VibeOS: Computer engineering student produces OS using Claude Code

mylesillidge

Journalist
Joined
Jul 29, 2021
Messages
3,912
Reaction score
4,145
University student vibe-codes an entire operating system from scratch

A computer engineering student has produced an operating system using the latest version of Anthropic’s Claude large language model that boots on real ARM64-based hardware and an emulator.

“VibeOS runs in QEMU aarch64 and Raspberry Pi Zero 2W. It browses the web, runs Doom, and has a C compiler and Python,” explains Kaan Şenol, a student at the Polytechnic University of Turin in Italy.
 
Vibe coder: chuck in a Python and C compiler while at it, ok.

What nonsense.
 
Last edited:
What this probably means is Claude is well on it's way to being able to generate the next leap in AI by itself. These companies are pushing the coding side of things hard so that their models have the capacity to create better versions of themselves.
 
good grief, an OS slapped together by AI and that world famous Italian work ethic

Jesus, take the wheel!


having said that, been putting ChatGPT, Grok and Claude through their paces a bit with analytical tasks ... Claude is ahead in that game
 
  • Haha
Reactions: Jan
good grief, an OS slapped together by AI and that world famous Italian work ethic

Jesus, take the wheel!
Not to undermine a well-crafted joke, but ackchyually he's Turkish. He's just attending an Italian university :P


having said that, been putting ChatGPT, Grok and Claude through their paces a bit with analytical tasks ... Claude is ahead in that game
Iiiiiinteresting. People were raving about Gemini at the end of last year. You should definitely add that to your comparison if you haven't already.
 
Vibe coding is perfect when you are learnng a new language as a hobby. When you do not want to invest in it to care whether you fully understand it the so next time you can be on your own. No production code and when released (for your own personal use) you will cross the code maintenance bridge when you come to it (use AI again). The AI has access to the vast coding documentation and all the libraries, add-ons, extensions etc (all dynamic and not easy to keep up with) and when things do not work together it usually works out why. It is also a quick way into the Development Tools which can be quite complex and full of pitfalls.

I created a C# Windows App in 3 months that would have taken a year or I would have just given up
 
VibeOS is a OS prototype made AI-assistance. Even Linus is having a good at it,


Another silly guitar-pedal-related repo​

...

Also note that the python visualizer tool has been basically written by vibe-coding. I know more about analog filters -- and that's not saying much -- than I do about python. It started out as my typical "google and do the monkey-see-monkey-do" kind of programming, but then I cut out the middle-man -- me -- and just used Google Antigravity to do the audio sample visualizer.

Going to Kaan's git:


A hobby OS vibecoded completely from scratch with Claude Code. Not everything works, some stuff is not even tested, but most things do.

He documented the process, not in detail, but it is a exercise nonetheless:


I don't mind vibe-coding where there is transparency. Even I had a good session(s) exploring how cheat developers develop cheats, and the techniques were telling. I wouldn't have known without AI. I did it, to research cheating in videogames.

Currently I am using vibe coding in all my prototyping. I am a noob when it comes to coding. Learning new things everyday.
 

A good thing about vibe coding is that you can take observations, and explore that behaviourally in code. That helps to establish baselines, determine patterns, and understanding those parameters. I can do it, then companies like Valve and Riot and can do it too, sure they are investigating that which is degrading the user’s experience?

Also, it is remarkably easy to train VL learning on almost any game without needing current top-end hardware. Neither do you need to do it because there are ‘researchers’ that make readily trained models available.

I was inspired by Basically Homeless to do this discovery. It is always best to understand the problem in order to help solve the problem.

Sometimes I don't think that gaming companies are too pressed to provide remedy, and I think it has a lot to with the burden. Especially when those users who are cheating bought deep into your ecosystem. Going nowhere slowly isn't helping.
 
  • Like
Reactions: Jan
Top
Sign up to the MyBroadband newsletter
X