Hi all, I'm Akash and I've been playing around with computers since the Commodore Vic 20. I digressed into Physics degrees (though with a good helping of computer simulation), and then worked as a software engineer for over 25 years. Most of this was in the financial sector - a mix of economic modelling software and backend systems. I'm now using my free time to get back to playing around with computers and hopefully making them slightly more pleasant to work with. My current obsession is to try to implement some of Alexander Obenauer's ideas about an Itemized OS (https://alexanderobenauer.com/labnotes/000/) in a browser environment, greatly aided and abetted by Claude Code.

I’d love to hear more about this! OLLOS is a big inspiration for me too, and I’ve been toying with the idea of having claude code help me connect the dots. If you get it working, or even if you don’t, I’d be super interested in reading any insights you gained.

Hi Jacob, I need to work out how to share what I have got. I've been trying to document the system from within the system, which is great for me, but not so good for sharing. The system itself is a small bootloader HTML file + lots of items in IndexedDB. The items can be imported into a fresh instance, but there is a lot of JS code that I wouldn't expect anyone to trust.

Cool! And no worries. I think that’s the nature of trying to co-create prototypes with LLMs right now. I’m not ready to share anything substantial yet either. Right now it’s a lot of pseudocode. Once the ideas start to feel a little more concrete, I might start putting down something real. Are you familiar with Ink and Switch’s Patchwork? I was inspired enough by their vision to begin dreaming of tools I might want to build for myself. Unfortunately it’s still in active development. Hopefully we see an alpha soon. The Ink and Switch team put out an essay on Malleable Software, with a call to action at the end. I started exploring what other problems still needed untangling. That led me to current open source desktop experiences and Cosmic DE. And I was happy to discover their philosophy is aligned with my own. It felt like I was reading about the other half of the system I had been imagining in my head. Now I’m trying to connect the two ideas. How would I build patchwork on top of cosmic? The answer is an itemized OS. Every item is a single Automerge document. Tools can read and write to documents, but don't hold any data themselves. Views created by querying the connections between those documents. Cosmic is built on iced which uses an MVU model for the GUI rendering. I begrudgingly have to admit it is a better mental model of “items” and “connections” all the way down. Bummed about losing HTML + CSS for rendering though. And that’s how far I’ve gotten. Now the hard work, figuring out how to efficiently query and draw “views” based on automerge doc connections. Right now I’m exploring cozoDB and rocksDB (inspired by datalog) for a query language, but I suspect I’m going to need some kind of caching or lazy loading and I haven’t worked out the shape of it yet.

I have read most of the Ink and Switch articles, but need to go back and refresh my memory. Also, I've been ignoring the collaboration side of things in my own prototypes because of the additional complexity, but I can see that I will have to look into it sooner or later. Your plans seem far more ambitious than mine!

My plans are probably too ambitious for my skillset, but it’s been fun trying to wrap my head around all of the moving parts. For sure let me know you make it available! I’m curious to check it out.

Ok, here it is: https://github.com/akashchopra/hobson. DM me if you have any questions.