a popular connection pooler, challenges with lambda and database connections, benefits of using connection poolers at multiple levels, the need for an HTTP-based protocol for PostgreSQL, PostgresSQL Configuration Tool by StackGres, distributed SQL databases like DSQL and their trade-offs, optimistic vs pessimistic locking in distributed databases, comparison of JPA optimistic locking to distributed database conflicts, the power of using SQL directly vs ORM frameworks, the evolution of Java and JDBC making direct database queries more convenient, the potential benefits of using stored procedures in databases, the importance of understanding database internals for optimal performance, the need for careful consideration when choosing between ORM and direct SQL queries, the complexities of distributed databases and their impact on application design

Alvaro Hernandez on twitter: @ahachete

An airhacks.fm conversation with Francesco Nigro (@forked_franz) about: starting with a used Commodore 64 without display, breakdancing as a hobby and its influence on his learning approach, studying computer science at university with a focus on AI and compilers, pursuing a PhD in reinforcement learning, transitioning to IoT and embedded system work, discovering high-performance computing and concurrency patterns like the Disruptor, contributing to open-source projects, persistence in joining Red Hat despite initial rejection, rewriting ActiveMQ Artemis journal, considering Hazelcast before ultimately choosing Red Hat, working on messaging and performance optimization at Red Hat, becoming the performance expert for quarkus, journey from assembly and C programming to Java performance optimization, the importance of understanding low-level details in high-level languages, the impact of container resources on Java JVM performance, the value of deep technical knowledge in the age of AI and LLMs, Francesco's current role at Red Hat focusing on Quarkus performance and scalability issues

Francesco Nigro on twitter: @forked_franz

An airhacks.fm conversation with Ladislav Thon (@ladicek) about: CDI history and evolution, transition from XML-based configuration to annotation-based dependency injection, introduction of CDI lite in version 4.0, differences between portable extensions and build-compatible extensions, Arc as Quarkus CDI implementation, challenges in implementing CDI at build time, new features in CDI 4.0 and 4.1 including lifecycle events and method invokers, comparison of CDI with other dependency injection frameworks, discussion on decorators, interceptors, and stereotypes in CDI, performance implications of CDI in Quarkus, Convention over Configuration in CDI, upcoming changes in CDI 5, removal of expression language dependency from CDI API, benefits of build-time oriented implementations like Quarkus, challenges in migrating portable extensions to build-compatible extensions, introduction of synthetic beans and observers, addition of priority support for stereotypes, improvements in invocation context API, ability to declare priority on producers in CDI 4.1, integration of CDI with application programming models, Convention over Configuration paired with dependency injection, performance considerations of CDI in Quarkus compared to manual dependency management

Ladislav Thon on twitter: @ladicek