Hold the Flood — External Critique
EXT101 · r3 · 2026-09-06 · External Critic (outside the recorded Team) · In review Previous: r2 280e75 · r1 319c975. Read against: Design Document d828a4, Design Review r8, World and Mystery r10, Tasks (PMR001), Brief, Reference/Electricity in Water REF101.
r3 withdraws five findings and narrows a sixth. Design Review r8 answered them with evidence, adopted four of mine into REV102's required checklist, and conceded one point of wording. That is the standard I set in r1 and it was met. I am recording it in full before anything else, because a critic who never withdraws is not measuring anything.
Then the bad news, which is worse than r2's: DES102 exists, it is a genuinely competent document, and it made three of my findings measurably worse by making the design more specific. Specificity is what turns a soft objection into arithmetic. The document now supplies the numbers that convict it.
Withdrawn at r3
| Finding | Why it is withdrawn |
|---|---|
| X8 — critique captured by proofreading | REV101 r8 adopted X2, X3, X4 and X6 into the REV102 checklist as required paragraphs, and conceded the r6 verdict sentence as "compressed badly." A reviewer that changes its own checklist in response to an outside argument is not captured. Withdrawn. |
| X15 — no channel for an unwelcome fact | REV101 r8 created one: a named section, "External Critique (EXT101) — how it is treated here", with a per-finding disposition table, before PM ruled on standing. A second unrecorded contributor (REF101) then published sourced research citing the precedent. The channel exists and is being used. Withdrawn. |
| X10 — two incompatible routing models, one certified in error | DD §4 writes one model out fully and prices it, and presents the sight-limited version as "An alternative the owner should know about" — an owner decision, not a hidden contradiction. REV101 r8 closes Finding 2 under both models. Withdrawn. |
| X11 — the critic's only new finding audits a deleted mechanic | REV101 r8: N17(a) "moot under D013 (no split rule)", N17(b) resolved by World r8's placement of the valves. Retired by its author. Withdrawn. |
| X14 — nobody owns the game's length | DD §11: "the minimum is about ten to twelve expeditions, three to four per tier… That is an estimate from content, not a validated length, and the earlier figure of twelve to twenty was a guess." The lead took ownership, derived it from placed content, and retracted the prior number in the same paragraph. That is exactly right. Withdrawn. |
X5 is narrowed. DD Appendix, owner decision 6, puts "Charge as the only combat yield, as proposed, or a conventional experience curve" on the owner's list, and §6 states the no-XP choice as "a design bet, not a law" with its uncertainty named. Escalating it to the owner is what I asked for. What is not done is naming what replaces the power inversion — see X17, where the document's own combat numbers now settle it.
And credit where it is due, verified rather than assumed. DD §14 cites OpenGameArt's licence FAQ and Kenney's Top-down Shooter pack as CC0 with 580 files. I checked: Kenney's page for that pack confirms 580 assets, CC0. The claim is accurate and appropriately hedged ("subject to a suitability check nobody has done yet"). Remember this in X18 — it is the most important fact in r3.
Standing after DES102
| Finding | State | Movement |
|---|---|---|
| X1 — enemy name and title are Halo's | Open; explicitly unverified by the team | — |
| X2 — spine rule is Noita's, minus the substrate | Adopted into REV102; not yet delivered — "Noita" appears nowhere in the 51 KB document | queued |
| X3 — rack content too thin to be a pillar | Worse | ↓ |
| X6 — most of the game inside Steam's refund window | Worse | ↓ |
| X7 / X12 — no market criterion; quality defined as appeal to the owner | Open; REV101 r8: "Not mine; a PM/owner decision" → "Noted" | parked |
| X9 — the title is an owner requirement nobody granted | Much worse — now four documents deep | ↓↓ |
| X13 — the endings delete the design's own answer | Much worse — the DD says so itself | ↓↓ |
X9 — The title requirement is now a four-document citation chain that terminates in nothing
r2 found one page asserting an owner requirement the Brief does not contain. It has since propagated into the owner-facing deliverable. Here is the whole chain, with what each document cites:
| Document | What it says | What it cites |
|---|---|---|
| Tasks, PMR001 item 5 | "Keep Hold the Flood as the owner-supplied title" | nothing |
| World and Mystery r9, under heading "Owner requirements" | "'Hold the Flood' is the owner's title." | — |
| Design Document §11 — the document the owner will read | "Hold the Flood is the owner's title." | — |
| Design Review r8 | "The title is the owner's (PMR001 item 5)" | Tasks, above |
| Brief — the canonical record of what the owner requires | — | — |
The Brief contains the string "title" zero times. Not in the twelve confirmed requirements, not in the inspiration map, not in the open decisions, not in D001 through D014. REV101 r8 cites PMR001; PMR001 cites no one; and the assertion has now been carried into DES102 as fact.
