FORKB_ANALYSIS_PRESENTATION_2026-07-27.md

pipelines/FORKB_ANALYSIS_PRESENTATION_2026-07-27.md

FORK B — PRESENTATION + UX under Josh's revive-choice direction (deep dive)

Axis: what the player SEES from HP 0 to back-on-their-feet, under the just-RULED Fork A (continuity

by default, cut earned) and Josh's 2026-07-27 direction (choose: revive at last checkpoint, or WAIT

for a support companion trained in resurrection; out of combat only).

Read-only analysis. No repo file edited. Every claim below carries its source; where I could not

find a thing I say so and name the search that failed.

Short-form sources:

---

0. HEADLINE — the option set was a false trichotomy; Josh's direction makes it a state machine

PD §0.B offered four competing camera treatments for HP 0: the shipped fade (Opt 0), the hard cut

(Opt 1, recommended), the diegetic dissolve (Opt 2), the hold on the body (Opt 3). Josh's direction

does not pick one. It converts all four into sequential states of one machine, each of which now

earns its place for a reason none of them had as a standalone fork:

Historical optionIts new job
Opt 3 — HOLD ON THE BODYthe DEFAULT state. The camera stays down because the player has a decision to make from down there, and because the wait has to be watchable.
Opt 1 — HARD CUTthe FAST PATH, taken only when the player presses for it.
Opt 2 — DIEGETIC DISSOLVEthe TIMEOUT. The withdrawal completing on its own — the one place a dissolve belongs.
Opt 0 — the shipped 0.55s fadeRETIRED. It is the greybox stand-in for the machine, not a member of it.

The single most important consequence: the hard cut becomes PLAYER-AUTHORED. PD §0.B's own

strongest objection to Opt 1 was that "making death the one routine hard cut trains the player to

read a cut as failure, which then poisons every other legitimate cut in §3.2 — a boss intro cut

would inherit a death-flavoured dread." That objection dies under Josh's direction. A cut the

player pressed is not a cut the game imposed. It reads as the player's own impatience, not as the

world breaking, so it never transfers dread onto the authored cuts. Fork A's scarcity doctrine

survives intact, and the one exception it had to carve becomes an exception the player writes.

Second consequence: the frequency problem solves itself. A death loop the player sees fifty

times cannot be a fixed ceremony (the AAA mistake PD §0.B names) and cannot be a fixed fade (which

is identical forever and therefore invisible). Under this machine the fast player presses once and

takes the fastest transition in the game; the patient player gets a beat that is different every

single time, because the beat is *the live world around the body*, and the live world is never the

same twice. That structural variety is what a fade can never buy.

---

1. WHAT THE FICTION ALREADY IS — the wait needs no new metaphysics

This is the strongest finding in the pass. The defeat model is already RATIFIED canon and it

already contains the interval Josh's WAIT lives in:

"Defeat model RATIFIED as canon: consciousness withdrawal to the last attuned recharge source +
defeat writes a path-variable journal observation." [SRC: B1R §4]

and its derivation:

"On defeat, the protagonist's consciousness withdraws to the last attuned recharge source ...
Respawn is therefore an in-fiction event at an in-fiction anchor, and attuning sources becomes
the diegetic checkpoint verb." ... "the defeat model is a substrate event, not a UI event"
[SRC: RP5 §5.2, §5.5]

Withdrawal is a process with a duration, not an instant. Josh's WAIT is simply *the interval

before the withdrawal completes*. Everything falls out of that with no invention:

(wait). One button is letting go. The other is not letting go yet.

design a failure state — the canon already has one, and it is the same destination.

no in-fiction justification beyond the obvious one, which is the best kind.

And the ability exists, named, gated, in ACTIVE tier canon — I did not have to invent a verb:

Breath of Life — Tier 5 Philosophus. L1 Truly Good + Vril Healing Tier 5. Vril cost high.
Cast pattern instant. "Single-target revival ability — restores a downed ally to combat
consciousness with reduced vitality. The integrity gate enforces canonical path discipline; the
L4 and L5 evil paths access alternate revival mechanics through different element routing."
[SRC: ABIL L848]
Vril Revival — Tier 5 Philosophus, L1/L2 band, Vril Healing 70-99. Same shape. [SRC: ABIL L922]
Vril Mass Revival — Tier 7 Adeptus Major, L1 band, vril cost extreme. [SRC: ABIL L924]

Three presentation facts fall straight out of those rows and I treat them as binding, not proposed:

1. The revive returns you at REDUCED VITALITY. The rise is not a full restore. That must be

readable on the body, not on a bar.

