Today, Europe is Hungarian.
We are stronger. More united.
To the people of Hungary - you've done it again.
You have spoken.
You have chosen Europe.
It's a victory for fundamental freedoms.
Main reasons I would try it now:
* back to basics and simplicity
* a European coding agent backed by a European company
* 30k+ stars and technical credibility from the author
* (too much) fun
A while ago, I saw a post on lobste.rs about someone building his own coding agent because he thought he could do better than existing tools, but I didn’t look into it.
What is changing my mind about Pi.dev, mainly:
* this talk: (youtu.be/Dli5slNaJu0
* this post: : mariozechner.at/posts/2026-0...
I have too many unfinished projects lying around already, so trying another coding agent is a test of my self-control.
But Pi.dev is starting to look hard to ignore.
Welcome!
I’m also using OpenAI models (and Opencode Go).
Next stop: Pi :)
It is is, doesn’t always stop with skill instructions either. I should probably look into (Opencode) plugins but where does it all stop right?
Good post on Scala adoption but also language adoption in general.
Being an excellent language isn’t enough. (I still like the Scala a lot but rarely use it these days.)
medium.com/@w.pitula/sc...
Harness engineering involves dealing with … John McLane? Keeps going
Had to add a simple instruction as well: ‘stop and ask the user’ when it detects 2 permission denieds in a row…
Avoid bouncing off the guard rails … and burning a lot more tokens. Harness are not free not easy!
Experienced the same when spending (way too much) time with Safehouse to box in my coding agents: on every ‘permission denied’ it tries to work around it and potentially goes on a tangent.
I've given up on writing conventions in markdown files and watching AI ignore them.
It's much more fun to do crazy stuff with Oxlint.
Here's what I've been upto: nick-tune.me/blog/2026-04...
Happy for any feedback or alternative approaches.
Coding is dead! Well maybe... the job of a software developer is certainly changing really fast at the moment due to AI tools. This is all very exciting and worrying at the same time. I just wrote this article about what TDD looks like in this new age: coding-is-like-cooking.info/2026/03/test...
Organisational Dysfunction of the Day
Stand-ups
Context: Your team is using Scrum and has been taught how important stand-ups are (daily coordination meeting), but it feels like a drag and a complete waste of time.
Warning to open source maintainers: the Axios supply chain attack started with some
very sophisticated social engineering targeted at one of their developers simonwillison.net/2026/Apr/3/s...
Proton Meet, secure video conferencing next to an image showing Proton Meet in action, a group of people on a video call.
Today we're launching Proton Meet: end-to-end encrypted video calls where no one can listen in, not even Proton. Built on the same zero-access architecture as Proton Mail. Read more: https://proton.me/business/blog/introducing-proton-meet
TIL: The Danish scale for grading (12, 10, 7, 4, 02, 00, −3) was deliberately chosen so the written numerals are hard to change into more favourable grades making grade-tampering difficult.
Plan Mode ...
No harm done but this is why I started running in Safehouse
My other condensed takeaways:
1. Modularity should make future change easier.
2. Coupling should be chosen deliberately.
3. Boundaries should reflect how teams actually work.
4. Modeling should optimize for usefulness, not completeness.
5. Architecture should record its reasoning and assumptions.
Worth a listen "Dear Architects" episode: "Why AI is the Third Coming of Domain-Driven Design".
My understanding is that, because AI changes the medium of communication to natural language, precise ubiquitous language will matter even more.
pod.link/1877884226/e...
If you use GitHub (especially if you pay for it!!) consider doing this *immediately*
Settings -> Privacy -> Disallow GitHub to train their models on your code.
GitHub opted *everyone* into training. No matter if you pay for the service (like I do). WTH
github.com/settings/cop...
Rename your team to its North Star
“They genuinely believe the spec document is the safety net. It isn’t. The spec document is the blueprint. The safety net is the test suite”
Julias Shaw via martinfowler.com/fragments/20...
Not sure about using Cucumber though. Experimenting with Testing DSLs as executable specifications myself.
If you installed or upgrade LiteLMM on March 24th all your secrets may have been stolen.
(Is it me or should their TLDR at the top be more explicit about that?)
docs.litellm.ai/blog/securit...
This is a screenshot of a GitHub dashboard page. At the top, there is a GitHub logo next to the word "Dashboard." Below, a blue information banner states that starting April 24, GitHub Copilot interaction data will be used for AI model training unless opted out, with links to "Review this update" and "GitHub account settings." The user's profile picture and username "chrismessina" appear at the bottom, with a label indicating it is their personal account. The overall design features a clean, minimal layout with a light background and icons for navigation and profile management.
GitHub is really going for it:
"On April 24 we'll start using GitHub Copilot interaction data for AI model training unless you opt out. Review this update and manage your preferences in your GitHub account settings."
I’ve been using my own skill to manage ADRs.
The main win is that this leads to implementing features better because of more relevant context.
When integrated well, agents handle iterating by amending, superseding, or creating new ADRs very well too.
github.com/lhohan/agent... 2/2
A good read on maintaining context across agent sessions:
martinfowler.com/articles/red...
Notably, it highlights documenting the why behind decisions using ADRs. 1/2
Excited to be attending Rust Week.
It’s been almost more than a decade since I attended a language-oriented conference (Scala Days).
2026.rustweek.org
While setting up sandboxes for my agents and to verify their working I ran some tests which confirmed: do not rely on agent permission prompts alone.
Did a short write up of the tests and the results.
hanlho.com/p/reminder-d...
Screenshot from fightchatcontrol with a template to write to MPs
🚨No Joke: Conservatives in the EU Parliament (EVP) want the vote on #ChatControl 1.0 to be repeated this Thursday - even though the Parliament already voted NO! 😡
Make sure your MP stays strong. Contact them now!
👉 fightchatcontrol.eu#contact-tool
Asked my coding agent to write after a file to ~/dev
Instead, it created ~/dev/1774107553.txt in my working directory.
I think I am not going to ask to clean that up ...
