Vibe coding has a credibility problem. Futurism AI reports on a revealing experiment: an entirely AI-generated operating system called Vib-OS that turned out to be, in the words of the YouTuber who tested it, “a pretty featureless operating system” that’s “kind of shocking” in how little it actually works.
The project sounds ambitious on paper. Hosted on GitHub and runnable on both ARM64 and x86 machines, Vib-OS claims to feature a custom kernel, a macOS-inspired GUI, and a virtual file system. The documentation even promises you can play Doom on it.
You cannot play Doom on it.
What Actually Broke
YouTuber Tirimid spent several hours stress-testing Vib-OS and catalogued a long list of failures:
- The OS couldn’t connect to the internet
- File Manager buttons didn’t respond to clicks
- The Notepad app refused to save documents
- Snake, a game from the dumb-phone era, barely functioned
- An app labeled “Browser” turned out to be an image viewer
- Python scripting support was listed in the docs but didn’t exist
- The Doom launcher icon did nothing when clicked, suggesting the AI hallucinated the feature entirely
That last point is worth sitting with. The AI didn’t just write broken code; it apparently documented a feature that was never built. That’s not a bug. That’s confabulation baked into the product spec.
Why This Matters Now
Andrej Karpathy, OpenAI co-founder, coined the term “vibe coding” just last year to describe the practice of rapidly building software through natural language prompts to AI models. The approach exploded in popularity among hobbyists, startup founders, and even professional developers looking to prototype fast.
Vib-OS is the clearest stress test yet of where that approach breaks down. Building a CRUD app or a personal finance tracker is one thing. Building an operating system, with a kernel, memory management, hardware drivers, and a functional GUI, is an entirely different order of complexity. The gap between “the AI generated something that compiles” and “the AI generated something that works” gets wider as the system grows.
What’s significant is that even tech investors are starting to voice skepticism. Futurism AI has noted growing fatigue among VCs with AI startups built on thin technical foundations. Vib-OS is a hobbyist project, but it illustrates the same dynamic at a smaller scale: impressive-sounding output that doesn’t survive contact with reality.
The Honest Takeaway
For AI practitioners and builders, the lesson here isn’t to abandon AI-assisted development. It’s to be rigorous about what AI can and can’t do right now:
- Use AI for acceleration, not abdication. Code generation tools are powerful assistants, not autonomous engineers.
- Test everything. AI-generated documentation is not a substitute for verification. If it says Doom runs, boot up Doom.
- Complexity compounds errors. The more interdependent the components, the more hallucinations can cascade into system-wide failure.
Tirimid’s closing note to his viewers was gracious; he explicitly asked them not to harass the Vib-OS creators. He’s right. This project isn’t a failure of ambition. It’s a useful data point about where the technology actually stands.
Whether AI will one day reliably produce working, complex systems is an open question. But for now, the answer is clearly: not yet. You can find the full story at Futurism AI.