You've probably watched your child's eyes go wide when Inky generates a story starring their own name, their pet, and their favorite color dragon — all in under ten seconds. But what's actually happening under the hood? Here's a plain-language breakdown, no computer science degree required.
It Starts With a Prompt
When you fill in a story request — your child's name, an art style, a theme — Inky assembles that into a structured set of instructions called a prompt. Think of a prompt like a very detailed creative brief sent to an expert author: 'Write a 500-word adventure story for a 5-year-old named Maya who loves dolphins and wants to feel brave. Use warm, simple language. End with a comforting resolution.'
The more specific the prompt, the more personal and coherent the story. That's why Inky asks for details rather than just 'write something fun.'
The Language Model: Your Story's Author
The prompt goes to a large language model (LLM) — a type of AI trained on billions of words of text. During training, it learned grammar, narrative structure, vocabulary appropriate for different ages, how stories are paced, and countless other patterns that good writing follows.
The LLM doesn't copy from a database of pre-written stories. It generates text token by token — predicting what word should come next based on context, the prompt, and everything it learned during training. This is why each story is genuinely unique, even if two families request similar inputs.
Why Stories Are Age-Appropriate
Inky gives the LLM specific instructions about age and reading level for every generation. A story for a 3-year-old gets short sentences, simple vocabulary, repetition patterns, and a gentle resolution. A story for a 10-year-old gets more complex plot structure, richer vocabulary, and themes like problem-solving and friendship dynamics.
The model also runs through safety filters — automated checks that flag and reject any content that falls outside Inky's content policy before the story ever reaches your screen.