UP-023 · Persistence and World · STATED CHECKABLE RUNNABLE EXECUTED STAGED

Exact Replay

A run of a sealed world is a film that replays byte-identically; a divergence on replay is the world's, not the player's.

Standing

STATED CHECKABLE RUNNABLE EXECUTED STAGED PUBLISHED REPRODUCED

Last checked

never run

Source

none — this record cites no check

Limit

No receipt stating a fidelity percentage was ever located (v0.2 review, finding 7), which is why this page carries a '?' where that number would go. TRVM's film replay is not verified by this build.

Next rung

in_tree — locate or produce the receipt. The claim is older than the search for its evidence, which is the defect this field exists to expose.

Intent

When identity is a seal and the observable is complete, a run is reproducible: the film of one run replays to the same bytes on another player. If it does not, the world hid state or the seal lied, and the replay has found a bug in the world, not in the player.

Problem

Reproducibility is claimed and rarely measured. 'It replayed' usually means a log was read back, not that a second player produced the same bytes from the same sealed world. When two players disagree, the player is blamed, because the world was never sealed to begin with.

Solution

Seal the world. Record the film with its reduction strategy. Replay on a second player. Compare bytes. Publish the fidelity with the method, or publish '?'.

Real-world analogy

A player piano roll. Two pianos, one roll, the same tune — or the roll is damaged, and it was never the piano.

Structure — on the surface

An illustration on a compute surface: loci above, carriers below. Press Play or Step; the takeaways collect as you go. Nothing here is evidence — the witness section is.

The chapter in WRL — and the chain so far

Chapter 23 of 32 — the fragment _patterns/wrl/chain/exact-replay.wrl, sealed alone by wrl.js

; EXACT REPLAY — its OWN clock domain into a relay chain and a Door; the runner reduces this world TWICE and
; the build refuses unless both runs produce identical Film hashes. er_a is Part V's router in the chain.
[pulser:er_clock](every 2){sig_out}
[relay:er_a]{sig_in, sig_out}
[relay:er_b]{sig_in, sig_out}
[door:er_player]{sig_in}

[er_clock] --sig--> [er_a]
[er_a] --sig--> [er_b]
[er_b] --sig--> [er_player]

module + bench seal to → sem-bccd2ab7d2bba016be4313f2d890cc2bf6299327626dce84454a14cbf7c3bb36

Run inputs — a ScenarioV1, the forge's own document, bound to this world's id and never part of it (D3) · reduced twice, hashes compared · ScenarioDigest scen-d1c7ffa2c480644c1d6… (the run inputs' own identity, computed by the forge)

No claims: the world runs on its clocks alone for 8 epochs.

Reduced by the native reducer (ic32); the reference reducer reproduces every epoch's film hash (0.007s).

Reduced twice from a fresh state: every epoch's film hash identical — exact replay, witnessed on one host.

FILM v0.7
t=1
pulser:er_clock:mode=periodic,p=2,phase=0,armed=0,done=0,nf=2
door:er_player:open=0,next_open=0
relay:er_a:cur_out=0,next_out=0
relay:er_b:cur_out=0,next_out=0
wire:w__er_a__er_b:cur=0,nxt=0
wire:w__er_b__er_player:cur=0,nxt=0
wire:w__er_clock__er_a:cur=0,nxt=1
admit:policy=admit_candidate_min_firstreceipt_v1,fact_capacity_fault=0,receipt_capacity_fault=0,capacity_fault=0

FILM v0.7
t=2
pulser:er_clock:mode=periodic,p=2,phase=0,armed=0,done=0,nf=1
door:er_player:open=0,next_open=0
relay:er_a:cur_out=0,next_out=1
relay:er_b:cur_out=0,next_out=0
wire:w__er_a__er_b:cur=0,nxt=0
wire:w__er_b__er_player:cur=0,nxt=0
wire:w__er_clock__er_a:cur=1,nxt=0
admit:policy=admit_candidate_min_firstreceipt_v1,fact_capacity_fault=0,receipt_capacity_fault=0,capacity_fault=0

FILM v0.7
t=3
pulser:er_clock:mode=periodic,p=2,phase=0,armed=0,done=0,nf=2
door:er_player:open=0,next_open=0
relay:er_a:cur_out=1,next_out=0
relay:er_b:cur_out=0,next_out=0
wire:w__er_a__er_b:cur=0,nxt=1
wire:w__er_b__er_player:cur=0,nxt=0
wire:w__er_clock__er_a:cur=0,nxt=1
admit:policy=admit_candidate_min_firstreceipt_v1,fact_capacity_fault=0,receipt_capacity_fault=0,capacity_fault=0

FILM v0.7
t=4
pulser:er_clock:mode=periodic,p=2,phase=0,armed=0,done=0,nf=1
door:er_player:open=0,next_open=0
relay:er_a:cur_out=0,next_out=1
relay:er_b:cur_out=0,next_out=1
wire:w__er_a__er_b:cur=1,nxt=0
wire:w__er_b__er_player:cur=0,nxt=0
wire:w__er_clock__er_a:cur=1,nxt=0
admit:policy=admit_candidate_min_firstreceipt_v1,fact_capacity_fault=0,receipt_capacity_fault=0,capacity_fault=0

