Your World Model
A generic chatbot starts every conversation from zero. Your board does not — it keeps a structured, private dossier of your world and reads it before any expert says a word. This page explains exactly what is remembered, how it gets there, and how you stay in control.
What the board remembers#
One structured dossier per user — not a transcript archive.
Your World Model is a single structured record, not a pile of chat logs. It holds:
- Identity — who you are, your role, and the register you prefer to be spoken to in.
- Company — name, industry, stage, business model, team size.
- Hard numbers — the metrics you have shared (revenue, churn, runway, price points) so advice is quantified, not generic.
- Goals and constraints — what you are driving at and what is genuinely off the table.
- Preferences — how you like recommendations framed and delivered.
- Open questions — the things the board knows it still does not know, so it asks when they become decision-relevant instead of guessing.
Every grounded session renders this dossier into a tight, bounded context block for each seated expert — so session #10 on a topic is structurally smarter than session #1.
How it gets written#
Seeded once, then updated by write-back after every grounded session.
- Seeded from your profile. When you first describe yourself or your company, that context bootstraps the dossier — no extra setup step.
- Updated from intake. When a board’s facilitator asks you clarifying questions, the answers merge into the dossier so no expert asks you the same thing twice.
- Updated from the memo. When a board delivers its call, what it learned about your situation writes back — the verdict, the numbers surfaced, the constraints that shaped it.
Grounded vs. Off the record#
Memory is a per-session choice, and the default is visible.
Grounded (default)
The experts read your World Model before answering, and what the session learns writes back. This is what makes the board compound — every call sharpens the next one.
Off the record
Nothing is read, nothing is stored. The session runs from your question alone and is marked with an “Off the record” chip in the session and on its dashboard card.
Over the API and MCP there is a third scope for automation: wm_scope accepts grounded, read_only (the board reads your dossier but writes nothing back — useful for experiments and batch runs), and off_record.
You stay in control#
The dossier is yours to read, correct, and prune.
- You can view everything the board holds about you.
- Edits stick: correct a fact or delete a goal and the board respects it — a removed item does not silently resurface on the next write-back.
- Memory is strictly per-user. It is never shared across accounts, never visible to other users, and never used to train models.
Where you meet it#
The World Model is not a settings page you have to maintain — it shows up in the work.
- Experts cite your real numbers and constraints instead of asking for them again.
- Intake gets shorter over time — the facilitator only asks what the dossier does not already answer.
- When you reconvene a decision, the new round opens already knowing the context the last round established.