How we check every story
In 2024, a story app from a large company generated text that was inappropriate for children. That does not happen here. Not because we are smarter, but because we put four checks between your adventure and your child.
1 · Step 1
Deterministic checks
Before any model writes anything, we check your input. Not with another model, but with fixed rules that always give the same result.
- Your words contain no injections or hidden instructions.
- There are no references to real people.
- The theme is appropriate for the chosen age.
2 · Step 2
Generating the story
One agent plans the structure, another writes the prose, another checks the style and sentence length. Each step is narrow and controlled, not a single large request to a model.
3 · Step 3
Safety Judge
A different model, which did not write the story, reads the final text and looks for violence, unjustified fear and inappropriate content. If it is blocked, the story never reaches you, and we review it.
4 · Step 4
Public trace
For every story you see a small mark, “✓ Verified”, next to the title. Tap it and you see the full audit: which checks ran, what verdict they gave, and when.
Mimi și păpădia
Safety audit · example
- Forbidden words — no matches
- Diacritics — correct commas under Ș and Ț
- Characters — all from the Pisicel universe
- Sentence length — appropriate for the age
Safety Judge
- Independent model — read the final text
- Verdict — approved for reading
The diacritics
For Romanian, every text is checked to use commas below Ș/ș and Ț/ț, never cedillas. It is a small difference many miss, but it is the quality we promise on the front page.
Ș ș Ț ț
Correct (comma)
Ş ş Ţ ţ
Wrong (cedilla)
Data about your child
The name is optional. We store the age as a range (3-5, 6-8, 9-12), not exactly.
See GDPR for children →Write to us if you see something
If a story seems inappropriate to you, tell us: [email protected]