2. The revive is INSTANT cast. No channel bar, no interruptible ritual. The ally kneels and it

is done. This is a mercy for the presentation: no progress meter on a death surface.

3. The dark path is not denied the revive — it gets a DIFFERENT ONE. "Alternate revival

mechanics through different element routing" is an explicit instruction that the revive image

FORKS on integrity band. The L4/L5 revive is its own picture, not an absence.

---

2. THE STATE MACHINE (the presentation spec)

Six states. Only S1 and S3 are new surfaces; the rest re-use shipped or ruled machinery.

              HP 0
               |
         [S0] THE FALL              camera inherits the fall; no cut, no lift
               |
         [S1] THE LOW HOLD  <-----------------------------+
           (world stays live)                             |
               |                                          |
     +---------+-----------+                        (the fight is
     |                     |                         still running;
 press RETURN         no press / press WAIT          player watches)
     |                     |                              |
 [S4] THE CUT         [S2] THE WAIT ---------------------+
  (hard cut,               |
   fastest in game)   +----+----------------+
     |                |                     |
     |          ally arrives          the draw takes hold
     |                |                     |
     |          [S3] THE RISE          [S5] THE DISSOLVE
     |         (diegetic revive)      (withdrawal completes)
     |                |                     |
     +----------------+----------+----------+
                                 |
                    wake at the attuned site (diegetic, camera already live)

S0 — THE FALL (frames 0-N, no player input)

The camera does not cut and does not lift. It inherits the fall. This is CC-4 (continuity of

subject, not of angle) applied at the one moment the subject stops being able to hold the camera:

the lens goes down with the body and settles low, near ground, off the shoulder or at the eye-line.

Explicitly NOT: a rise-and-orbit. An orbit around your own corpse is a film-camera read — it says

"look at your body," which converts a subjective camera into a spectator camera and makes the beat

a *death screen*. Under Fork A this is exactly the class of thing the doctrine exists to refuse.

Camera yaw selection is a real build requirement, not a detail. If the defeat camera keeps the

player's last look direction, roughly half of all deaths will stare at a wall, a rock, or the inside

of a cave — and the whole wait beat is unwatchable. The defeat camera must choose its facing from

the ARENA: bias toward the largest open sightline, and where a living ally exists, bias toward the

ally's lane.

That bias does a second job for free, and it is the best UX in this whole design:

**The availability of the WAIT option is communicated by the CAMERA, not by a greyed-out menu
item.** If someone is out there, you can see them. If nobody is out there, you see empty ground.
Show, don't tell — on the surface where a UI availability flag would have been most tempting.

S1 — THE LOW HOLD + THE CHOICE (the new surface)

The choice appears immediately — frame one, no entry hold. This is what makes the fast path

genuinely fast; a player who wants out is never made to sit through a ceremony first (PD §0.B's

over-bold failure mode).

Shape: the ruled fork-as-scene is the exact precedent and should be inherited wholesale rather

than re-invented. It already is: letterbox via the ONE shared DrawLetterboxFrame path, a diegetic

prompt centered under the top bar, de-authored player_option texts numbered and centered, **no

integrity integer, no internal vocabulary**, skip supported, and the scene subsystem enforcing the

single-framed-overlay invariant [SRC: HUD DrawForkScene L370-416; PD §2 CC-6, §3.2 fork row].

Differences from an ordinary fork, and why:

the opposite claim: the world did not stop, you did. A thin or absent bar keeps the live world in

frame, which is the entire point of the state.

is wrong here: it kills the two things that make waiting mean anything — the visible fight and the

visible approach. If a treatment is needed to separate the frame from ordinary play, take it at

the EDGE (a slow closing vignette, which doubles as the diegetic timer — see S2) and never from

the center of frame.

S2 — THE WAIT (the state Josh's direction actually adds)

Held on the body; the world runs. Three things happen in this frame and all three are presentation

work that does not exist today:

(a) The world alive. Rain still falls, the fire still burns, the last enemy still paces, the

water still runs. This is the load-bearing difference from a fade and it is why this beat can be

seen fifty times without going stale.

(b) THE INT AXIS MADE WATCHABLE. The single best presentation opportunity in Josh's direction.

