Memory is not retrieval — it is inference

We talk about memory as though the brain were a filing cabinet: an experience goes in, sits unchanged, and is pulled out later intact. Almost everything we know about how remembering actually works contradicts this picture.

Memory is reconstructive. Each act of recall is a fresh inference about what probably happened, assembled on the fly from partial traces, priors, and the current context — and altered by the very act of recalling it. This is not a bug. A system that re-derives the past rather than replaying it can generalise, update, and imagine futures it has never seen.

This piece is the opening note in a longer line of thinking — that learning and memory are best understood as dynamical, subjective inference. More to come.