Is There Really a 'Vienna School of Agentic Coding'?

Austria. That country that brought us Schnitell, Schwarzenegger, and Strudel, has now brought us Peter, Armen, and Mario, the minds behind the Vienna School of Agentic Coding. Yeah, some schmuck on the internet came up with the term Vienn Vienn School of Agentic Coding, and I have still no idea what it means. In early 2026, Peter Steinberger broke the internet when he released Open Claw. It's an open source AI agent that has taken over the tech world in a matter of days, exploding in popularity, reaching over 180,000 stars on GitHub. Over the last few weeks, everybody's been going crazy over it. It's racked up over 65,000 GitHub stars in record time. Mario, Armen, and Peter have known each other for years. In 2025, the stars aligned, and these three Austrian engineers all had some free time. Peter made me re-evaluate Claude code in particular and and through that that the agents are actually much further than than I was giving it credit at the time. They started vibe coding in a Vienna flat resulting in a little open source project called VIP tunnel which is an entirely vibe slop and you should never use it. It's garbage. Mario's first tinkering with claude code made him realize he wanted to build his own agent. The reason I built Pi is that I like being in control of my tools. Armen was enthusiastic. So I fully switched to Pi around Christmas 25. Peter was also enthusiastic. He quickly adopted Pi as the underlying agent of OpenClaw. The hype around OpenClaw became hype around Pi. And Armen, my partner in crime, thought it was a great idea to write a blog post saying, "Oh, OpenClaw is actually based on pi. You should use that." That's basically the downfall of the entire thing. And then I wrote a blog post in early January um sort of on the when when open claw was was really popular like hey just do away with the claw and just use the underlying pie instead. As the claw took over the world, people started to notice that the Austrians were thinking a little bit differently about how we should build software in the AI era. While Armen, Peter, and Mario seem reluctant to accept or define exactly what the Vienna School of Agentic means, there seems to be some core beliefs that unite them. The first thinking is essential. In industry, we have token use leaderboards in big corporations and anyone who doesn't use enough tokens is on a cutting block. That's just really ridiculous. Everybody kind of knows that this is currently unsustainable. If I were to say like what's the Austrian way of of doing a chantic software engineering, it is a maybe a little bit more of a measured approach. Don't leave your brain at the door. Kai is very primitive and that forces you to kind of think a little bit more before you use it which hopefully leads to you understanding the technology which I think is paramount to become effective at using it. Second, minimalism over maximalism. I just can't bear the thought of having a million lines of code that do something that can be done in 5,000 lines of code because I think complexity is the mind killer. We basically just don't have the training data for the models to become actual software engineers instead of coders. you need design, you need uh system architecture and all of that stuff. And they are really bad at that still because like if you look at what Peter is doing now, um he's is he's is a a trailblazer in in in trying to see like how far can you push it and and I think me and Mario are more like this is really cool, but also we don't want to maintain this. But I think like even if you sort of even if you spend enough time with Peter, Peter actually despite the fact that he goes all in on this is actually pretty critical himself on on some of the consequence of the technology. Third software should be malleable. If agents can code, why can't it modify itself? If you see what kind of software um a secretary has to work with in a doctor's practice, typically not great software. All this agency that the office worker has on the desk with physical paper to create their own workflow is actually really good and you lose out a lot of that when you adopt a software product that tells you how you should be doing that. You as a as a user of the software should not feel like you have to go to a software engineer to make it behave like you want. Fourth, tools should be open and neutral. It should be possible to build a commercial company and also still value this idea of like open protocols and and and open source software and and finding ways to reclaim a little bit of this space that was otherwise held by uh large corporations where like the main interest is just to grow one closed ecosystem. Ultimately, what we want is intelligence for everyone um maybe even locally. uh with neutral platform neutral tools like pi or open code allowing you to access that intelligence and make it useful a long enough horizon everything will turn into open source so what do the guys think about the Vienna school of agentic coding I personally it's charming somebody came up with that phrase for me personally it means nothing while they don't vibe with the term Vienna school of agentic coding they certainly have a vision for the future there's a there's a bad future and there's a good future. And I think the bad future is where our industry continues doing the same crap that it has been doing for the last 20 years, which is creating creations that are having a net negative impact on society. One of the reason we spend so much time on a genetic coding tools is because they're addictive. If it just becomes performative art, what you do with them, and you actually don't really develop anything of value, then that's a bad future. And I don't want that future to happen. I I want the future to happen where people that are really good at that job and have enjoyment in it find the job more enjoyable because the boring part about it can be in parts be automated with AI. Somewhere in there is hopefully a really nice future where we actually get some value out of this that helps people