Ally INTELLIGENCE is a build number the player spends stats on and then never sees. Here it becomes

visible behavior, once per death, in the moment it matters most:

INT bandWhat the player watches
lowthe ally is still swinging at something, or standing where it was. It does not read that the fight ended. It arrives late or not before the draw takes.
midonce the last enemy drops, the ally turns, registers the body, and comes.
highthe ally breaks off early — reads the room, clears or pulls the last threat off your body first, and arrives fast and with intent.

That makes "team composition is CONTENT" and the AAAAA balance doctrine's "a never-picked option is

a defect" *felt* rather than tabulated. A player who dumped INT sees the cost without a tooltip; a

player who invested sees the payoff in a close-up. No other surface in the game shows an ally's mind.

The approach must be staged, not pathed. A companion who jogs over on a nav mesh reads as an NPC

bug, and a bug on a death surface is the worst place in the game to have one. Four beats, all

readable from the ground:

1. The register — the head turn, the stop. They saw you go down.

2. The break-off — they leave what they were doing. Visibly a decision.

3. The traverse — in frame if the camera bias worked; if off-frame, carried by audio and by a

silhouette crossing the light.

4. The arrival — the kneel. They come into frame from above, which from a ground camera is a

composition the game gets exactly once.

(c) THE HONEST-FAILURE PRESENTATION — the biggest UX risk in the whole direction, and its fix.

A WAIT that resolves into nothing is indistinguishable from a bug. The player pressed "wait," sat

in the dark, and the screen dissolved. Did the system fail? Is there no companion? Did they not

learn it? Is it broken? The player cannot tell, and *not being able to tell* is the failure.

Rule: waiting always shows you something. Every negative outcome gets an image:

Why no reviveWhat the player SEES
ally is still fightingthey are visibly fighting. The player watches them try and fail to disengage. Fully legible.
ally cannot reach (distance/terrain)they are visibly moving toward you and the draw takes first. The near-miss.
ally has no vril leftthey arrive, kneel, and cannot. The reach that fails.
ally never learned itthey arrive, kneel, hold on, and stay with you while the draw takes. Presence without power.
no companion in this chapterthe camera bias found no ally, so the frame is open ground from frame one. Honest from the start.
the companion is DEAD (named NPCs stay permanently dead)empty ground where they would have been. Say nothing. The absence is the memory.

The "arrives and cannot" cases are *better* presentation than a successful revive, because they

carry the relationship. This is where the direction earns its emotional keep.

S3 — THE RISE (the revive moment)

Diegetic, no cut, and per the §5 Presentation-Ladder register.

probably backlit. It is the one held close-up in the game that the player receives rather than

performs.

"the novice's rough cast vs the Magus's liquid one," and Josh nodded it as a hard 5090 requirement

("every spell and ability and movement needs to be beautiful and fluid art"). The revive is the

ONE ability the player watches from the receiving end, in close-up, held still. Nowhere else is

the tier ladder more visible. A Philosophus-tier revive should look effortful; a mastered one

should look like nothing at all.

through different element routing" [ABIL L848]. The dark-path rise is its own image.

and the gameplay camera end at the same pose — CC-3 (return to the LIVE pose) is satisfied

trivially. The return to control IS standing up.

seconds, the edges of frame. Grounded in the ability rows, not invented.

Ordinal: the rise is SLOWER than an ordinary stand-up and FASTER than any scene exit.

S4 — THE CUT (the fast path)

Hard cut, immediate, then the diegetic wake at the attuned site with the camera already live. This

is PD §0.B's Option-1 recommendation kept intact — but now taken only on a press. It stays the

FASTEST transition in the game (CP-6 unchanged).

The consolation read rides here, skippable from frame one (CP-2).

S5 — THE DISSOLVE (timeout / the draw takes)

The one place a dissolve belongs, and the one place the whole loop needed one: the withdrawal

completing on its own. It is the slowest of the three exits and it should be, because the player who

reaches it chose to be there.

The diegetic timer. A countdown bar on a death surface is a UI object that destroys the hold.

The primary channel should be the world itself closing — a slow vignette, the light going, sound

thinning — the draw taking hold. See §6 for why that alone is not accessible, and what backs it up.

---

3. THE ONE PLACE I RECOMMEND AGAINST THE OBVIOUS READ: do not gate the OPTION on combat state

"Out of combat only" invites a static availability check at HP 0: enemies alive → hide the WAIT

option. That is the wrong build, on presentation grounds.

