I find the hallucination metaphor in the context of LLMs very misleading (perhaps deliberately so). For humans, hallucination is an illusion in the absence of an external stimulus, i.e. malfunction. For LLMs, hallucination is all they can do. It just happens to be correct 🤷.

One of the most annoyingly unnecessary things I've seen too many times recently (including projects using http4k) is passing around base URL and credentials for each function that makes an HTTP call 🙄 In http4k world you can do `SetBaseUriFrom(. ).then(MyAuth()).then(client)`.

This is the recording of @duncanmcg and me doing the "47 refactorings in 45 minutes" talk at #KotlinConf Aka "47 refactorings in a week with PRs". (Note that it's missing slides, so the intro and outro might not make a lot of sense).

Assuming systems are doing their job perfectly if PRs are not about continuous integration (or code quality), they must have another function. What about PRs are about control (or rather an illusion of control)? And that could fundamentally be insecurity/anxiety. #noPRs.

How "clean" are two weeks of commits on a branch? 🤷‍.

As Duncan McGregor speculated, we might be using the same part of the brain to track whether our hands are dirty and to track the size of code changes. So people with different hygiene standards and definitions of "dirty" will tend to disagree?.

@duncanmcg Looks like the live stream video is not available anymore.

For anyone interested, I pushed the code from the "47 refactorings in 45 minutes" talk by @duncanmcg and myself at #KotlinConf to this branch (including references and mini-plugins). And here is the live stream video

"If you don't know what your organisation is optimising for, then it's probably control" TIL from #KotlinConf keynote by Diana Montalion.

If you happen to be at #KotlinConf today, come over to the Kotlin Foundation Booth at 13:45 where I will do a demo of LivePlugin.

I updated the blog post with the non-default IntelliJ settings I use for 2025.2 EAP. Please let me know if you disagree with them 😅.

I wish IntelliJ could put a breakpoint in bytecode and step through it. (The actual problem being several versions of a class without attached source code namespaced by classloader while figuring out who sets empty " on Kotlin embedded compiler).

I miss the 'Convert to lazy property' Kotlin intention after moving to K2. (It looks like it should be available in both K1 and K2 but I don't see it :( .

I wish the Kotlin 'Make private' intention was more like a refactoring and would only work if making var/function wouldn't break the code. Also, an inspection to make everything as private as possible would be great.

I love how the Shelve Silently action in the latest IntelliJ comes up with the change description by default (using shelveSilentlyTitleProvider via AI Assistant plugin). The next thing is to make it default for commits, so we’ll just be giving our approval to the generated text.

Random IntelliJ tip. You can compare any directories (including the content of files) by selecting two directories in the project view and invoking the 'Compare Directories' action (ctrl / cmd+D). You can then change the directories to any location outside the project 🧐.

This is a niche rant, but how come the presentation assistant in IntelliJ shows shortcuts from the keymap instead of the keys that I press on the keyboard 🤷 It had one job! I had to fix it by writing my own assistant using LivePlugin

Software that "just works" long enough without any intervention, naturally becomes hard to build/release/support. OTOH software that breaks often enough, creates a sociotechnical system to support its existence. Should it be a deliberate practice though? 🤔.

The meetup looks a bit busy, but if you're in London and would like to see @duncanmcg and me pre-practicing our KotlinConf talk by doing 31 refactoring in 30 minutes, then I'm sure there will be a seat and a free M&S snack/drink

If you happen to be in London on Thursday 17th April, this is the event to attend with two talks by the authors of @http4k.