Author
117 articles
Quarkdown, a documentation system using "markdown with superpowers", has been released. Changes include quite a few useful documentation elements (key bindings, path specifications, other useful presentation elements) but the core focus is on the project and its scope.
Dragonfly is a multithreaded ground-up reimplementation of the Redis wire protocol, described as being designed to scale vertically where Redis can't. The interesting thing is that Dragonfly is selling itself as "in-memory infrastructure for caching, feature stores, and job queues," which is the IMDG shape from the cache side of the river and a quiet admission that the cache framing was always too narrow.
JetBrains just published a survey on why AI adoption lags in CI/CD. Their read is that teams are being appropriately cautious; the harder read is that pipelines have no feedback loop for evaluating AI and no outer validator for catching it when it errs. The survey's own numbers quietly confirm this: among teams using AI in CI/CD at all, most aren't delegating. The real split isn't AI versus no-AI. It's passive versus active.
Leaning Technologies released CheerpJ 4.3, a WebAssembly-packaged JVM that runs unmodified Java applications — including applets, JNLP applications, and Oracle Forms — in a browser without a plugin. This release adds ergonomic features like in-window file upload and download, long-press-as-hover for Swing, and stability gains, and sets up a major startup-time improvement for 5.0.
Timefold Solver, OptaPlanner's successor, hits 2.0 with list variables promoted, a new Neighborhoods API, and a quietly consequential license shift: the explainability APIs you built on in 1.x now require a commercial edition. The solving core stays Apache 2.0, but the answer to "why did the solver pick this?" is now a paid feature.
TypeScript 7.0 Beta represents a fundamental architectural shift for the language, with the entire compiler rewritten in Go rather than bootstrapped TypeScript. The rewrite delivers approximately 10x performance improvements over TypeScript 6.0 through native code execution and shared memory parallelism, while maintaining strict semantic compatibility with the previous version.
The "10x engineer" may be a real phenomenon, in that some people really do vastly outperform their peers along certain metrics, but "10x output" is the wrong thing to optimize for: you want to find the people who create exponential output for the whole team. AI makes this kind of optimization urgent, because these people not only help their team members, but curtail the worst aspects of AI aid, too.
Chicory Redline delivers native-speed WebAssembly on the JVM by compiling the Cranelift compiler to Wasm and running it through Chicory itself - native code, bundled in a JAR, with no JNI required. Already shipping in JRuby with a 10x parsing improvement. Early but credible, and the cleanest Panama FFM use case to appear so far.
Zig 0.16 is a significant release that breaks a lot of existing code on purpose. Major I/O abstractions have been deleted and redesigned, the type system has been overhauled for better ergonomics, and the toolchain continues its march toward self-hosting. Worth a look even if you're not a Zig developer, because this is what a language looks like when it's still willing to be wrong on the way to being right.
JetBrains' Koog agent framework now integrates directly with Spring AI, layering agent orchestration - multi-step workflows, fault-tolerant checkpointing, history compression - on top of your existing Spring AI setup without replacing it.
Pilot 0.1.0 is a Maven plugin that provides a text console for the Java build tool, providing features like update checking, conflict detection, CVE auditing, dependence tree display, and more in a text console application.
MethodHandle has been in the JDK since Java 7, and most Java developers have a vague sense that it exists somewhere near invokedynamic and lambda internals. David Lloyd's ongoing series covers the mechanics thoroughly. What it doesn't cover is when you'd actually reach for it - and why.
Terminal Trove mentioned Matt Hartley's `netwatch` tool as its "tool of the week:" netwatch is a network diagnostic tool for the console, and it's pretty comprehensive, showing a ton of information about network traffic in a quite reasonable, useful interface.
Most teams don't trust their code - they trust the people who wrote it, or the streak of days without incident. That's not the same thing, and the difference is becoming harder to ignore. Trust in a system isn't social capital or accumulated momentum: it's demonstrated, repeatedly and verifiably, by the system itself. If you have to read the code to know it works, you don't know it works.
pg_6502 is a fully functional MOS 6502 emulator running in pure PostgreSQL - registers, memory, and all 151 opcodes implemented as stored procedures. It passes functional test suites. Could you load a Commodore 64 ROM into it? Probably. Should you? That's not our call. But someone's going to port DOOM into PostgreSQL eventually, and this is how it starts. Has anyone seen SkyNet around?
John Loeber wants idiomatic design back, and he's right - but the loss of idiom wasn't laziness, it was physics. IBM's Common User Access standard worked because the OS enforced it and the hardware was fixed; move to the browser, then mobile, then AR, and the contract doesn't bend, it voids. We're not in a period of design failure; we're in the same chaotic pre-standardization phase that preceded CUA in the first place. The question isn't how to restore what we had - it's how to recognize the improvised patterns that are already winning, and hold the line when we find them.
Scott Chacon and Markus Feilner have introduced GitButler, a tool that changes how developers work with branches in Git: instead of working on a branch, possibly introducing changes that belong in other issues, developers can work on a cohesive codebase and split changes into lanes that make more sense in context, which can make working with git much cleaner.
Optimization isn't a single measurement and a fix. It's whack-a-mole: find the loudest problem, fix it, then look at what the problem was hiding. Jonathan Vogel's second installment in his Java performance series shows this process live, with JFR recordings and flame graphs - including a contention bug that was completely invisible until he pushed the concurrency high enough for it to matter.
We've decided, collectively, that the things we love should be free and paid for by someone else. Thunderbird just put a number on how that's working out: 3% of their users actually invest in the project. We want good software, honest journalism, and open source tools that last, but we just don't want to be the ones paying for them. That's a problem.
Three projects had releases that generate scaffolding for database access: spring-crud-generator uses an entity specification as the director for generation, scythe uses SQL queries as story drivers for data definition and API endpoints, and fluent-repo-4j provides a more direct database access model for Spring, not as reliant on domain-driven design for access patterns as the formal Spring Data modules.