Ten days ago I had an experience with Gemini 3.8 flash that made me wonder if I was being routed to a different model under test. I was trying to use rocm with llama.cpp on my 128gb Strix Halo but could only get it to run Vulkan. I pasted the error message into agy and it proceeded to attach GDB to my GPU driver, reverse-engineer the kernel queue ioctl interface, and author an LD_PRELOAD C shim to get ROCm llama.cpp working on my Strix Halo. My jaw was hanging open the whole time.
Edit to add the fix: <a href="https://gist.github.com/birep/6f2c8d490c7a29820997d57bd654c351" rel="nofollow">https://gist.github.com/birep/6f2c8d490c7a29820997d57bd654c3...
This is why I think llms are a killer app for Linux desktop. They’ve been trained on Linux very hard, and it cleanly solves the “how do I make it do $thing” problem since everything is open and the llm can manipulate it. For example: Sound not working? Just tell the llm.
I’ve been playing with the ITS operating system on a simulated PDP-10. Despite it being from the 1960s, Claude has been able to do sysadmin work on it, including a very similar journey where it started disassembling code to identify the source of an obscure problem. I don’t want to take away from the learning experience (the whole reason I’m interested in this stuff), but it’s often almost impossible to find what I’m looking for online, and the LLM turns out to be a better resource when I want to learn how a particular subsystem works. I ended up making a custom MCP server so it could connect to the simulator more efficiently. <a href="https://github.com/masto/pidp10-mcp" rel="nofollow">https://github.com/masto/pidp10-mcp
Drifted off my point a bit, which I guess was meant to be it’s not necessarily Linux-specific training.
taylorfinley · · focus · HN ↗
Edit to add the fix: <a href="https://gist.github.com/birep/6f2c8d490c7a29820997d57bd654c351" rel="nofollow">https://gist.github.com/birep/6f2c8d490c7a29820997d57bd654c3...
plasticchris · · focus · HN ↗
masto · · focus · HN ↗
Drifted off my point a bit, which I guess was meant to be it’s not necessarily Linux-specific training.