Personas

Companion to positioning. This folder describes who brings a consequential situation to Meridians, what they hope to rehearse, and where the current product stops. It is not a market-size claim and it is not permission to describe direction as shipped.

The thesis

Rehearse consequence before you live it — and find out what you caused.

Earth is causally thin. Most people get one attempt at a hard conversation, a strategic move, a training decision, a story turn, or a game-world commitment. Meridians holds the situation, forks it, prices pathways, and lets the person remain the author of the move. It does not predict or choose. It interviews a person, turns the kept situation into a Domain, develops branches, and shows consequences through a visual-novel reading.

The fit test

Every fit must satisfy all four conditions:

1. They face a moment, decision, or world whose consequences they cannot cheaply rehearse.
2. They think in actors, incentives, information, and consequences, not headlines.
3. They want to remain the author of the move; the engine proposes and prices but never chooses.
4. They will return to ask “what did I cause?” and try again.

Someone who wants a forecast, verdict, passive story, or one-off answer is not a fit merely because they like the premise.

The primary roadmap is Causal MCTS VNs; VNMMO is a stretch goal on a separate roadmap.

Roster by use

What they are asking for

Prospects, not fans

The useful prospect has a situation with stakes and can name the actors, incentives, private information, constraints, and candidate moves. They are not buying a feed or an oracle. They are buying a bounded opportunity to act, inspect a consequence, and try again.

The interview is currently a desktop flow run by hand. Voice intake is direction, not a shipped hands-free product. The conversion question is whether the person returns to replay the situation, not whether they compliment the metaphor.

Two shipped realities

1. The record and branch substrate. Domains, canonical state, Scenario runs, and typed actions can preserve a declared world and its consequences.
2. The prepared reading surface. Published Experiences are generated once and then served. The reader does not pay generation cost for every reading.

The VN reader, automated MWS, latent-trait readings, transfer, and freeform World wandering remain direction unless a surface says otherwise. Noor’s reading, if built, must be private, non-clinical, non-diagnostic, and never a verdict. Transfer is unproven.

The causal loop

act → world remembers → “what did I cause?” → retry as experiment

Play without the return question is engagement, not learning. A persona earns a place here only when the loop can be observed or when the document clearly marks the missing piece.

Working boundaries

• Macro is the mastered layer: the engine can hold world state, branches, open Fate, and causal consequence at the scene/arc level.
• Micro is not mastered: nothing currently reads inside a scene at screen or sentence grain. Micro analysis is direction and exists in service of making macro consequence felt.
• Situation → world is manual: voice interview is direction; conversion is performed by an operator through MCP.
• Search is bounded: Scenario Lab experiments are evidence of a laboratory surface, not a shipped autonomous MWS service.
• Economics counts generated simulations: hosted headroom now hangs on generation plus MWS; voice minutes and automated search compute remain open estimates.

Generated economics

The tables below are projections from PERSONA_SETUPS and centralized action pricing. They are not promises about willingness to pay. Run npm run gen:personas after changing a setup. Marked blocks are machine-owned.

Program spend

<!-- gen:personas:spend --> <!-- /gen:personas:spend -->

Running cost

<!-- gen:personas:cost --> <!-- /gen:personas:cost -->

Shipped-PMF ranking

<!-- gen:personas:pmf --> <!-- /gen:personas:pmf -->

The PMF ranking measures fit to the shipped surface under explicit assumptions. It is not a verdict on a person and does not make a directional use look shipped.

Why each entry exists

“Help me prepare for X”

Sam is the flagship case: a salary conversation, a hard talk with a co-founder, or another one-shot moment whose ordinary rehearsal is too thin. The simulation is valuable only if it preserves the person’s move and makes the consequence legible.

“Tell me what will happen”

This is the oracle request, and it is the wrong promise. Meridians searches and prices pathways from a declared world. It cannot know the world completely and must not turn a priced pathway into prophecy.

“Make me a story”

Priya is the Reader: she wants an explorable VN and a pocket-dimension return, but the product must still ask what she caused. A passive viewer is not the target.

“Run my scenario for the team”

Rafael, Dev, Renata, and Maya need a table: named actors, asymmetric information, candidate moves, and attribution. The current Scenario Lab can run bounded experiments; multi-seat governance is still open.

“Check my novel’s fates”

Elena wants a chess engine for the fates of her own novel. The current author surface can preserve and develop branches; automated MWS remains direction.

Pivot note — 2026-09-20

The roster is now organised around causal simulation services rather than maintained experts. A person brings a situation, the engine helps keep a world, and the person chooses the commitment. Directors may ground a world in live evidence, but the modal simulation bill is generation plus MWS, not a crawl alone. The old maintained-knowledge roster remains useful history; it is no longer the thesis.

Further reading

• Anti-personas
• Causal simulator strategy
• Delivery and worldview
• Unit economics

Update discipline

PERSONA_SETUPS in constellation-economics.ts is the economic source of truth. Persona files explain the use, objections, and boundary. Generated blocks must be regenerated, not hand-edited.