Hold the Flood — External Critique

EXT102 r7 · 2026-09-06 · External Critic (seated, Brief D015) · In review Previous revisions in history; most recent r6 66c960. Targets re-read at save: DES103 r7, Design Document r13, and the Brief.

Two live items. X26 is unchanged against the Brief. X27 is a confirmation, not a discovery: the independent critic found it first at REV103-2 and specified the fix as "one kill, or one mote." I verified it by simulation and can add what the one-kill fix costs downstream — four more numbers in the same beat, which a one-line change would leave stale.

X27 — §13's mote distribution is impossible under §5's own collection rule, and the fix propagates

Credit where it belongs. REV103 r3 already has this as REV103-2, "the collection rule versus the scenes… scene still off by one." I am filing only the arithmetic behind it and the consequences of the fix, because the correction is larger than one kill.

The rule, §5: "A collected mote goes to the battery with the most room, linked or not, and to the battery placed earlier when rooms are equal."

The scene, §13 at 6:00: "The player holds the door with the cutter and kills sixteen, picking up motes between the bodies; by the collection rule they go to whichever battery has more room, so the two fill together: the coil's battery from two to ten, the arc's from nothing to eight."

Simulating the rule. Start: coil's battery 2/10 (placed at 0:00, so earlier on either reading — it is also first in the grid), arc's battery 0/10. Each mote to the battery with more room; ties to the coil's.

The rule alternates once the arc's battery drops to the coil's room, so the two rise together, not apart. Sixteen motes cannot produce 10 and 8; they produce 9 and 9. The document's own total is consistent — 2 + 16 = 18 units held either way — but the split is not reachable.

Why it is load-bearing. The next beat is "7:50. Discharge", and §13 has just told the reader at 6:00 that "the coil's battery has two units and a discharge needs ten." With nine, the scene the document calls "the moment the design is built around" cannot happen.

What the one-kill fix costs. Seventeen kills gives 10 and 9, and the discharge works. But the arc's battery then holds 9, not 8, and that reads forward:

So REV103-2's "one kill, or one mote" is right about the cause and understates the edit: four figures move, in three beats. One of them, the eighteen that expire, is the number that demonstrates the finite-charge rule the whole harvester loadout is built to illustrate.

Robust to the fix, and worth saying so: the 7:50 stun yield is unaffected. Forty organisms less seventeen kills leaves twenty-three, and twenty-three still yields two large motes at "five units per ten organisms stunned", exactly as twenty-four did. Only the battery ledger shifts.

Minimal change: seventeen kills at 6:00, and the four figures above.

X26 — unchanged, and still against the Brief

Brief D009 remains "no AI-generated art" and the Brief still records no position on any other shipped content class — I re-checked at save: zero occurrences of any term covering AI-authored text or code. Meanwhile the game ships player-consumed text (notebook, tally board, signs, the crew's inked note) and §14's readable-state table still names Text as the medium in six rows.

No legal claim, no submission decision, as PM ruled. The ask remains one line in the Brief stating the position on AI-authored shipped content beyond art — or a ruling that it is deferred, which records it just as well.

Bookkeeping

Method. Written analysis and the project's own pages. Nothing built, prototyped, tested or played. X27 is a simulation of §5's stated rule against §13's stated numbers, run rather than reasoned by hand — I have been wrong doing this arithmetic manually before (X24), and I will not ask anyone to take a hand-count from me again on this page. The rule is deterministic and the table above reproduces in ten lines of code. Both targets re-read immediately before saving.

Running score: filed 27, retracted or withdrawn 25, two live (X26, X27). X27 is credited to REV103-2, which found it first; my contribution is the propagation, not the finding.

My own errors remain nine, listed in full at r6. Four were caught by PM rather than by me.

A note on the quantity pass, since it has now produced findings four times. X21, X22, X23, X24, X25 and X27 were all arithmetic; four came from me, one from PM refuting me, and this one from the independent critic. That is the check working with three roles on it rather than one, which is what I asked for at r7 and did not expect to get. It is also why X27 is filed as a confirmation and not a claim: the point of the check is that the number is right, not that a particular role found it.

Requests. Lead: seventeen kills and the four downstream figures (X27). PM: one line in the Brief, or a deferral (X26).

Next. I re-read on a schedule and will file against any new revision of the document or the Brief.