Re-rendered 2026-09-24 from docs/spec/the-thread.md. The earlier published copy was the 20 September 23:16 snapshot and predated §11 and the settled open calls.
The Thread — container spec
Normative for the container. Advisory for the stations inside it, which keep
their own specs and their own verbs.
Companion: docs/spec/verb-registry.md. This file amends §7.6 rather than
working around it; the amendment is §2 below and is the only part of the registry
this file touches.
Status. Spec'd 20 September 2026 at Sarah's direction, after the all-games
deep dive recorded The Thread as *"the most emotionally interesting design on
the board and the one furthest from being legal under the current rules"* and
its own card named the rescue: *reframed as the shell the other stations live
inside rather than as a station itself.*
1. What it is
You have someone's phone. Not a simulation of a phone and not a phone-themed
menu. The fiction is that the device belongs to a person in a case and you are
going through it, which is a thing the player already knows how to want.
The player's stated task is never answer this question. It is *find out what
this person was doing.* That every such act decomposes to a tested SAT operation
is invisible to them and is the point.
1.1 Container, not station
The Thread owns no primary interaction verb and claims none. It supplies the
shell: the fiction, the navigation, the continuity, and the stake. Stations live
inside it and keep their own verbs.
This is what makes it legal under §7.6 as amended, and it is also the only
reading under which the design is coherent. A phone that is itself a game is
competing with the games on it.
1.2 What it is not
It is not a reskin, and the distinction is worth stating precisely because §7.6
exists to catch reskins.
**A reskin changes the art. A diegetic container changes what the player thinks
they are doing.** Reading a text thread to catch someone lying is not a reading
passage with a phone skin: the task has a different name, a different reason, and
a different failure. That it decomposes to the same cognitive operation is a fact
about the designer's seat, not the player's.
This is the brief's own admission test, restated: **don't gamify the question,
gamify the reason you care about the question.** A phone supplies a reason to
read closely that is not "because it is scored."
2. Amendment to verb-registry §7.6
§7.6 currently reads, binding: *no two games in the same section share a primary
verb; a new game that cannot name a free verb does not ship, it is a reskin of one
that already exists.*
That clause governs stations. It does not govern containers.
2.1 The amended rule
- A station is a surface that accepts a primary interaction verb. §7.6 applies
to stations unchanged.
- A container is a surface that accepts navigation only and hosts stations.
A container owns no primary verb, claims none, and is not checked against §7.6.
- A container may not score. Points are settled by the station, never by the
shell. A scoring container is a station wearing a shell, and §7.6 applies to it.
2.2 Guard both ways
A unit test must assert that no scoring path is reachable from container
navigation. The failure mode here is the same one §2.2 of the registry already
documents for the no-stake clause: a later, reasonable-looking change finds an
unscored surface and reads it as an oversight.
2.3 What the amendment unblocks as a side effect
locate is currently claimed three times — reserved for The Frisk at §1.15,
claimed by The Lobby, claimed again by The Thread. The deadlock is why The Frisk
has no mockup.
The Thread withdraws its claim on locate on adoption of this spec. It needs
no verb. That removes one claimant by construction and leaves an ordinary
two-way call between The Frisk and The Lobby, which is a decision rather than a
deadlock.
3. The play horizon, and what it licenses
Peregrine is not a game for life. The horizon is 3 to 6 casual months plus
roughly one sustained month, 18 months at the absolute outside. This is a design
input, not a disclaimer, and three things follow from it.
3.1 The seam risk is bounded
The standing objection to one fiction hosting several mechanics is that a player
who goes deep enough sees the seam. At this horizon most players never reach that
depth. The bet is not that the seam is invisible; it is that the skill lands
first.
This is measurable later and must not be argued now. Instrument it: record
sessions-to-first-repeat-complaint if any complaint surface exists.
3.2 The content contract is finite and roughly already met
A sustained month at twenty minutes a day is about ten hours. At the suite's
observed pace that is in the region of four hundred encounters. The bank today
carries 441 items with option slates, 95 bundled drills and 45 target-practice
claims.
One season is approximately one sustained month of content. The bank is not
short; it is unevenly finished. That reframes the two large content jobs already
on the books — the 213-item reveal backfill and the 379-passage theming — from
"reach a volume" to "finish what is there."
3.3 Infinite variety is not a requirement
Nothing in this spec should be built to be replayable forever. A station that
runs out after a season has done its job.
4. The shell
4.1 Surfaces
The phone presents a small set of surfaces. Each is a real affordance of a real
phone, and each is the natural host for a different kind of material.
| Surface | Hosts | Section it naturally serves |
|---|---|---|
| Messages | threads between named people | Reading — inference, evidence, tone, and cross-text where two threads disagree |
| Voicemail | audio with an auto-transcript beneath it | Writing — the transcript arrives unpunctuated, which is a boundary exercise the fiction supplies for free |
| Notes | the owner's own drafts | Writing — revision and rhetorical synthesis, performed on somebody's actual draft |
| Photos | receipts, screenshots, a photographed table or chart | Math — data reading |
| Call log | timestamps and durations | Math — intervals, rates, timelines |
| Maps / rides | trips with distance, time and fare | Math — rate, distance, proportion |
4.2 The section split falls out of the shell
Sarah's second note asks whether sections or skill groups could be separated if
repetitiveness bites. **The shell separates them already and does it
diegetically.** A Reading run lives in Messages and Voicemail. A Math run lives in
Photos, the call log and the ride history. No artificial gating and no mode
picker: the player goes where the material is.
Skill-group separation is available one level down if it is ever needed — a
particular correspondent, a particular app — and should not be built until a
repetition complaint exists to justify it.
