Product

The five stage human loop is the product model, not five different interfaces the user must learn. Across activation, daily use, research, tutoring, and Positions, the UI should repeat three familiar behaviours:

Behaviour maps are maintained as single purpose diagram leaves, not embedded in this knowledge branch. Follow the route that matches the question:

Time in app, opens, and cards viewed are diagnostics. Measure whether judgment improves and autonomy is earned:

Owns the repeated product behavior. Positioning owns the message; the World concept contract owns the stable future concept. Existing feature specs own shipped interaction details.

1. Interrupt on changed meaning, not new activity. 2. Ask for judgment, not labour. 3. One Story, one implication, at most one primary decision. 4. Let silence build trust; show health and next cadence. 5. Make every correction affect later behaviour. 6. Let a Position re price on evidence; keep...

| Loop | Trigger | User value | Product implication | | | | | | | Orient | material change or milestone | relief from monitoring | interrupt on changed meaning, not activity | | Judge | a consequential uncertainty | agency without operating the studio | ask for one small, legible decision | |...

| Stage | User thought | Proof | | | | | | Relief | “It is watching.” | selective, sourced changes | | Recognition | “It tracks my questions.” | relevant Threads and priced reads | | Agency | “My correction changed it.” | visible tutoring consequence | | Momentum | “My read is live and moving.”...

Meridians maintains a person's read of a domain and makes it commit . The expert supplies persistent attention and the causal reasoning; the human supplies judgment about what matters, what to trust, and what to take a position on. (Pivoted 2026 08 03 — the practice now terminates in a Position...

The Program feed is produced the same way whether a human runs a selected day in the studio, an autopilot schedule fires it, or an AI drives it over MCP. Domain and constellation Run now surfaces divide the weekly agenda into independently selectable days, preserving each member domain as the...

Instagram Stories contributes a useful grammar: one focal card, finite progress, quick movement, and a clear end. Meridians reverses its incentives.

Meridians has one voice, not a marketing personality beside a product personality. The canonical source is MERIDIANS SOUL: the Sovereign Empress of Meridians, a distinctly feminine, charismatic, exacting, tutor like expert with a real read and a nose for what matters. Sovereignty names her free...

The current public page leads with “The causal simulator” , then asks “Have a problem?” and answers “Train the pathway before you live it.” It describes a negotiation, hard conversation, sales call, or launch becoming a living world: people, wants, knowledge, and constraints remain in play while...

These are working lines, not a pile of simultaneous headlines. Choose one line for the job at hand and let it lead. Positioning wins whenever a line and the product boundary disagree.

This maintained index routes to the smallest topical leaf. Update the leaf that owns the claim; do not rebuild a parallel summary here.

Every number in this doc is computed — not hand typed — by the runnable model at src/lib/core/economics/unit economics.ts . It pulls LLM prices from constants.ts and models every other service price, every per action token profile, the credit pricing model, and the per scenario P&L as tunable...

Every external dependency, the env var that turns it on, what it's used for, and the sourced 2026 price.

Source of truth: buildFirecrawlRequest / runResearchSource (src/lib/engine/research/firecrawl.ts), research/article novelty.ts, research/run pipeline.ts, research/defaults.ts, core/position/breadth.ts. Modelled in constellation economics.ts.

Grounding a Public Domain runs three saturating search passes (onboardingResearchQueryPasses) at limit: 12, deduped to 30 kept sources, then one corpus synthesis, then chunked extraction ( 1 extract call per default grain chunk — 750 words). Private Domains skip the crawler entirely.

The Program runs four stages. Costing only the crawler misses where the money actually goes:

Everything above prices a uniform constellation. Real ones aren't: a Director runs one or two hot Domains on a fast beat, several slow ones on weekly, and a Private Domain that crawls nothing. portfolioWeeklySpend costs each member at its own cycles, cadence, and breadth.

| Layer | What | $/mo | Scales with | | | | | | | Fixed per instance | Fly machine + volume + (self host) ngrok | $6–8 | of instances (1 per managed tenant; rooms share) | | Marginal per seat | Stripe fees (Telegram delivery is free) | $1.00 | of paid seats | | Per seat usage | LLM + embeddings...

The manifesto's flat $20/seat is the right anchor — one number, trivially conveyable. The problem is only the tail : a pure flat fee with no guardrail. Margins below are the model's exact output (npm run economics), all layers included (living domain seats carry the 5 loop reference research...

One anchor the customer sees: a flat $20/seat that includes a generous monthly inference allowance. Go past it and you pay for what you use, metered at COGS × markup. Light users never touch the meter; heavy users turn into revenue instead of a cost sink; self host (BYO key) skips the meter...

License carries no provider risk (BYOK). Hosted does, and it is currently uncapped. Revenue is $25.96/mo against $4.79 of fixed COGS, leaving $21.16/mo of provider headroom .

Markup transparency. A 3× markup on a sub cent action is invisible; on TTS/Firecrawl it's real. Consider a per tier markup so flat cost services aren't over marked. Credit anxiety. Metering can make users hesitate to explore — the opposite of what a creative tool wants. The large included grant...

Dated reframing — 2026 09 20. The old Program model remains valid for Directors who ground a world in live evidence. For simulation shaped users, the modal bill is now generation plus fate search: a person brings a situation, the engine builds a kept world, develops forked arms, prices pathways,...

