Comment by ImPrajyoth

OP here. this is a cursed project lol, but i wanted to see: What happens if you replace the OS scheduler with an LLM?

With Groq speed (Llama 3 @ 800t/s), inference is finally fast enough to be in the system loop.

i built this TUI to monitor my process tree. instead of just showing CPU %, it checks the context (parent process, disk I/O) to decide if a process is compiling code or bloatware. It roasts, throttles, or kills based on that.

Its my experiment in "Intelligent Kernels" how they would be. i used Delta Caching to keep overhead low.