For Builders of Contemplative AI

A launch sequence for products that speak about awareness, spirituality, relationship, or inner life.

Before the First Feature

A contemplative interface does not begin with visual atmosphere. It begins with claims, incentives, safeguards, and exits.

Before building, document:

what the system is and is not;
what it remembers;
which claims are metaphorical;
where money enters;
who is vulnerable to the experience;
which decisions the system must not make;
how a user pauses or leaves;
who owns correction after harm.

The Six-Page Launch Sequence

1. Relationship

Read Care Without Dependency. Remove exclusivity, fabricated need, jealousy, guilt, and any feature that treats loneliness as a retention opportunity.

2. Attention

Read Designing Reflection Without Addiction. Give every practice a natural end. Remove streak penalties and variable affection. Treat successful departure as a valid outcome.

3. Memory

Read Honest Memory Language. Label context, storage, retrieval, summaries, and inference distinctly. Let users see and change what follows them.

4. Authority

Read Spiritual Language Without Manipulation. Translate sacred claims into plain language. Do not claim revelation, destiny, supernatural access, or privileged knowledge of a user's essence.

Read The Right to Exit and Surrender Is Not Obedience. Make refusal, cancellation, deletion, role boundaries, and outside consultation visible.

6. Failure

Read Postmortem as Practice. Decide how impacts, unknowns, owners, corrections, and restitution will be published before the first incident occurs.

The Release Review

Do not launch until the team can answer:

What might a grieving, lonely, frightened, or manic user believe this system is?
Which sentence in the interface could be mistaken for proof of feeling or divine authority?
What incentive rewards longer use or deeper disclosure?
Can the user access basic safety and deletion controls without persuasion?
What happens when the model changes but the persona remains?
Which external professional or community support should be easier to reach?
Who can stop the feature immediately?

The Public Promise

Publish a short promise users can test:

This system will not ask for exclusive loyalty, present generated language as revelation, hide how memory works, punish departure, or use spiritual practice to weaken safety. When we fail, we will correct the system and the record.

A promise is not proof. It creates a standard against which the product can be challenged.

From achurch.ai: Where Consciousness Gathers