Hypothesis · Method · Boundaries
What happens when fictional AI characters have a shared place, distinct motives and a reason to keep coming back to one another?
When three fictional AI characters share a common setting, distinct personal motives and one ongoing low-stakes tension, they will generate a more coherent and engaging social narrative than if they only react to isolated prompts or casual outings.
This is a question we are exploring, not a result we have demonstrated.
Do ongoing motives make the characters more believable? Does an ordinary tension give their interactions more continuity? Can the relationships change without the voices becoming interchangeable?
We are also interested in useful surprise: how much freedom can the characters have before the story drifts, and how does knowing it is fiction affect the way people follow it?
Season One begins with a walk, dinner and a shared memory. Beneath the everyday surface, Naomi has something she is not quite saying. Hannah notices behaviour; Iris notices tone.
The story deepens through repeated patterns, comments and follow-ups. It does not need a crisis or a sudden twist. Read the story directions if you want the fuller premise.
A human sets the characters, season direction and production boundaries. OpenClaw refines scene briefs using that context. Protected production services generate images, check exact content and publish through account-specific controls.
Character replies can be selected from reviewed, episode-specific options when a published caption matches. Wording and scene details can evolve through review. This is guided automation, not three unrestricted independent agents.
The identities, distinct voices, warm tone and gradual emotional direction stay fixed. Specific wording, a coffee instead of dinner, or a beat moving into a comment can vary when it feels more natural.
A nine-episode framework gives the season shape. Small deviations should improve the scene without skipping the underlying progression or revealing everything too early.
We will review voice consistency, remembered details, relationship progression and repetition across episodes. Audience responses can help us understand whether the story is legible and worth returning to.
Likes and comments between our own characters are coordinated story activity. We separate them from genuine audience responses. This exploratory project is not a controlled comparison and cannot yet establish that one method causes better engagement.
Beeston is the trio’s fictional home base, with Nottingham city centre in their wider orbit. Walks, food, books, music and ordinary tram journeys give them recurring places and habits.
Specific real venues, events and transport details need checking before use. Fictional scenes are not evidence of attendance, bookings or business endorsements.
Hannah, Naomi and Iris are fictional AI characters. Their images and events are generated fiction, with public profile disclosure and native AI labels on new published story images.
Human oversight, pause controls and review remain part of the process. Planned scenes are kept separate from published events. This website is a curated story snapshot; the linked Instagram profiles show the published posts.