The fight can END while the player waits — most often because the companion finished it. A player

watching their ally kill the last enemy and *then* come for them is the single best expression of

this entire system, and a static gate at HP 0 makes it impossible.

Recommendation: the OPTION is always offered; the REVIVE ACTION is gated. The surface stays

constant; the world decides. This also removes the greyed-out-menu-item problem entirely (§S0's

camera bias is the honest signal instead), and it makes the "out of combat" rule invisible — the

player never learns there is a rule, they just observe that nobody kneels next to them mid-fight.

Bounded risk: a fight the ally cannot win means the player watches a hopeless fight for a while.

The draw always resolves it. That is the cap, and it is already the timeout.

---

4. THE STRINGS

4.1 What is on the surface TODAY, and why it must not survive

FString::Printf(TEXT("DEFEATED (no fail wall) -- %s"), *Consolation)   // HUD L548

Three defects in one greybox line, all worth naming because the final surface will inherit whatever

is not deliberately replaced:

1. "no fail wall" is internal design vocabulary on a player-facing frame. CP-3 violation. The

player must never learn the game has a *policy* about fail walls. Naming the mercy destroys it.

2. "DEFEATED" is system voice, and arguably taxonomy. A person does not think the word

"defeated."

3. The consolation payload is system voice too. BE_0002's death_reward reads: *"logs his stalk

cadence for the retry; unlocks a polo-corruption lore fragment."* Every load-bearing word there

— "logs," "retry," "unlocks," "fragment" — is a machine's word.

This is a live natural-voice-doctrine gap that the 2026-07-27 rewrite pass did not reach. The

pass covered "all 96 sidecar strings ×4 chapters + 14/16 boss display names" [SRC: commit ab5d32c].

The death_reward column is neither a sidecar string nor a display name. The highest-frequency

player-facing text in the game is still in machine voice.

And the natural-voice rewrite makes it strictly better, because it makes it *teach*:

The first promises a log. The second *is* the log, and it hands the player the counter. NVD §1's own

rule — "quest text is what someone SAID or what the protagonist THINKS — never a system's summary" —

applied here converts a receipt into a lesson. Same word count.

4.2 The register, calibrated to the shipped corpus

Ch-2's opening beat sets it exactly: *"It bit me and ran, and from the crack in the rock I watched it

stop moving. Get up. Climb toward the light."* [SRC: build/spine_rows/CH_02/beats.csv CH02_B01

player_objective]. First person, present tense, short, bodily, no explanation.

Two structural notes on that line:

telling themselves to get up. The death loop is a re-run of the arrival. That rhyme is free and it

is good — but it must be used with the Ch-58 HELD reveal intact (the Child Familiar canonically

escaped and is believed dead; the death surface must never re-stage that image or gesture at it).

4.3 Proposed strings (PROPOSAL-tier seeds; examples-not-limits; Josh/director authors final)

The prompt — the protagonist's own thought, not a question the game asks:

The two options — de-authored, in the fork-option register (compare the shipped Ch-2 fork

options: *"Let it hurt. Feel all of it, and keep walking anyway."*):

SeedWhy
WAIT"Not yet."two words, entirely the protagonist, names no mechanism, and is exactly what a person thinks at that edge.
"Hold on."the double reading — hold on to life / hold on, someone is coming — is doing real work.
RETURN"Go back."says what the player means. Does not name the water, the site, the substrate, or the withdrawal.
"Let go."the honest pair to "not yet"; slightly more final than the mechanic actually is.

Recommended pair: "Not yet." / "Go back." They are opposites, they are both what a person thinks,

neither names a system, and neither carries a promise the world may not keep. Critically, **neither

one promises rescue** — "Not yet" is a refusal, not an expectation, so the honest-failure cases in

§S2(c) do not read as broken promises.

Strings to refuse, and the reason:

"wait for revival." Mechanism-naming is both CP-3 taxonomy and (see §5) an L3 leak.

advisory signature in harness/check_reveal_discipline.py WHO_SOFT. Trivially avoidable; noted so

a future author does not walk into it.

4.4 The journal observation is a SECOND artifact — do not double-present

B1R §4 ratifies that defeat writes a path-variable journal observation, and RP5 §5.2 specifies

that each observation "concretely records one telegraph-and-counter the protagonist witnessed."

