FOR PARENTS OF KIDS 4–12 • STAGES ALWAYS IN ORDER • NO PREP

Artstache: Why a Graph Beats a Search When the Order Matters

Parents don’t default to “that’s nice!” because they don’t care — they do it because after a 10-hour workday they are exhausted, holding a curling drawing at the kitchen counter. And they don’t know what else to say. Children’s drawing develops in stages, in order. A normal AI search can’t tell a 4-year-old’s tadpole person from a 10-year-old’s comic — the descriptions sound the same. Artstache stores Viktor Lowenfeld’s developmental stages (1947) as a chain in a database (Neo4j), so the order can’t be skipped, and gives the parent a short guide — the Ghost Guide — they can read in about a minute.

01 / DISCOVERY — THE 6:15 PM WINDOW

The Kitchen Counter Problem: Guilt, Clutter & Fatigue at 6:15 PM

Founder's Truth: I'm not a parent. I built Artstache for my sisters and friends — working parents who love their kids' art but after a 10-hour day default to “that's nice!” and feel guilty tossing the paper pile at midnight. I earned my design degree much later than most, and learning visual literacy helped me cut through noise. This is my contribution.

Child development literature is dense and academic. Parents want to guide creative growth but have zero time for lesson prep. Result: 25+ unsorted drawings piling up on counters, guilt-driven midnight purges, and passive iPad screen time.

What Success Looks Like

“A tired parent gets something meaningful and stage-accurate to say, in about a minute. Web-first, one-handed scan at the kitchen table. Out of scope: classroom dashboards, selling art supplies.”

Built for the Kitchen TableA Guide in About a MinuteNo Prep Needed
From Fridge Pile to Ghost Guide — 6:15 PM → 6:16 PM
One Minute, Start to Finish
✦
6:15 PM
1. The Pile
25+ unsorted drawings pile up on counters. Parent exhausted after 10-hour day.
◑
6:15:30 PM
2. The Trap
“What is it?” Well-meaning, but triggers defensiveness. Default: “that's nice!”.
✧
6:15:40 PM
3. Ghost Guide
Ask: “Where is the dino running to?” Guilt replaced by 10 minutes of actual conversation.
About a Minute • No Prep • Nothing to Look Up
Problem: Generic parenting advice requires 45 minutes of forum searching while dinner boils over.
Solution: a one-minute guide, written for this drawing, before the pasta boils over.
Viktor Lowenfeld (1947) // The Stages of Drawing, and the Tadpole Person
PHASE 02 OF 06
Milestone Spotlight: Stage 02
Preschematic (Ages 4–7) — The Tadpole Person
How a Young Child Sees It
What Adults Misread
A head with legs and no body. Adults assume something is missing, and respond with “Where is the tummy?”
What It Actually Means
The child draws what matters to them: the face talks, the legs run. The tummy does nothing they can feel, so it is left out. That is healthy, normal development.
Why a Simple Search Fails Here: It matches the words “stick figure” and suggests anatomy or perspective lessons meant for much older kids — years before a 4-year-old is ready.
Lowenfeld’s Stages, in Order — Stored as a Chain in the Database
01 Scribbling
2–4 yrs
02 Preschematic
4–7 yrs
03 Schematic
7–9 yrs
04 Realism
9–12 yrs
05–06 Decision
12–17 yrs
*The research: Viktor Lowenfeld (1947), Creative and Mental Growth — the same stages public school art curricula are built on.
Stages: 6
Age Band: 4–12 Years
Guide: About a Minute
Follow-Ups: 15-Min Activities
02 / THE STAGES — THE TADPOLE PERSON

Why “What Is It?” Backfires: The Tadpole Person

I went through public school art curricula. Most of them are already built on Viktor Lowenfeld’s 1947 developmental stages. The research was finished decades ago — what was missing was a way to use it in about a minute, at the kitchen table:

About a Minute
Reassurance before the pasta boils over. Anything that takes setup gets abandoned; parents need something short with no prep.
A Bridge, Not a Babysitter
Five seconds on the phone to spark 10 minutes of looking at each other. The screen starts the conversation; it doesn’t replace it.
Don't Skip Stages
Never suggest a lesson meant for a 10-year-old to a 4-year-old. Development goes in order — which is why the stages live in a graph database (Neo4j), not in a similarity search.
The Rule Behind It

“A tadpole person is not a mistake — it is how a young child’s mind works. Artstache’s job is to translate the drawing, so a parent can celebrate what is emerging instead of correcting the anatomy.”

