Summary of Part I

Four claims carry the weight, and they are worth separating by how much support each has.

1. The lower rungs are cheap. Driven nonlinear systems under constraint generically produce structured attractors; among structured states, bounded ones persist differentially; boundary maintenance under uncertainty requires a compressed predictive state. Each step is standard physics or standard control theory, and the affect geometry that comes with the last of them appears in every seed the experiments have run.
2. The upper rungs are architecture-gated. Self-modeling becomes useful when actions close a viability-relevant loop — but usefulness is not sufficiency. Self-model salience, counterfactual capacity, and high integration arose only where the architecture was installed to support them. "Consciousness was inevitable" therefore weakens to: consciousness-relevant dynamics require specific architectures that are not generic.
3. Geometry is eigenvalues; dynamics is eigenskeleton. What modes a system has is cheap and universal. How those modes couple and twist as the system moves is expensive, and it is fixed by the sequence of crises survived rather than by substrate or task pressure. The forcing functions hypothesis conflated the two and was contradicted at the level of geometry by V10. In these substrates, populations put through repeated stress-recovery ended up with more mode coupling than controls matched on baseline integration (V19, V31) — which is evidence that the bottleneck does something other than filter, and is not evidence that stress is necessary for integration, uniquely productive of it, or good.
4. Normativity is grounded, and only agent-relatively. Valence is a real structural property, so the fact/value gap closes for the system bearing it. It does not close between agent-relative disvalue and agent-neutral obligation; it migrates there. Every downstream "ought" addressed to a third party is conditional on the identity thesis and on that unbridged step.

The chain, stated plainly: driven systems form persistent structure; some persistent structures maintain boundaries; boundary maintenance under uncertainty rewards compressed predictive state; when actions close a viability-relevant loop, prediction may reward self-modeling; counterfactual reasoning and strong integration require further architecture. These systems possess objective viability conditions, and valenced agents may instantiate states that are genuinely better or worse for themselves. Whether those dynamics are experience, and what obligations follow if they are, remain the questions of the parts that follow. This is not a further metaphysical question layered on top of the physics. It is a question about what integrated cause-effect structure is, intrinsically, when description shifts from outside to inside.