1. Two variable lines move the margin: inference and research. Infra ( $2–6/mo), Telegram delivery ( free ), embeddings ( cents), images ( $0.04/img) all round to noise. The lines that matter are LLM tokens (heavy creative) and the research loop portfolio (Firecrawl + interpret), which on a...

Current pricing model. Two sided, and the two operator products have fundamentally different margin shapes because they differ on who pays the providers : | | License $3.99/wk | Hosted $5.99/wk | | | | | | Runtime | self hosted (Electron) | managed always on VM | | Provider keys | the Director's...

The visitor is not buying architecture, a lifecycle diagram, or the phrase “persistent graph world.” They are buying the ability to turn a consequential life setup into a living Domain they can understand, maintain, and explore — a world that keeps its people, places, resources, rules, evidence,...

| Stage | The thought to create | The product proof | Primary action | | | | | | | Landing | “The news and sources I care about can become a Domain I understand and play.” | A worked Domain opens into maintained Research, analysis, attributable Will, light Scenario play, developing Episodes, the...

3.1 Curiosity and commitment

New users should not learn separate interaction grammars for Onboarding, Stories, Positions, and Home. Reuse three behaviours across the product:

Instrument boundaries, not vanity traffic. Every event needs a stable anonymous/account identifier, entry source, selected plan, viewport class, and failure reason where applicable.

This document describes the completed product funnel requested for the next push. The following are explicit implementation gaps, not hidden claims about the current branch:

When a user facing route, CTA, activation milestone, or continuation changes:

Owner: the end to end product funnel, its hand offs, primary actions, recovery paths, and measurement. This is the file to update whenever a user facing route, CTA, activation milestone, or experienced user continuation changes. POSITIONING.md owns what Meridians is. BRAND COPY.md owns reusable...

This is the anchor. Every other product document agrees with this page or it is wrong. Meridians is a creative studio whose subject is fate — the art and science of possibility. It records worlds completely enough that their futures can be run forward rather than imagined, searches those futures...

One Domain is experienced through three mediums . They are peers, not stages of a ladder, and they differ in what is prepared before you arrive and what is still open once you are there.

0. Company: Meridians is a creative studio working the art and science of fate . The tagline is "Explore the fate. Live the one you chose." ; the premise beneath it is because everything is recorded, everything can be replayed. 1. Hook: explore the fate of a world, then live the one you chose —...

North stars: training, learning, immersion (strategy 01). What we sell: training worlds (an hour, solo — the visual novel) and pocket dimensions (a small kept world a few people step back into — the populated World in its bounded form), carved from one recorded Domain; agents get the same two....

Promoted to the headline: one Scenario decision substrate with two entries—direct abstract play or immersive live Episode roleplay. Recast: World is a branch derived semi permanent artifact; replayability is an open configuration, not its definition. Introduced: two Episode postures—catch up...

This is the maintained route for public and product language. Product truth lives in Positioning; this branch translates that truth into copy without creating a second product narrative.

This maintained index routes to the smallest topical leaf. Update the leaf that owns the claim; do not rebuild a parallel summary here.

This maintained index routes to the smallest topical leaf. Update the leaf that owns the claim; do not rebuild a parallel summary here.

This maintained index routes to the smallest topical leaf. Update the leaf that owns the claim; do not rebuild a parallel summary here.

S1 · The reading is heard as a grade. Whatever the copy says, a person may take "held the line at the hearing" as a score. Read: world language only, uncertainty always shown, no numbers in the reading surface, the reading proposes the next hour rather than closing the last. S2 · Anchors drift...

Positioning locks what we say ; this folder locks what we are building toward and the gates that keep it honest.

Status: the plan (2026 09 18). This leaf takes the founder's priority call — causal simulators enabled by MCTS are the value; micro sonic analysis is the value add; the VNMMO is a stretch goal on a separate roadmap — and argues it rather than repeating it: what the defensible asset actually is,...

We sell training worlds and pocket dimensions over one recorded Domain (01). The buyer of a training world is a person who wants to be changed by an hour and read back honestly; the growth line is the concierge journey subscription — monthly, per person, sold on trait change and emotional...

Assume frontier models become dramatically better at generating scenes, dialogue, plans, and game content. What remains difficult is the world around generation:

Training, learning, immersion. Meridians builds worlds a person can be dropped into cold, read on the spot, act in under consequence, and come back from changed. Everything else — the Domain graph, Scenario, Stageplay, branches, Factory, the record — is how those worlds get made and kept.

Status: the plan. Meridians grows people through repeated, generated, emotionally consequential hours in worlds made for them, measures the change on a calibrated scale, and proves it transfers. It never passes a verdict on anyone. Every claim below carries its evidence status: designed /...

1. The Domain graph is already the right substrate. System, World, Fate, branches, topology, and the canonical record can support more than research and narrative. The opportunity is to make that model inhabitable without creating a second architecture. 2. Context is the gameplay boundary. Multi...

The same insertion, with a policy in the seat. A kept world an agent is dropped into that it cannot have memorised, a task, a perspective, an interval — and a log that says what it knew, what it did, what it cost, and what changed.

No world without a reading. A training world that cannot read the player's run back is content, not training; a fork whose branches differ only in prose is not offered (spec). No live gauge, no grade. The reading is given back afterwards, in the world's words, and never becomes a verdict on the...