Give a story its beginning and ending. Five agents take it from there — each watching the others' work.
Give a character a start and an end — then watch five AI agents build the life between them. A Fact Generator invents the key events. You review and approve them. A Planner outlines the path, a Critic challenges it until the logic holds, a Writer turns it into prose. Add new facts later and an Explainer shows exactly where — and why — the story diverged. A live demonstration of goal-directed planning, human-in-the-loop approval, and agent feedback loops.
Start by giving your story a beginning and an ending. The Fact Generator creates life facts shaped by your settings — you control contradiction level and world type — creativity is derived automatically from the world type. You edit, reorder, or regenerate freely before approving. Once approved, those facts become the baseline for the original story.
The Planner builds a story outline. The Critic evaluates it and can reject it — forcing a replan until the logic holds. The Writer then turns the approved outline into full prose. This feedback loop is what makes the system genuinely agentic rather than a simple prompt chain. To generate a revised story, add more facts on top of the approved baseline — the Explainer will then show exactly where and why the story diverged.
Generates life facts shaped by contradiction level (low/medium/high) and world type (grounded/realistic/fantastical/outlandish). Creativity is derived automatically from the world type. Output is reviewed and optionally edited by a human before proceeding.
Reads the world model and produces a structured 5–7 milestone outline. On retry attempts, the Critic's rejection reason is injected so the Planner knows exactly what to fix.
Evaluates the Planner's outline for timeline plausibility, fact consistency, and logical gaps. Returns APPROVED or REJECTED with a specific reason. Max 3 attempts before falling back to last outline.
Takes the approved outline and generates full prose. Prose style is shaped by the world type — from plain literary fiction (grounded) to epic other-worldly narrative (outlandish).
Runs automatically after v2 is generated. Compares both stories, identifies what changed, and explains which new facts caused each divergence. Cost and time tracked independently.
Human-in-the-loop: After the Fact Generator runs, the human reviews the facts — they can edit, add, or regenerate. Only after explicit approval does the Planner→Critic→Writer pipeline proceed.
Real world, real rules.
Heightened but believable.
One impossible element.
Different universe entirely.