The consolation read and the journal entry are two different things and must not be the same words

shown twice. Recommendation: **the on-screen consolation is the observation's first line; the

journal holds the full entry.** One authored string, two lengths, one write.

Timing note that is nearly free: the journal activates at Ch 13 [SRC: VOL2 §11k slice-honesty

note] and Ntala joins at Ch 13 [SRC: CAST §1]. Both halves of the mature death surface arrive in

the same chapter. The surface can honestly be designed as *the thing that grows at Ch 13* — before

it, defeat is bare and solitary; after it, defeat has a witness and a record.

---

5. REVEAL DISCIPLINE ON THE DEATH SURFACE

The death surface is the highest-frequency player-facing text in the game and it fires from Ch 2,

long before almost every reveal. Five specific risks, in severity order:

R1 — the L3 theology leak (the big one). The ratified fiction is that consciousness withdraws to

an attuned recharge source. The player at Ch 13 does not know this, and the game's whole reveal

architecture is show-don't-tell with no thesis declaration. Any string that *explains* what death is

— substrate, vril, withdrawal, tether, consciousness — hands the player Layer-3 metaphysics on the

one surface they will see fifty times. The strings stay bodily and immediate. The player should

work out what the water is doing over many chapters, not read it on a death card.

R2 — Thread 6 contamination. "Dying-Rising God" is a canonical thread with authored anchors

(Osiris Ch 15, Eleusis/Dionysus Ch 23, and forward) [SRC: T1_Threads_22_Master_References L354,

L363]. A mechanical revive dressed in dying-rising imagery — light from above, a mythic register,

resurrection language — pre-empts those authored beats with a routine loop. **Keep the revive vital

and medical, never mythic.** Someone catching you, not someone raising you.

R3 — the boss name on the consolation. BE_0002 carries boss_name = "Rangku Wera, the Hollow

Hunter" and display_name = "the one who tracks me." The consolation surface must draw the

diegetic display name (or no name), never the registry name. The death_reward line itself must not

embed a canonical name.

R4 — the Ch-38 antagonist-pattern leak. Consolation lines about a boss must never attribute a

pattern or a hand behind events. check_reveal_discipline.py WHO_HARD/WHO_SOFT applies to authored

prose; these strings are short, repeated, and read at a moment of high attention, which makes them a

worse place for a leak than body prose, not a better one.

R5 — the Ch-58 HELD familiar reveal. Per §4.2: the death-and-wake loop rhymes with the Ch-2

opening. That rhyme must not re-stage the familiar's death image or imply the familiar.

---

6. THE ACCESSIBILITY FLOOR

Attunements is RATIFIED canon, not a proposal — Josh closed the naming 2026-07-04 and ratified

the layer: "always-available, judgment-free; includes the non-negotiable colorblind vril-polarity

multi-channel readability" [SRC: B1R §4]. So this section has a canonical home to land in, and the

death surface is squarely inside its remit ("shape how the demand reaches you").

The choice must be readable without color and without audio.

fork surface already gets this right: numbered, text, centered.)

sound off must still be able to tell that someone is coming: a visible silhouette / motion in frame

(the S0 camera bias is what buys this) AND an audio cue. Either channel alone fails a real player.

is a *low-contrast* one — players with low vision will not read it. Attunements carries an explicit

duration indicator for players who want one. Judgment-free, gates nothing, changes no content.

construction. The revive close-up needs a reduced-intensity variant. This is a hard floor item, not

a polish item — it is on the most-repeated surface in the game.

shake / motion sliders.

THE DEFAULT ON NO INPUT MUST BE "WAIT". Three reasons and they are all floor-level:

1. A player who is stunned, distracted, away from the pad, or using assistive input takes no action

and is not punished for it.

2. The timeout resolves WAIT into the checkpoint anyway, so no-input can never softlock. This is

CC-8 ("the hold is a floor, not a gate") and the no-dead-end floor, both satisfied by

construction.

3. It preserves the beat for the player who did not know there was one.

The checkpoint must therefore be an ACTIVE press. That ordering is what makes the whole machine both

merciful and fast.

The fast-loop preference. A player forty deaths deep who knows no companion is coming should not

re-take the beat every time. Attunements carries a presentation default — "at defeat, pre-select:

wait / return." It governs which option is pre-selected, never availability, so it is a

presentation preference and not a content gate, which is the line the judgment-free layer must not

cross.

