Ran into some friends on the way to #rubykaigi!

Also, never be at the edge of a wide angle photo! 🫃

The artwork for #RubyConfAT is amazing! I'm SO excited to visit Vienna for the first time!

bag labeled "w" as in wege (veg)

bag labeled "M" as in meat

Lunch bags at #wroclove.rb day 2... `w` as wege (veg) vs M as in meat. 😂

I loves me a good hybrid JVM application!

Twenty years working on JRuby and delivering hundreds of talks and here's the top five questions I still get:

What is JRuby?
Why should I use it?
Do I need to know Java?
Does it run Rails?

And after my presentation:

Why isn't everyone using this?!

Excellent, densely-packed talk by @twnsnd.com at #wroclove.rb, starting with a short history of Rails and leading to modern best practices for tight, fast, nearly JS-free (no React!) web frontends. Makes me want to build something with a spicy-hot JRuby backend!

SHOCKER! For a simple piece of code like "Array.new(10_000_000) {Object.new}", JRuby 10.1 with JDK 26 and compact object headers uses LESS REAL MEMORY than CRuby 4.0. That's just crazy. Real parallel scaling AND lower memory use?!

Lettering "Sponsors" and eyes emoji on a branded background.

This year we’re going big — two conferences instead of one.
That also means we need more support. Give us a hand: make a repost, tell your community, pitch to employer — anything helps, and we’ll make it worth it.
Proposal: https://balticruby.org/
#ruby #conference #sponsor

Well, it's closer at least.

absurdly incorrect flight map generated by chat GPT showing AMS in New York and Japanese airports in South Asia

Note to self: don't ask AI (ChatGPT) to generate realistic flight maps. How many errors can you spot? I'll get you started: MSP is not in Georgia.

Feels good to be on the road again! For the first time since 2007 I'm traveling all the way around the planet! Currently at Schiphol en route to Wroclaw, Poland for wroclove.rb. Saturday I fly to Japan for RubyKaigi. Then home... a full orbit! I'll be two days ahead of you! 😵‍💫

Big announcements coming from JRuby next week... we are ready to unleash JRuby 10.1 with a ton of improvements and some really exciting new technology. Stay tuned.

Well I mean it looks like a pretty simple mathematical surface that could be generated from a formula.

At what point do I just start ignoring dependabot, and then I'm right back where I started?

Looks decent! I'm guessing you could probably procedurally generate that surface as well.

Frequent conversation:

> Why don't you use AI for this task?

I tried. It failed to do it better than me..

> You must be doing it wrong.

Great, here's the task, try it yourself!

> I see the problem. These models don't do that task well.

Isn't that what I just said? 🙄

My son has a low-cost ticket for #RubyKaigi available right now! Help him out and buy this ticket off of him!

x.com/i/status/204...

Feels that way sometimes! To be fair, I paid nothing for the printer because it's a 6-year-old Creality Ender-3 and printing parts to upgrade it is only costing me pennies.

Owning a 3D printer is saving me so much money printing parts for my 3D printer. Wait a second... 🤔

makeshift raspberry pi camera mount using a wine glass and a binder clip

This is why AI will never defeat me.

JRuby 10.0.5.0 is released! This round we fixed some memory leaks when embedding multiple JRuby instances, a few pesky Windows bugs, and several Ruby compatibility blockers. Enjoy!
www.jruby.org/2026/04/06/j...

We have been saved from the evil and cruelty of this administration only by their rank incompetence.

That's a great way to put it! I've been struggling to find the right words. My experience with these tools has been that they mix simple and correct answers with dangerous, confidently incorrect mistakes. The users depending on them are either too inexperienced or too busy to audit everything.

Tonight, our nation can celebrate the removal of this horrible, shallow, incompetent person from the top of the US Department of Justice. The Trump administration dumpster fire continues, but at least this idiot is gone.

This one started to stray into really insulting language, making me doubt my own analysis and understanding of the system.
It told the user that my example proving it wrong "hid" the actual issue, followed by more completely broken analysis. It's dangerous... and starting to feel just plain evil.

I just spent two days in mortal combat with an AI agent that convinced a user my analysis and fix for a bug was incorrect. It provided convincing descriptions of the problem and declared its fix "obvious".

It was all nonsense.

Your AI coding agent is LYING TO YOU. Beware.

Things are changing fast, but I would say no. My experience so far with AI is that the code it generates and iterates on eventually becomes almost unreadable by humans, and at some point it's too tangled of a mess even for AI improve. Sometimes write-only code is okay, but not often.

Then you are not one of those people I referred to.

It is a bit ironic that many folks who demanded strict static typing and who refused to even consider using dynamically-typed languages are now advocating for verifying the results of agentic coding using little more than unit tests and specifications. So the result *is* actually all you care about?

I loathe April Fool's Day. The only news I want to be an April Fool's prank is the winter storm warning for Minnesota this evening.