Here’s a fair question from any parent looking at an AI storytelling app: if an AI writes my child’s story, how does it know, three weeks in, that the dragon’s name is Miro, that he’s afraid of thunder, and that he still owes the hedgehog a favor from chapter four? Is it listening? Recording? Remembering everything?
The honest starting point is the opposite of what most people expect: the AI doesn’t remember anything on its own.
The blank-slate problem
The language models that write stories start every session from scratch. Left alone, an AI asked for “the next chapter” would cheerfully rename the dragon, forget the hedgehog exists, and contradict last week’s plot without noticing. This isn’t a flaw we work around reluctantly; it’s just how the technology works, and any app claiming its AI “simply remembers” is glossing over the machinery.
So the memory has to be built deliberately, outside the AI. Ours is built the way television has solved this exact problem for decades.
The story bible
Long-running shows stay consistent across years and changing writers because of a document called the story bible: who every character is, what has happened, which threads are still open. No writer relies on memory. They rely on the bible.
Lullawick keeps one per saga. After each chapter, the new facts get written down in a structured way: characters (names, quirks, that fear of thunder), facts about the world, open plot threads (that favor owed to the hedgehog), and a short record of what happened. It’s notes, not prose or recordings, closer to index cards than to a transcript.
What happens each night
When it’s time for a new chapter, the relevant parts of the bible are handed to the AI along with the request, so the writer starts the night already knowing the world. After the chapter is written, a second step reads it back against the bible, checking for contradictions the way a continuity editor checks a script: did anyone’s name drift, did a settled fact change, did a closed thread mysteriously reopen? A chapter that fails the check gets fixed before your child ever hears it. Then the bible is updated with what’s new, and the cycle repeats tomorrow.
That’s the whole trick. The AI never remembers. The bible never forgets.
What’s actually stored about your child
Less than you might guess, because the bible is about the story. It knows a great deal about Miro the dragon and very little about your family: a first name, an age (chapters are written and sized to it), and the preferences you’ve chosen, like themes you’d love woven in. There are no recordings, no conversations, no microphone. The richest thing Lullawick accumulates about your child is a story they love, and that’s by design.
Where it can still slip
Honesty requires this section. Continuity checking catches most drift, not all of it; a minor character’s scarf might change color, a small detail might wobble between chapters. When something slips, it’s a fix in the bible, and the story carries on. We’d rather tell you that plainly than promise a flawless memory no system actually has.
The result, on a good night, is the thing that makes this whole approach worth it: your child asks what happens next, and the story actually knows. See how it works.