---

7. WHAT IS ACTUALLY REQUIRED TO BUILD (presentation-side)

Positive-controlled where I claim absence.

**7.1 There is no combat-state authority in the engine. The "out of combat only" clause has nothing

to read.**

grep -rn "InCombat|bInCombat|CombatState|IsInCombat" over C:/dev/Humanity/Humanity/Source/

zero hits. A second pass for combat_active|EncounterActive|bEngaged|IsEngaged → zero hits. What

exists is per-enemy: EHumanityEnemyState {Idle, ..., Pursue} and

IsAggroed() { return bPerceivedThisFrame || State == Pursue ... } [SRC: ENEMY L49-53, L93], plus

AHumanityBossCharacter::IsDefeated() [SRC: HumanityBossCharacter.h L193]. So an aggregate is

*derivable* but does not exist. This is the gating build item for the entire WAIT branch, and it

is a presentation dependency, not just a rules dependency — under §3 the option renders always and

the world decides, so the world needs a mouth.

Sub-question that has no answer in the code and needs one: **do enemies de-aggro from a dead

player?** They will, on DeAggroRadius hysteresis, since a corpse stops being perceived — which

means "out of combat" may become true almost immediately after death for the wrong reason, and the

companion would walk over while a wolf is still standing two metres away. The combat-state read must

be arena/encounter-scoped, not per-enemy-perception-scoped.

7.2 The defeat camera does not exist. Today: StartCameraFade(0→1, 0.55s, black, hold) then a

timer to RespawnAtNearestSite then StartCameraFade(1→0) [SRC: CHAR L303-318, L352-359]. There is

no camera possession, no framing selection, no yaw bias. All of S0/S1/S2 is new.

7.3 The choice surface does not exist, but its precedent is shipped and should be inherited:

DrawForkScene + the shared DrawLetterboxFrame + NotifyForkOverlayActive's single-framed-overlay

authority [SRC: HUD L314, L370, L851]. Watch the invariant: the death surface is a *second*

blocking framed overlay, and BUG-0017's rule is "AT MOST ONE may draw." Dying during a held scene or

with a fork pending is a real collision and it needs an explicit precedence ruling — my read is that

the defeat surface outranks both, because the fork's own justification (it is the blocking choice)

applies harder to a choice made from the floor.

7.4 No ally actor exists. grep -rn -il "companion|ally|familiar" over the game Source returns

matches only in combat/QA/audio files (i.e. incidental words), with no companion class, no ally

pawn, no follower controller. Everything in S2 — the approach, the kneel, the INT-driven behavior —

is downstream of an ally runtime that has not been started.

7.5 The WAIT option cannot exist in the pre-5090 slice, and that is a ruling, not a gap.

Josh ruled "Ch 2-7 stays SOLO as built" — "the thinnest-witnessed stretch is a deliberate

arrival/assimilation choice; no early recruit" [SRC: CAST ▶ RULINGS]. The staged slice is Ch 2-5.

Therefore:

two with one greyed out. Under §S0's camera bias the player simply sees empty ground.

still, because the ability is Tier 5 Philosophus of the twelve-tier ladder [ABIL L848/L922].

gains a record at Ch 13, and gains a witness who can reach you sometime after. That is a good

story about the arc — the world starts empty and fills — and it should be built deliberately

rather than discovered as a bug report.

7.6 The consolation string layer needs a natural-voice pass (§4.1). Data-side, not engine-side:

the death_reward column across the boss registry, plus the HUD's own wrapper string.

7.7 The Presentation Ladder needs a tier key on the revive. §5's presentation_tier_ref seam

already exists as the design; the revive is its highest-visibility consumer and should be named as a

flagship case in the DR-2 tiered-asset request, since it is a held close-up rather than a

combat-speed flourish.

---

8. THE DOCTRINE DELTA — the replacement §3.2 row

PD §3.2's "Death / defeat" row is superseded. Proposed replacement, in the table's own register:

Beat classkeyENTRYBODYEXITReason
Death / defeatplayer HP 0DIEGETIC — the camera inherits the fall and settles low; arena-biased yaw; no cut, no liftthe LOW HOLD: world stays live, the choice is offered from frame one, and the WAIT state runs on the live world (the approach, the kneel, or the honest failure)player-authored: HARD CUT on the return press (fastest transition in the game) · DIEGETIC RISE on a revive (camera rises with the body; reduced vitality readable) · DISSOLVE on timeout (the withdrawal completing)The fiction breaks continuity here and nowhere else, but the break is now the PLAYER'S to take — which is why the one hard cut in ordinary play no longer trains dread onto the authored cuts (§3.2 Fork-A objection retired). Grounded in the RATIFIED defeat model (consciousness withdrawal to the last attuned recharge source) and the ACTIVE revival ability rows. REPLACES the shipped 0.55s fade.

