Weekly notes from running local AI on my own hardware: the models and agents I self-host, and all the gotchas. Tested first on a Mac Mini M4, an RTX 5090, and a VPS.
