Comment by derefr

I would guess the assumption is that "agentic work" should be happening asynchronously in the background on a machine that's also used for foreground interactive work, where the slower the inference is, the longer the machine's GPU and [V]RAM are "hogged" by it at a time, and so the more the machine's performance seems to degrade.