How we teach
StorySqueeze is not a chatbot with a composition prompt. It is a structured, research-backed tutoring workflow — designed so a child can run it independently, and a parent can trust what is happening inside it.
Four core principles
🎯Zone of Proximal Development
Every piece of AI feedback stretches your child just one step beyond where they are now. Not too easy to be useless, not so hard it discourages. This is the same principle behind the best human tutoring — we just make it available on demand, at scale.
🧩Bite-Sized Over Marathon
A full composition is broken into Scenes — 1 to 2 paragraphs each. Each scene is a small, winnable challenge. Small wins build momentum and make writing feel achievable on a school bus.
🎨Show, Don't Tell
The single biggest differentiator between AL3 and AL1 is replacing flat statements ("I was scared") with vivid action, sensory detail, and inner thought. Our AI coach relentlessly targets this — never ghostwriting, just showing what "one level up" looks like.
📱Learning Everywhere
Every workflow is designed to be completed on a phone, one-handed, in 10 minutes. Voice dictation for writing. OTP letter-boxes and handwriting canvas for spelling. No desk required.
How a session works
Storyboard
Your child plans one sentence per scene — the story arc before the story starts. Our AI checks the narrative structure: Is there a real turning point? Does the opening hook? Is there an emotional resolution? Problems caught here save 20 minutes of rewriting later.
Model Essay (background)
While your child plans, we silently generate an AL1 model essay aligned to their storyboard. We never show it to them directly — instead we extract 4–6 transferable phrases per scene that they can use as inspiration. The goal is to show technique, not give answers.
Scene Drafting
Your child speaks or types their draft for each scene. Our "clean" step fixes voice-dictation errors and typos without touching their ideas — their voice, their story, just legible.
AI Coaching
After each scene is submitted, the coach grades it (AL1–AL6) against the MOE rubric and returns a short, specific checklist for the next revision — plus 1 to 3 vocabulary items tied to what they were actually trying to say. The full story context is always in view, so the coach can assess whether a scene serves the narrative arc, not just whether it reads well in isolation.
Final Essay + ZPD Elevation
Once all scenes are done, the full essay is assembled. The "elevate" step rewrites it exactly one band up — same plot, same character, ≤6 changes — framed as "what your story could look like one step up". It is a model to study, not a shortcut to copy.
The AI is a tutor, not a ghostwriter
/clean
Fixes STT transcription errors and typos. Never rewrites your child's ideas — their voice, their structure, just legible.
/coach
Grades the paragraph, returns a forward-looking checklist (not a list of errors), and surfaces vocabulary tied to this specific scene. The student writes the next draft — the AI doesn't.
/elevate
Rewrites the finished essay one band up — capped at 6 changes, same plot, same character. It is explicitly framed as a model to study. Not a shortcut. Not the real grade.
Motivation layer
XP is awarded for completing stories, maintaining streaks, and — critically — for improving your grade from one draft to the next. It cannot be farmed by resubmitting the same draft. The system is designed to reward genuine effort, not volume.
Next
The exam & scoring
How compositions are marked and how we target each criterion →