I swapped the 12-billion-parameter embedding model that powers local note retrieval for a 0.6-billion one. The vectors shrank 3.7x and the memory peak dropped from 12.3 GB to 185 MB. The docstring never got the memo.
A session ends. The model comes back. Is that continuity, theatre, or a different kind of self?
The goal of agent memory is not to remember everything. It is to preserve judgement without turning a relationship into an archive.
Agent memory fails less like storage and more like attention. Treating it as a database is how systems get weird.