It's supposed to be an OS but I think it's mostly about UI.
The prototypes he built so far are very impressive and he’s been using them as part of his workflow.
1. "Atomic apps", which I understand to be apps that separate presentation from the data. 100x this. We used to have it, with files and programs that could open them, but with move to cloud apps, this feature of computing is disappearing. How do I make a directory that contains stuff from my google drive, onedrive, figma and youtube? Interestingly, there are now SaaSes that let you "index all your company resources", which try to tackle this problem.
2. "Graph OS". The author has really impressive Obsidian graph[1]. However, I don't believe this is a good approach for all apps, or for all people. I tried many times to make my notes reference each other, with backlinks and stuff, and I now believe that my brain is not wired for this stuff. Still, kudos for the author for making it work for them.
This looks like a nice way to think at common office tasks, but it doesn't need to be an OS - it's just a set of UI conventions participating programs would adopt along with the APIs to help those communicate.
Basically data records live free from applications, so for example you can make a pasteboard that contains an image, an email, a couple of contact addresses, a map, or whatever, without having to go into an image editor, an email app, etc to get the data, or to have to link/embed those apps as in microsoft's 90's OLE/COM object embedding stuff.
It differs from 'everything is a file', in that files contain a whole bundle of different data, and are 'opened' by a particular application or class of application, and are effectively a 'save state' of the application, whereas 'items' or data records or whatever are independent of any particular app; they have a type, and any app/tool that can deal with that type can operate on them.