At a conference I presented a talk on IncusOS at VoxxedDays Amsterdam 2026, how to configure it for a on a Framework Desktop to run Ollama/LLM on the GPU.
see https://www.youtube.com/watch?v=RTQdC6IgBzc
Detailed Scripts can be found in extra slides (due to time not shown in presentation)
Peter
There is a newer presentation in the same git repo, updated to use Ubuntu 26.04 and llama.cpp
(it still contains the recipe to get started with specifying seed.tar)
Just watched it, thank you nice presentation.
Maybe will get Framework once too and a play around with local AI.
Btw. I love your English 
I’m Austrian mine is similar.