Consequential edits elsewhere in PD:

historical field stays as provenance.

WAIT state can outlast it. Resolution: the WAIT state is not a HELD-FRAME beat class — it is a

player-paced live-world state closer to CC-9's dwell — so it sits outside the hold-class ordinal

ladder rather than topping it. Stating this explicitly prevents a future author reading a

contradiction into a ruled doc.

binds two exits instead of one.

same pose) and by the existing diegetic-wake rule on the other two.

exist), and Attunements.

home. It is the natural first real item for that doc, alongside the vril-polarity readability

floor that DECISION_DOSSIER already earmarked for it.

---

9. RESIDUAL RISKS AND THE STRONGEST OBJECTIONS

Objection 1 — "the wait is a slot machine." The player presses wait, and sometimes a thing

happens. Randomness on a repeated surface trains superstition, not understanding.

*Answer:* it is only a slot machine if the player cannot see the cause. §S2(c)'s honest-failure

presentation is what converts it from randomness into a legible system — you always see WHY. If the

honest-failure images are cut for budget, this objection becomes correct and the feature should not

ship without them. They are load-bearing, not polish.

Objection 2 — "two options at death is one option too many." Every death now asks the player a

question, fifty times.

*Answer:* the no-input default is WAIT and the fast path is one press, so neither audience is taxed

— but this is the objection I would most want tested in the QA loop's feel telemetry rather than

argued. It is a wall-clock question about a repeated surface, which is exactly the class the feel

strips measure.

Objection 3 — "it punishes the dark path twice." Companion loyalty is integrity-driven; a low

integrity player has fewer companions, so they get fewer revives on top of an already

harder-textured path.

*Answer, and it is canon's answer not mine:* L4/L5 "access alternate revival mechanics through

different element routing" [ABIL L848]. The dark path is explicitly NOT denied revival — it gets a

different one. Absence should occur only where the player specifically drove *that* companion away,

which is consequence, not tuning. If the build ships the dark path with no revive image at all, it

has contradicted an ACTIVE tier-doc row.

Objection 4 — "the low camera will look bad in half the arenas." True, and it is the reason

§S0's arena-biased yaw is a build requirement rather than a nicety. A defeat camera that keeps the

player's last look direction will stare at geometry constantly, and the wait state will read as

broken.

Objection 5 — "this is a lot of new work for a failure state." Also true. The honest framing is

that it is not a failure state: under the ruled Hades loop, death is a *content beat that advances

the world*, and it is the single most-repeated authored moment in a 30M-word game. Per-frequency,

it is among the highest-leverage presentation surfaces in the project.

---

10. FINDINGS FOR THE DIRECTOR (things I found that were not the question)

1. Live natural-voice gap. The death_reward consolation column and the HUD's

"DEFEATED (no fail wall) -- %s" wrapper were not in scope for the 2026-07-27 rewrite pass and

remain in machine voice, with internal design vocabulary ("no fail wall") on a player-facing

frame. CP-3 + CP-4, on the game's most-repeated string.

2. T1_UI_UX_Spec [DRAFT v0.1] is a 2-line file. Two ratified floor items already point at it

(vril-polarity multi-channel readability; now the defeat surface) and it has no body.

3. No combat-state authority exists in the engine (positive-controlled, two greps, §7.1) — the

"out of combat only" clause has nothing to read, and the de-aggro-from-a-corpse question is

unanswered.

4. The single-framed-overlay invariant (BUG-0017) has an unruled collision: dying during a held

scene or with a fork pending. Needs a precedence call before the defeat surface is built.

5. PD CP-6's ordinal ladder will read as contradicted by the WAIT state unless the

outside-the-ladder clause in §8 is added.

6. The Ch-2 opening beat and the death loop rhyme ("Get up. Climb toward the light.") — a free

presentation asset, with a Ch-58 HELD-reveal constraint attached.

Generated by harness/site/structure_site.py — the URL path is the repo path. review root