Exploratory / Not experimental results
Ideas to sharpen a test, rather than conclusions drawn from one.
14 October 2026 / Exploratory framing
A record can survive a restart. A later context can use that record. A task can improve because it was used. These are separate propositions. Storage verification supports the first; input audits address the second; matched evaluation is needed for the third.
This framing is methodological, not a new measured result. A larger memory score cannot establish subjective experience.
14 October 2026 / Exploratory framing
A developmental program might build many sizes, but an ordinary loop can do that too. The proposed test should identify which capability is missing from the conventional builder, count supplied interpreter knowledge, and separate initial preservation from useful adaptation.
No training or search experiment is reported here. Draft protocols remain proposals.