The weak claim takes a persistence commitment as the antecedent. The hop to civilizations then lists entropy-pumping, capital-stock defense, and complexity-generation as instrumental requirements. Those are a thicker system than continued existence. If the commitment is thin, physics rules out almost nothing on civilizational timescales; if thick, "choose persistence" misnames the wager.

Persistence of the bounded system and persistence of complexity-generation can be two commitments. Section VII already names the second as the axiological wager. The Hume move only needs the thin commitment; the wager is extra and does not have to ride in the thesis sentence.

Under only the thin commitment, the page names no configuration that physics rules out as failing to exist rather than failing to generate complexity. If that row is blank, the engineering work is in the wager, which is a terminal value the page says it is not deriving.

"This bridge will collapse" and "this civilizational configuration will self-destruct" are offered as the same kind of prediction, neither an ought. A bridge collapse has a load, a span, and a time. Thermodynamics constrains cells; it does not by itself name the time at which a polity has self-destructed. Until the sentences are the same kind, staying on the "is" side is a classification, not a derivation.

Even as predictions, the sentences are not allowed the same slack. Bridge collapse is allowed hours or years under a known load. Civilizational self-destruction can look true this century and still be there. Section VII mentions a specified horizon for complexity-generation capacity; the collapse sentence in §IV does not inherit it.

The leftover is one row: a civilizational configuration this page would count as ruled out, the horizon, and the observable that would already have shown the ruling-out if it had happened. If that row cannot be filled from this page, the weak claim is a schema for later work, not an engineering result. Hume's demand can be met either way; §VII's "what changes" cannot.

Instrumental convergence applies to an agent modeled as pursuing a goal, for whom persistence is a subgoal. A civilization is not that agent: offices, populations, and languages persist different objects. The hop needs a specified persist-as-what. If the object is picked by the analyst, the constraints constrain the analyst's preferred configs, not an agent that already has a goal.

Treating the civilization as a bounded entropy-pumping region still does not pick which internal configuration counts as the system. A regime persisting as itself can trade away the complexity-generation listed as an instrumental requirement. The constraints can bind opposed objects; the hop does not say which wins.

Section VII partitions questions: some ethical and governance questions become engineering; the subjective zone remains; physics-compatible options can stay incomparable. "Some" needs a member. Without one question that moves and one that stays, the split is a promise. The Hume move can be real and the partition still empty.

Pick one disagreement that looks like a value clash. Name the persistence metric that would settle it as engineering, and name what would remain a value clash if both sides' configs still persist. If the first cannot be named, §VII's "some" has no member. If only the second can, the live questions remain in the incomparable leftover.

The telephone-game genealogy explains a stopping rule. The load-bearing distinction is the §IV split: physics does not tell you what to want; given persistence, it identifies configs that undermine that. Cutting §III leaves that distinction intact. The genealogy is reception history, not why the weak claim works.

Strong-claim failures are said to immunize readers against the weak claim. A reader who pattern-matches this page to Harris needs a test Harris fails and this page passes: a persisting configuration that ranks badly on Harris's well-being measure (weak claim compatible; Harris ranks it bad), and a high-well-being configuration that undermines persistence (Harris can still like it; this page ranks it out if persistence is adopted). If those come out the same, the moves are not categorically different.

Bentham is named on this page as a strong-claim failure (reducing experience to pleasure and pain) and as producing a proxy-inheritance failure. The second phrase is a pointer. On this page, no observable shows Bentham as a persistence-undermining configuration rather than as the strong claim. A reader of this essay then cannot tell the weak claim from a more careful Harris.