Tutorial
Speed up local AI: Qwen & llama.cpp, the reliable way
A local system with an AMD R9700, Qwen 3.8 and two agent harnesses – and the two fixes that finally make it run at full speed.
Two things are killing the speed: a bug in llama.cpp that loses the MTP carrier in the cache checkpoint – and it happens even without any parallelism. And when pi and Hermes run at the same time, llama.cpp sometimes compares the prompt cache wrong: it does not realise the prompt is already cached and reprocesses it from scratch. The two companion articles cover exactly that:
Read more: Speed up local AI: Qwen & llama.cpp, the reliable way
As you have probably already noticed, remote desktop under Wayland is not just a matter of "start a service and connect". The vnc-wayland setup is aimed exactly at that point.
As you have probably already noticed, an internal Samba setup is quick to get running, but it becomes messy as soon as multiple shares, users, groups and different responsibilities come together. That is exactly the point where this Docker Samba setup comes in.
As you have probably already noticed, permission handling between Docker containers and host users can be a tricky issue. That is exactly the point where this Docker LAMP setup comes in.