Variable landing: does recall geometry reflect temporal identity?

Neuroscience already knows memories are reconstructed, not replayed — who you are when you retrieve something shapes what comes back. We tested whether that holds for an AI system with persistent memory: does the geometric signature of recalling the same thing change once real experience sits between encoding and retrieval?

A four-arm, pre-registered design on Qwen3.5-27B found the predicted ordering, with a clean zero floor on the no-intervention arm. At eleven memories, the primary comparison doesn’t clear correction — but the effect size is real enough to specify exactly how much power the next version needs.