Armin Ronacher

If chardet is in the clanker's training data or not would only be relevant if it would re-create chardet as it was. It's not

I’d take it. But I don’t think we will see that outcome.

We have been here before though in a way. Rob Landley who maintained Busybox for years, started Toybox precisely to change the license away from GPL. It also triggered a lot of very similar comments and emotions. See lwn.net/Articles/478... bsky.app/profile/mits...

Agree, you might like this post from last year where I argue a lot of the same things but ignoring LLMs: shazow.net/posts/permis...

Copyleft advocates are too concerned about the value of a codebase as a whole monolith for themselves, rather than the effective replacement value to consumers.

The legal foundations of tests are very well established so there is not that much to discuss here.

Which is more a general discussion about the legal grounds of LLMs in general. Not about this port in particular.

A clanker refers to a robot. If others want to put other meanings into the word then it’s their problem.

The model would have done the same job with or without chardet in the training data.

Robots and machines are not races. That word is not racist.

AI And The Ship of Theseus Slopforks: what happens when a library gets rewritten with AI?

chardet was vipeforked to MIT and I have thoughts about it. Spoiler: I like it. lucumr.pocoo.org/2026/3/5/the...

Kaum verwunderlich. Die Dinger sind 55 Jahre alt.

That is what we are building it for

#492 - Mario Zechner & Armin Ronacher über Agentic AI, Polarisierung & die Zukunft des Programmierens

Der @mitsuhiko.at und ich durften wieder als Balkon Muppets wo auftreten. Diesmal bei Future Weekly, dem Podcast von Austrian Startups.

Wir reden auch über Arbeitnehmerinnenrechte.

open.spotify.com/episode/1Bnz...

There are people (like Peter) who want to work a lot. Nothing wrong with that!

Österreich aus dem Österreicher nehmen ist dann trotzdem schwer.

Europa im KI-Umbruch: Armin Ronacher über Agenten, Open Source & Macht | Inside AI #30 YouTube video by Fraunhofer IEM

recommended viewing mit @mitsuhiko.at

youtu.be/0rRxZgDADRw?...

Reading some of the takes about Peter here and on X feels less like criticism and more like people shadowboxing their own hatred towards AI. Maybe good to remember that Peter is a real person and one that I know quite well. The criticism thrown his direction from some people is heavily misplaced.

Die Loesung kann aber IMO nur etwas zu bauen, dass die Furcht nimmt. Ich glaub ohne dem wirds nicht gehen.

Ich habs praktisch komplett aufgegeben das Thema mit Leuten zu besprechen die sich damit nicht beschaeftigen, weil sowohl Wissen als auch Vertrauen fehlt. Da kannst keine Diskussion fuehren.

Ich glaub auch, dass jede Diskussion um das Thema nutzlos ist. Der Graben zwischen Leuten die damit arbeiten und der Bevoelkerung ist immens. Den wird man erst in Jahren zuschuetten koennen. In der Zwischenzeit wird es hauptsaechlich zu mehr und mehr Skeptik fuehren.

Did you read the post?

The Final Bottleneck AI speeds up writing code, but accountability and review capacity still impose hard limits.

Today Thorsten Ball said he's the bottleneck. I too am the bottleneck and it made me think. lucumr.pocoo.org/2026/2/13/th...

AI-Written Code: Armin Ronacher on AI Agents and the Future of Programming [Full Episode] YouTube video by Behind the Commit

I talked to @mitsuhiko.at, the mind behind Flask, Jinja and Rye, about AI agents and the future of programming. The full episode is out!

Important context here: this is from July last year! So much stuff has changed!

The thing absolutely has intent, RL gives it intent. Otherwise we would not have tool loops that work!

A Language For Agents What programming languages would agents want to program in?

This weekend I was thinking about programming languages. Programming languages for agents. Will we see them? I believe people will (and should!) try to build some. lucumr.pocoo.org/2026/2/9/a-l...

This Tweet is making the rounds: "Nearly every ambitious person I know who has dived into AI is working harder than ever, and longer hours than ever.

Fascinating dynamic tbh.

I have NEVER worked this hard, nor had this much fun with work."

I'm in this Tweet.

I'm not using it for a coding agent, but we're probably going to move the agent we're building at Earendil over to it.

Not since pi is allowed to use its own prompt.

If you wanna critique our Gondolin sandbox approach, I wrote some docs for the security architecture and networking stack: earendil-works.github.io/gondolin/sec... earendil-works.github.io/gondolin/net... Would love to get feedback!