FILM v0.7
t=5
pulser:er_clock:mode=periodic,p=2,phase=0,armed=0,done=0,nf=2
door:er_player:open=0,next_open=0
relay:er_a:cur_out=1,next_out=0
relay:er_b:cur_out=1,next_out=0
wire:w__er_a__er_b:cur=0,nxt=1
wire:w__er_b__er_player:cur=0,nxt=1
wire:w__er_clock__er_a:cur=0,nxt=1
admit:policy=admit_candidate_min_firstreceipt_v1,fact_capacity_fault=0,receipt_capacity_fault=0,capacity_fault=0

FILM v0.7
t=6
pulser:er_clock:mode=periodic,p=2,phase=0,armed=0,done=0,nf=1
door:er_player:open=0,next_open=1
relay:er_a:cur_out=0,next_out=1
relay:er_b:cur_out=0,next_out=1
wire:w__er_a__er_b:cur=1,nxt=0
wire:w__er_b__er_player:cur=1,nxt=0
wire:w__er_clock__er_a:cur=1,nxt=0
admit:policy=admit_candidate_min_firstreceipt_v1,fact_capacity_fault=0,receipt_capacity_fault=0,capacity_fault=0

FILM v0.7
t=7
pulser:er_clock:mode=periodic,p=2,phase=0,armed=0,done=0,nf=2
door:er_player:open=1,next_open=0
relay:er_a:cur_out=1,next_out=0
relay:er_b:cur_out=1,next_out=0
wire:w__er_a__er_b:cur=0,nxt=1
wire:w__er_b__er_player:cur=0,nxt=1
wire:w__er_clock__er_a:cur=0,nxt=1
admit:policy=admit_candidate_min_firstreceipt_v1,fact_capacity_fault=0,receipt_capacity_fault=0,capacity_fault=0

FILM v0.7
t=8
pulser:er_clock:mode=periodic,p=2,phase=0,armed=0,done=0,nf=1
door:er_player:open=0,next_open=1
relay:er_a:cur_out=0,next_out=1
relay:er_b:cur_out=0,next_out=1
wire:w__er_a__er_b:cur=1,nxt=0
wire:w__er_b__er_player:cur=1,nxt=0
wire:w__er_clock__er_a:cur=1,nxt=0
admit:policy=admit_candidate_min_firstreceipt_v1,fact_capacity_fault=0,receipt_capacity_fault=0,capacity_fault=0

Composes with the 22 chapters before it

The chain through this chapter — every earlier fragment, this one, and the links — seals to sem-57d2e3599c2eac7a3db6ad8deb606dbc712c04a9cdacce4bc8071c69c4208215: 82 objects, 72 edges (was 78 / 69; every earlier object and edge is still present — checked, or the build refuses).

The board so far: one band per Part, signal flowing left to right; relays that fan out are routers, doors are switches, pulsers are clock domains. Hover an object — or click the board and walk it with the arrow keys — for its role, its Part and what it is wired to. This board is the chain’s sealed shape; no Film drives it, so it has no state to report, and the whole board in the conclusion is where every object’s state is read epoch by epoch. Wheel zooms · drag pans · double-click fits.

Syntax — quoted from the tree at build time

Law 4 (TRVM/LAWS.md) TRVM/LAWS.md:54

### Law 4 — CANONICAL (ratified 2026-07-22)
- **Statement:** *Every reported reduction cost must name the reduction strategy
  under which it was measured.* Eager/`first`, lazy-reference, or any later

Forces

Exact replay needs everything Shared Observable demands and a cost model that names its strategy, since a replay that takes a different reduction path is a different run. The number that would witness it is a receipt this tree does not yet hold.

Applicability

Verification of transitions, debugging distributed runs, any claim that a computation is deterministic.

Transformations

A refusing transformation is not one that is discouraged: it is one that, applied, makes the invariant above false. The word is the tree's, and it is the same word the join uses.

Consequences

Bugs in worlds become reproducible by construction. The honest status: STATED, with an explicit '?' — an acceptable answer under the doctrine, and the only honest one today.

Failure mode it answers

Replica Theater — Calling independent mutable copies 'the same thing'. Paid for at: CLAUDE.md: the served box-and-box copy is a MANUAL COPY and nothing syncs it

Witness

No witness. This pattern is STATED — the tree has no check for its invariant.

Counterexample

No counterexample shipped (required only when WITNESSED).

What to take away

1. from the animationTwo players, identical films — and a '?' where the fidelity number would be.
2. from the syntaxA cost without its reduction strategy is a rumour (Law 4); a replay without its strategy is not exact.
3. from the literatureDeterministic replay (rr), event sourcing, and the reproducible-builds movement.
4. from the witnessSTATED with '?': no receipt states a fidelity, and the page says so instead of inventing one.

Prior art — and what is not claimed

Novelty not claimed. Deterministic replay is a shipped, mature technique. TRVM's film replay is not verified by this build, and no receipt stating a fidelity percentage was ever located (v0.2 review, finding 7) — the page carries a '?' where that number would go.

Realizations in the tree

• TRVM/LAWS.md Law 4

Relations with other patterns

Cross-Machine Skill Replay STATED · Meaning Is a Hash WITNESSED