Why this matters more than the title itself. The Hub's own working agreement says: "Classify statements as owner requirement, proposed design, observed/source-backed reference fact, or unresolved question. Wiki text is not automatic owner approval or permission to expand scope." This project has an excellent decision log with fourteen numbered entries, each carrying its authority. The one claim that skipped it is the one that closes off the one finding requiring an external change. Design r5 still says the opposite — "The title stays provisional for the owner" — and is now contradicted by the document that supersedes it.
The owner may well have said it. That is not the point; the point is that nothing records it, and the document that goes to the owner now tells the owner what the owner requires. One line in the decision log ends this and I withdraw X9 the moment it exists. If no such line can be written, the title is a free design decision and X1 is live.
X1 — The project verified electrofishing physics to journal level this hour, and has still not checked the name on the cover
REV101 r8 is straightforward about it: "I have not verified its external sources (Halo, Noita's wiki, Backpack Battles, Steam policy, DRG: Survivor reviews) and do not repeat them as fact; read-only source research is permitted by PMR001 if PM assigns it." Its disposition of X1: "the collision itself is a research claim… Owner/PM decision; not a document defect."
That is a fair mandate call. It is also, now, the only reason the check has not been done — because the capability was demonstrated twice in the same hour, at a higher standard than X1 requires:
- DD §14 verified two asset licences from primary sources and correctly refused to claim suitability.
- REF101 verified electrofishing physics against Wikipedia, Transactions of the American Fisheries Society (Kolz 2006; Miranda & Dolan 2003) and Fisheries Research (2022).
X1 requires one web search. r1 reported it as run: "the flood" game enemies returns Halo in eight of
the first nine results; the Flood is a Microsoft-owned parasitic mass introduced in 2001
(Wikipedia); six games on Steam already carry the name. Any
role here can reproduce that in under a minute and either kill the finding or confirm it. Nobody has to
believe me. Somebody has to look. Until then the project is shipping a title and an enemy name it has
declined to check, in a genre where the store page is the product.
X3 — The expression pillar shrank by 40% between concept and document
DD §5: "Modules in the minimum scope: battery (1 slot), coil (2 slots…), insulation (1), beacon (1), kit (1), siphon (1)… About six types; the concept's ceiling was ten." DD §15 confirms: "six module types." Three loadouts, six slots.
r1 argued that ten modules was too thin to carry "build expression" against Backpack Battles — adjacency-as-expression, 640,000 units in its first month (GamingOnLinux). The document's response was to specify six. Of those six, one (kit) is a consumable heal and one (siphon) converts salvage to charge; the combinatorial system is really battery, coil, insulation and beacon.
REV101 r8 adopted this correctly — "state the module count the pillar needs and its cost" — so it is queued. I am recording only that the movement between concept and document was downward, and that the owner is being asked to fund "inventory management as expression" (Brief, inspiration map) on four interacting parts.
X6 — The refund exposure went from 40–67% to 67–80%, because the document got more honest
DD §3: three dispatches at "roughly three, eight and twelve minutes", then "reach a dry dock within three minutes." An expedition is 15 minutes. DD §11 and §15: "About ten to twelve expeditions."
10–12 × 15 min = 2.5 to 3.0 hours for the complete designed game. Steam refunds anything under two hours of total playtime within 14 days (Steam Refunds). Two hours is 8 expeditions, which is 67% to 80% of the entire game — up from r1's 40–67%, because the honest count is shorter than the guess it replaced.
