hestermofet

Thank you. I'm the author. It's easy to get lost in the slop that AI can produce if you try to read the code or output, but the results speak for themselves. Other one-person OS hobby projects exist, but it typically takes years just to get to a shell that you can interact with. I got to a fully graphical operating system that runs on dozens of ISAs, can connect to the modern web using TLS, ships with its own native WiFi drivers, and even supports 3D accelerated graphics for playable video games in just a few months.

Yes, I am very clear that this is written entirely by an LLM. I "wrote" it using plain English.

Thank you for the kind words, I'm the author. Yes, this was a huge labour of love that took 100s of hours.

"i'm assuming if you know what the code is doing you don't need to audit them..." I frankly disagree. Really depends on how you define "what the code does". If you need to know what a particular function will result in when you put an input into it, sure, yeah, you need to know how that code works. But if a function is one in a hundred thousand of other functions that all work together to do "something", it matters less how that function is implemented. What the code "does" here is provide a platform to run whatever applications you could want. A music player. A web browser. A game.

An OS is an incredibly complicated system. I don't think anyone who is part of writing one these days knows what every line of code does, nor do they really need to be able to pick a random line and understand it. So what is the use of good code style and good documentation if nobody will read will ever be expected to pick a random line, read the documentation about what it does, and understand it? Made sense when code bases were relatively small, and humans were the only entities that could read code. Doesn't make as much sense now when an LLM can read an entire OS's hundreds of thousands of lines of code in mere seconds, and understand it almost instananeously.

All I care is that this operating system I designed all by myself is functionally equivalent at the minimum to Windows 95, which at least 1000 people working together wrote.

I don't care what the code does, and 99% of my repo is not meant for humans to read, it's meant for other LLM agents to read. That's why it's largely MD files, in a language only other LLMs can understand. What is meant for humans to read is here: https://unodos.arinbakht.com/manual/ How do you open a DOC file? How do I play an MP3? How do I browse the web? Those are the parts of an OS humans actually care about, and those are explained more plainly in the user manual. How does the bootloader hand over control to the kernel? Frankly, not that important.