Author
117 articles
Kerry Ivan Kurian's 'Exponential Failure' says AI agents multiply parts faster than they improve them, so projects fail by arithmetic. He's right about the math and wrong about the fix: the answer isn't fewer parts, it's parts you can describe.
Paul Parks has released PUDL, the "Pleasantly Usable Design Language," a specification and toolkit for applying consistent and usable application design for applications.
Git 2.56 has been released, with some major improvements in rebasing support.
Midden is a single-binary heap dump analyzer for the JVM, written in Rust. Tested against two deliberate leaks and checked against MAT, it found both and got the sizes right, from a shell prompt rather than a workstation, and very quickly.
Helidon 27 is out, and it ties its version number to the JDK release cycle, including for LTS support. Helidon is Oracle's microservice library, released under the ASL 2.0 license; this version has some feature upgrades and separates the MicroProfile support apart from Helidon proper.
Sunil Pai's “senior engineer death spiral” describes what happens when engineers start managing perception instead of reporting reality. The deeper problem may not be individual pride or fear, but organizations that make uncomfortable information too expensive.
Kenta Cho built a Z80 REPL: type an instruction, execute it immediately, and watch the machine state change. It's eight years old, still delightful, and a neat example of what happens when good tooling makes experimentation cheap.
ByteCode.News has been quiet for a month. Part of that is ordinary life, but part is something more familiar: the point where everything happening in technology starts to look the same. Here’s what happened, and what BCN is changing to get moving again.
A release to Maven Central used to cost more than the code being released. Jason Lee's JReleaser walkthrough shows what replaced that process: one Maven profile, one command.
A recent discussion on how AI was used effectively made me wonder what "effective" or "better" meant, because there's no way to test the alternative any more to compare. I still don't know how to characterize my own results, but I think it makes me better because it gives me a chance to work out who I am and what I mean to say, but not really how to say it.
Andrew Oram asked why programming languages rise and fall, and quoted the answer without noticing. Programming is a vocation, an art, and a job. Make any of those harder than they need to be, and your language survives exactly as long as it can't be replaced.
Sean Goedecke argues that domain expertise is the real prompting skill. He's describing two quantities as one: a floor, available to anyone who can describe what acceptance looks like, and a gain that multiplies whatever you bring.
Xberg 1.0 has been released. Xberg is a document extraction engine, describing itself as a "document intelligence framework." It's written in Rust, with fifteen generated language bindings, and a very descriptive data model that provides a *lot* of flexibility.
There's no standard for marking AI contribution, and there can't be, because contribution isn't a scalar. What readers actually want is attestation, and that's what a byline always was. The question was never whether a machine touched the text. It's whether anyone was behind the controls.
Baldur Bjarnason argues that development tools win by making developers replaceable. The JVM ran that experiment three times, against Visual Basic's precedent, through EJB's committee, and into Spring's reversal, and the results say his theory is exactly half of a whole.
Imran Nazar recreated a JavaScript animation demo on 1982 hardware and wrote up every decision along the way: 6510 assembler, fixed-point math with no floating point available, a CPU bug turned into a cosine lookup. When he lands at 30fps, he can tell you exactly why without running it. It's very cool.
An essay from the creator of Primate argues that the JavaScript ecosystem traded cohesion for composition and has been paying for the seams ever since. The diagnosis reaches well past JavaScript: Java's own graveyard of JSF and GWT proves the seam problem is universal, and WebAssembly is about to make the question unavoidable for everyone.
Andi Roberts wrote up Alex Pentland's research on why some teams outperform others. The research is good; the practical advice mostly assumes a watercooler you may not have. Some additions from the trenches: what Brooks actually said, two cheap levers for building the idea flow, and when you want the opposite.
Reader Hideki Idoru argues that AI is a decent information distiller and a bad tool for nearly everything else in software, because no one can cheaply verify that generated code is correct. The deeper claim is that most programming was already trivial, unabstracted busywork, and AI has only torn the mask off. It's worth reading and thinking about.
Ray Myers wrote up the Bun rewrite and the fight around it. It's emotional in places, but there's real signal in it about how agentic work is sold versus how it's done, and one review technique worth stealing.