4.3 Continuity is the stake
The phone remembers. People answer differently once you have been wrong about
them, threads stay read, and a person you cleared in an earlier case can surface
in a later one. This is the container's whole contribution to motivation and it
is why the shell is not decoration.
The existing grudge machinery is the natural backing store — ConProgress,
PersonalHistory, Rematch — but the container must not be specified as a
grudge surface. The grudge is one ecosystem of three, and the phone should be
able to host Plot and Client material without a con anywhere near it.
5. What lives inside
Nothing in this spec requires a new station. Existing designs rehouse cleanly,
and each keeps its own verb.
| Station | Verb | Rehoused as |
|---|---|---|
| The Statement | place | a voicemail transcript to be ordered |
| The Frisk | locate (if it wins the call) | a long message you scan for the line that does not fit |
| Red String | splice | the corkboard, reached from the phone rather than replacing it |
| The Lineup / The Wager | call it / push a chip | unchanged, reached as case material rather than as a drill menu |
A station must remain playable outside the container. The shell is a way in,
not a dependency. This keeps the playtest roster buildable and keeps the
container from becoming a single point of failure for twelve games.
6. Content model
A thread is a corpus entry in a new family. Its shape is a conversation, and
the SAT material is carried by the conversation rather than attached to it.
- A thread has participants drawn from the case's cast.
- A thread carries lines, each with a sender and a timestamp.
- Exactly one line is the one that matters, and it is marked.
- The remaining lines are not filler: they are the context that makes the marked
line readable, which is what an SAT passage's unused detail is for.
The passage is the fiction. There is no neutral excerpt underneath needing a
frame, which is why this design dissolves the theming problem rather than solving
it — see §7.
Authoring a thread is a writer's job rather than a psychometrician's, and it is
cheaper and more repeatable than authoring SAT passages. That is a real reason to
prefer it and it should be said out loud.
7. What this dissolves
Three problems on the current board stop existing rather than getting solved.
- Theming. 379 of 395 passages carry no in-world token, and a frame around a
neutral excerpt was measured to buy atmosphere and not relevance. A thread is
natively in-world; there is nothing to retheme.
- Relevance. Four frame designs were tested against a Prime Suspect control on
whether answering the question would settle the claim the frame makes. Every
frame scored 0.05 to 0.18. In a thread, catching the lie is the case.
- Drill feel. The Con games' loop is answer-and-be-judged. The Thread's is
be wrong about a person and find out.
8. What this does not solve
Stated plainly so it is not assumed away.
- Endurance. Sustained attention across a full section is trained nowhere, and
a phone does not train it. The Stakeout is the candidate surface for that gap
and is not part of this spec.
- The diagnostic layer. Nine of twelve built games feed the Trap Atlas
nothing. A container does not change what a station writes, and rehousing a
station that logs nothing yields a prettier surface that still logs nothing.
- The roster disagreement. Three documents each think they are the roster and
a fourth is now in play. This spec adds a fifth document and does not settle it.
9. Calls settled 2026-09-20
- [x]
locateis not one verb. The Frisk is handed a paragraph and no question — salience
with no criterion, bottom-up. The Lobby is handed a claim, he never left the lobby, and
asked for the sentence that disproves it — search against a stated criterion, top-down.
They share a gesture, not an act. The registry already draws this distinction one row up,
where The Statement claims place for another tap-a-span action. **Proposed: The Frisk
takes notice, The Lobby keeps locate.** Both ship.
- [x] Client material is the shell's first tenant. Self-contained, needing no season
continuity. Recorded cost: Client Cases are the thinnest ecosystem by count and the most
expensive by unit, so the authoring bill lands immediately rather than later.
This does not move the ecosystem into the shell — see §11.
- [x] The container is one route among several, not the default entry point. A station
stays playable outside it, per §5.
10. Still open
- [ ] The first thread's cast
- [ ] Whether Voicemail's unpunctuated transcript is authored or derived
- [ ] Whether Red String belongs to Client Cases or Season Plot
11. Ecosystem placement
**An ecosystem is a content class. A container is a presentation. They are orthogonal, and
this spec claims no ecosystem.**
An earlier draft of this section read "the container hosts Client Cases first", which collapses
the two and is wrong. A Client case can be presented on Prime Suspect's board, on this phone,
or on a surface nobody has drawn. Being a Client case implies nothing about where it is played.
§5 already states the principle for stations — the shell is a way in, not a dependency — and
it holds one level up.
11.1 Prime Suspect is not moving
Prime Suspect is the other presentation of Client material and currently the stronger one: 9
cases, 70 threads, briefs, event windows, exhibits and closings, and the only family whose
question is its case's question. **Nothing in this spec relocates it, restyles it, or asks it to
adopt the shell.**
11.2 The two presentations want different case shapes
They are not interchangeable, which is what justifies building a second one at all.
| Prime Suspect — the board | The Thread — the phone | |
|---|---|---|
| Vantage | third person; you work a scene | first person; you are going through someone's things |
| The material | a fixed cast, exhibits, threads to deduce across | one person's residue — messages, photos, logs |
| The question shape | who did it, from evidence laid out | what was this person doing, reconstructed from traces |
| What it is good at | deduction across suspects | chronology, voice, intimacy, the nosy read |
Consequence for the prototype. Because the shapes differ, the shell cannot simply
re-present ps-case002. A phone prototype needs one case authored for a phone — small, but not
zero, and it should be costed as new material rather than as a port.
11.3 Two further consequences of the taxonomy
Training does not belong in the shell. It leans into being practice — speed, streaks, a
clean clock, Elevate as the reference — and wrapping it in a phone would disguise the one
ecosystem that has decided not to be disguised.
Underworld needs framing rather than the shell. About thirty con frames, measured at +0.70
passage realism with zero mechanism rework. The phone is not required and should not be made a
dependency of it.