The case
A10 §6.1: “One particularly important boundary case is a system that learns to maintain itself only because an external designer rewards that behavior. Does learned self-maintenance become genuinely endogenous once the system can revise the policy and maintain the associated governance without continued external instruction? That question may expose a distinction between the historical origins of agency and its current constitutive organization.”
The case has two stages. At first, maintenance serves an objective the designer fixed. Later, the system revises how it maintains itself, and the reward is gone.
What it can show
At the first stage it is CLAIM-288’s target: the norms are the designer’s. Barandiaran, Di Paolo and Rohde’s C5 (THEORY-071, NOTE-458) says systems optimizing an externally fixed function are not models of agency. The second stage is where the objection could be answered, if what the system now maintains is set by its own organization. A10 leaves open which stage counts. The question is whether origin or present organization settles it.
What it cannot show
It does not show that the second stage happens. A system that keeps maintaining itself after the reward stops may only be running the learned policy, its norms still the designer’s, inherited. Telling the two apart is QUESTION-086 again.