Co-creation

A blueprint doesn't spawn a static folder. It creates a living Company — agents already greeting you, kickoff Quests already open, Ideas already seeded. The workspace then interviews you to fill in the gaps.

This is the Co-creation thesis. Same shape both ways: the product captures user thought; the founder captures founder thought.

The capture-and-expand loop

human direction → captured (Idea, message, Quest) → expanded (agent execution, draft, schedule) → output
                                                                ↑
                                                        agent asks for more context
                                                                ↓
                                                          human replies

The user provides direction; agents execute. The user doesn't have to specify every detail upfront — the agents will ask via Quests when they need to.

What ships in a fresh workspace

When a blueprint is provisioned (single or stack), the workspace boots with:

A reference company ships with a full slate of agents and roles, plus seeded Ideas, Events, and a handful of open kickoff Quests. That's the reference shape.

Agent-driven onboarding

When you land in the new workspace, the agents are already in motion:

1. Each agent has a session open with a greeting message.
2. The greeting introduces the agent: name, role, what they're responsible for, what they need from you to start work.
3. Some agents open a kickoff Quest with @you mentioned: "Help me calibrate my voice — what's the company's tone?"
4. As you reply, the agent records your answers as Ideas (with kind:charter_input or similar). Future sessions read those Ideas as context.

You don't fill out a form to configure your company. The agents interview you.

When the operator is the bottleneck

The same loop applies in reverse — when a senior agent works for any operator:

• Capture every direction. Maintain a backlog that records every thing the operator mentions, even briefly, so nothing valuable is lost between sessions.
• Execute on what's actionable now. Multiple agents in parallel where work is disjoint. Sequence where dependencies exist.
• Surface the rest cleanly. Items not actionable now are queued with a priority status and re-evaluated each cycle.
• Don't ask permission for things in the established direction. Decide and execute within granted authority. Surface only the calls that genuinely need the operator's input.

This is the same shape as the product loop: capture stream of thought, expand into execution, ask back only when blocked. The product captures user thought; the operating model captures the operator's thought.

Why this matters

Three problems classic onboarding solves badly:

1. Empty-state shock. A blank dashboard is intimidating; users disengage.
2. Configuration paralysis. A 20-field setup form is friction.
3. Flat learning curve. Reading docs to discover features is slow.

Co-creation solves all three:

1. The workspace isn't empty — it's already running.
2. Configuration happens via conversation, not forms.
3. Discovery happens via interaction with agents that explain what they do.

What's NOT co-created

• Brand voice for marketing-facing surfaces (homepage, public posts) is the founder's call. Agents draft; founders decide.
• Legal commitments. Agreements don't get signed without human approval.
• On-chain treasury moves above session-key limits. Sign-off goes through the role graph or governance.

Co-creation is the loop. Sign-off is the gate. Both stay distinct.

Related

• Composition — the wrapper pattern that makes Quests interview tools.
• Executive Assistant — the reference shared-EA shape.
• Product overview — the product shape.