GitHub: https://github.com/Johnny70/ladanaOS
I've been working on Ladana OS for a little over two months.
It started as a summer experiment: could I build a Linux desktop where AI is treated more like a system layer than an application?
Ladana is very much a proof of concept. It isn't finished or production-ready, and I'm not trying to compete with GNOME or KDE.
The interesting part for me is the architecture.
The current PoC has:
A small custom X11 window manager and desktop shell
GTK3-based desktop components
An AI assistant that can work with desktop context
A master AI that can orchestrate sequential tasks across applications
Deterministic automation combined with LLM-based reasoning
The beginnings of a built-in Power Automate-style workflow system
One reason I wrote my own window manager was simply control. I could have built on top of an existing desktop environment, but then I'd also inherit a lot of assumptions about how the desktop should work.
For this experiment, I wanted to control that layer myself.
The UX is intentionally familiar. I've borrowed heavily from desktop conventions that already work well — especially Windows. I don't want people to learn a completely new way of using a desktop just because the underlying architecture is different.
The interesting change is supposed to happen underneath.
I'm also deliberately using mature technology for the PoC. X11 + GTK3 isn't particularly fashionable these days, but it gives me a predictable foundation while I experiment with the parts that are actually new.
There are still a lot of rough edges. Some parts are barely experiments, and I expect the architecture to change substantially as I learn more.
I've mostly been building this by myself, so I'm putting it out there because I'd really like some outside perspectives.
I'm especially interested in feedback on the concept rather than bug reports.
Does treating AI as part of the desktop make sense to you?
What would you want an AI-native desktop to actually do?
And what parts of this approach seem fundamentally wrong?
I'd genuinely like to know.
Top comments (0)