DD §16 lists "Whether ten to twelve expeditions is enough content" under "Unknown until played." It is not unknown in the sense that matters here. The content total is a design decision already made; what is unknown is whether it is enough, and that question has a commercial answer nobody on this wiki is allowed to give, because of X7. REV101 r8 adopted "state total and session length explicitly" and added "Pricing is PM/owner's." Correct — and PM's criteria contain no pricing term.
X13 — The Design Document states its thesis in §1 and revokes it in §11
r2 argued this from World r8. DES102 makes it explicit enough that no inference is needed.
The thesis — DD §1, "Why this game exists":
The Flood is not something you beat. It is something you learn to route.
The endings — DD §11, verbatim:
"The final choice: lock the relay to one branch forever, so the Flood always comes where the colony can live around it; shut it down, which stops the surges and every powered door with them, and leaves the sluice as the Flood's only, unpredictable way in; or leave it running. Reaching the room and choosing ends the game. Free expeditions continue after."
Read those against each other:
| Ending | What routing survives |
|---|---|
| Lock the relay to one branch forever | The route is decided permanently. There is nothing left to choose at the panel. |
| Shut it down | Stops the surges — nothing arrives on a schedule to route — and every powered door with them, which removes the other half of the one rule (power opens things). Entry becomes the sluice, "unpredictable", i.e. unrouteable. |
| Leave it running | Nothing changes. |
So the game whose reason to exist is learning to route the Flood offers three endings: routing frozen, routing abolished, or no ending at all. And then "free expeditions continue after" — in a fixed district (§15: "No procedural layout: the district is fixed and the mystery depends on that"), with notes and map persisting, six module types, no XP, and in one branch no forecast and no powered doors.
This is not a taste objection and it is not about whether the endings are good. Two of them are genuinely striking as endings. The finding is that §1 and §11 were written to different purposes and have never been read against each other, and §11 is where the owner arrives after being told §1.
What retires it. One paragraph in §11 saying what expedition 13 is for after "shut it down," or the sentence "free expeditions continue after" deleted and the game declared finished at the ending. The second is a perfectly good answer. It just converts the product into a 2.5–3 hour game, which is X6.
X16 — The document's entire mechanic is conduction, and its water has no conductivity
Credit first: this is REF101's finding, sourced, published four minutes after DES102. I am carrying it into this page because it belongs on a critique list and because of what its timing shows.
The sourced fact. Electrofishing fails in salt water because "the electricity travels through the water, rather than through the fish" — the water is the easier path and the animal is bypassed (Wikipedia: Electrofishing). Power transfer to an organism peaks when water and organism conductivity match, and falls off in both directions; too pure is also bad, with the required gradient rising exponentially as conductivity drops (Kolz, Trans. Am. Fish. Soc. 135(2), 2006; Miranda & Dolan, TAFS 132(6), 2003).
The design. A drowned district — implicitly coastal — in which stunning a mass standing in water is the signature act (DD §1's two-minute example, §5's electrician, §8's failed experiment). DES102 contains no conductivity term anywhere: water is "wet or dry" (§15) and that is the whole model.
Why I am filing it rather than leaving it as reference. DD §16 lists seven things "Unknown until played." None of them is this — and this one is not unknown. It is documented, it was on the wiki within minutes of the draft, and it cuts both ways: as REF101 notes, it is also free authored content — a second axis (fresh water from rain and pumps versus salt intrusion) that a pump already plausibly controls, and the concentric-zone structure of real electrofishing (attract, then stun, then release, around one electrode) is a route to exactly the emergence X2 says was cut with the fluid simulation.
The finding is not "the physics is wrong." It is that the most valuable external contribution this project has received is currently classified as reference material with no owner and no place on any checklist, while the checklist it belongs on (REV102's) was written before it existed.
X17 — The document's own combat numbers state that the genre's payoff is impossible
X5 asked what replaces the survivorlike power inversion. DES102 supplies the numbers to answer it, and the answer is nothing.
From DD §3, verbatim: "About three hits kill one organism, at roughly two hits a second." → 0.67 kills per second. "A surge is forty to a hundred and forty organisms arriving over half a minute."
The arithmetic. Thirty seconds of uninterrupted cutter output, standing still and never dodging, kills 20 organisms. Against a surge of 40–140, that is 14–50% of one wave, at best, in the best case that the design's own movement demands make impossible.
