The same cue, a different context.

Context changes the familiar action

Stage 2 used perceived colour, shape and relative order to distinguish two scenes. The same cue selected a rightward action in one context and a leftward action in the other. Four teaching episodes and two cold-retrieval episodes completed in the physical simulator.

Those six new observations belonged to staging. They added no new primary observations. The mirrored scenes and repeated deterministic trajectories do not establish robustness across varied settings.

A quiet checkpoint before waking retrieval

Stage 3 checked provenance and evidence sufficiency during a bounded quiet phase. Two additional cold simulator actions reproduced roughly +301 and −297 millimetres of displacement. Staging now held eight observations; the primary store still held one.

The checkpoint did not improve physical performance over the earlier motor result. Recurrent replay remained off, and no new association was learned through autonomous consolidation. This is a verified checkpoint and retrieval pathway, with the more ambitious quiet-learning claim still open.