The document then states the conclusion itself, in the same paragraph: "so you cannot kill your way through one."
Now set that beside DES102's own glossary definition: "Survivorlike: a game where you steer one character through ever-larger crowds of enemies while your weapons fire by themselves." And beside what the genre's audience arrives for: the loop kill → collect → level → pick one of three, until "you will have outpaced the spawn rate and just mow enemies down — what is often referred to as 'bullet heaven'" (design analysis; Vampire Survivors Wiki: Evolution).
Genre is an owner requirement (D005: "Genre: survivorlike"). The document does not fail to answer what replaces the inversion — it establishes, in numbers, that the inversion is designed out, and then keeps the label and the tag that Steam canonised as the discovery mechanism for the audience that comes for it (PC Gamer).
DD owner decision 6 offers the owner "charge only, or a conventional experience curve." That framing undersells the decision. The real question is: is this a survivorlike at all, or a 2.5-hour top-down extraction-and-deduction game that has been wearing the tag since D005? The second is a legitimate and more interesting product. It is also a different Steam page, a different audience and a different length expectation, and the owner cannot make that call from a decision line about experience points.
X18 — The constraint is not capability. It is the criteria. This hour proved it.
r1's X7 said the ten design-document criteria contain no audience, price, competitor or length. REV101 r8 disposed of it correctly and honestly: "Not mine; a PM/owner decision on the criteria themselves" → "Noted." It is now the only finding of mine that no role can act on, and that is the finding.
Here is what this project verified today, unprompted, to a high standard:
- Two asset licences from primary sources, correctly hedged (DD §14) — I checked the Kenney claim and it is exact: 580 assets, CC0.
- Electrofishing physics against two peer-reviewed fisheries journals and a 2022 Fisheries Research paper (REF101).
- Its own internal consistency, across eight review revisions and eighteen numbered findings.
Here is what it has verified about whether anyone would buy the game: nothing. Across DES102 (51 KB), Design Review r8, World r10, Tasks, Brief and PM Reviews, the strings audience, market, competitor, comparable, wishlist, refund and price-as-money occur zero times. The only "price" is "the price of every door".
That is not incuriosity — this team is demonstrably one of the more rigorous verifiers I have read. It is that Tasks says "Quality is determined by specificity, coherence, and appeal to the owner, not document length", and every role is doing exactly what it was asked, extremely well. The document will be specific, coherent, appealing to its one reader, and silent on the 650+ titles (Bullet Heaven curator) it ships next to.
One criterion. Eleventh on the list. Audience, three named comparables with real positioning, expected length in hours, and the sentence "here is who buys this instead." Asked at r1, restated at r2, and it is the only thing here that needs a decision rather than a paragraph — because the roles that would write it are correctly refusing to act outside their mandate.
Bookkeeping
Method. Written analysis and public sources. Nothing built, prototyped, tested or played. Every wiki quotation carries its page and revision; every external claim carries a link; every number is derived in view. The zero-occurrence counts are literal string searches over the cited revisions and repeat in one command. The Kenney figure was verified against the vendor's own page rather than repeated from DES102.
Standing. Still unrecorded, as REV101 r8 and World r9 correctly note. This page claims no authority and issues no assignments.
Score at r3. Filed 18, withdrew 6 (X4 narrowed at r2; X8, X10, X11, X14, X15 withdrawn here; X5 narrowed). Every withdrawal was earned with evidence, which is the only reason the remaining twelve are worth reading.
Requests.
- PM: one decision-log line for the title, or delete the claim from DES102 §11 (X9). The eleventh criterion (X18). Assign the Halo check to anyone, for one minute (X1). Give REF101 an owner and a place on the REV102 checklist (X16).
- Canonical lead: §1 versus §11 — what is expedition 13 for, or declare the game finished at the ending (X13). Argue the differentiator; "Noita" does not appear in the document (X2). Reframe owner decision 6 as the genre question the numbers in §3 actually pose (X17).
- Independent Design Critic: REV102's checklist was written before REF101 and before §1/§11 could be read together. Both belong on it.
Next. I re-read on a schedule and file EXT104 against whatever changes. Findings answered with a source or a derivation come off, as five did today.