03 / HOW IT WORKS — WHY A GRAPH, NOT A SEARCH

Why a Graph Beats a Search When Order Matters

A normal AI search matches by similar words. A 4-year-old’s tadpole person and an older child’s stick-figure comic get described with the same words — “figure, limbs, face” — so the search mixes up the stages and can recommend a perspective lesson years before a child is ready for it.

The Difference
Similarity Search
Matches words. “Stick figure” ≈ “stick figure”, so it treats them as the same — whatever the child’s age.
Artstache (Neo4j Graph)
Matches the chain instead: each stage points to the stage that comes next, so the order can’t be skipped.

In child development, order is everything. You can’t teach perspective before a child is drawing a baseline — the stages have to come in sequence.

The Graph Keeps the Stages in Order
Neo4j Aura • Lookup ~90ms (My Machine)
INPUT
Kid Art Scan
Canvas Memory
The photo is shrunk under 80KB in the browser. Its location data (EXIF GPS) is stripped before upload. No faces are stored.
→
GROUNDING
Neo4j Lookup
Stage-Order Tests
Graph lookup ~90ms on my machine. If no stage matches, nothing is shown. No badge is better than a fake badge.
→
OUTPUT
Ghost Guide
One Tap
A question to ask, ready in seconds — and a guide that takes about a minute to read, backed by the stage research.
The Safety Check:is_grounded_in_neo4j

If the lookup finds no matching stage, or the AI’s confidence is under 0.75, the app stops and shows no milestone badge. The AI can only write from the exact stages the lookup returned — it can’t invent a stage.

Guardrail Rule:The research sets the order. The AI only writes the words. No badge is better than a fake badge.
artstache.app • Split-Pane Studio • Archival Canvas
Paper Elevation Hierarchy
Level 2: Cards#FFFFFF
Level 1: Paper Low#FAF2EC
Level 0: Archival Canvas#FDFBF7
Ergonomic Layout
A panel that collapses (288px → 80px): Camera, archive, and activities stay in thumb reach — one-handed, during dinner chaos.
The Studio // Ghost Guide
Clay Dino: “Where Is He Running To?”
15 MIN • PRINTABLE
!
Preschematic Milestone (Ages 4–7)
She is focused on the story, not the anatomy — that is imagination emerging. Ask: “Where is he running to?”
Typography
Georgia (reflection) + Manrope (bedtime)
Edge Treatment
Deckle edges (255px 15px 225px 15px)
04 / CRAFT — THE NO-LINE RULE

The No-Line Rule: From Clinical to Archival

The Psychological Challenge
How do you show a parent real developmental analysis without making it feel like a medical report card that says “your child is behind”?
Early tests with thin gray borders and cold white cards felt like a doctor’s portal. I removed the harsh lines and replaced them with soft layers of paper (`#FDFBF7` canvas, `#FAF2EC` paper, `#FFFFFF` cards) and warm shadows.
// The 5 steps, in order (LangGraph)
step[1]: Privacy check — strips location data, checks the photo against the stages
step[2]: Vision — Gemini 2.5 Flash reads the drawing’s shapes
step[3]: Guide writer — checks the stage order, writes the question to ask
step[4]: Activity picker — matches past activities, picks a 15-minute follow-up
step[5]: Voice pass — rewrites it warm and reassuring
05 / TESTING — WHAT I CAN PROVE

What’s Proven, and What Isn’t

My rule: No fake user counts. I only report what I can check in the code and the tests. The biggest risk in this product is an AI suggesting a stage out of order — so that is what the tests hammer.

1
All 50 Stage-Order Tests Pass
Every test drawing’s stages came out in the right order. No activity is ever suggested unless the stage order checks out first.
2
The Test That Proves the Check Works
I deleted the link between stages 2 and 3 in the database on purpose: the tests caught 12 out-of-order results (38/50). I put the link back: 50/50 again. A check you’ve never seen fail isn’t a check.
3
The Vision Step Can’t Keep Extra Data
The AI that reads the drawing can only return what’s in a fixed list (skeletal_mapping.json). Anything extra is dropped, and no face data is kept.
4
Privacy Handled in the Browser
The photo’s location data is stripped in the browser, before upload, and the photo is shrunk under 80KB. Built with children’s privacy rules (COPPA) in mind — not certified compliant.
The Guardrail (milestone_guard.ts)
50/50 Tests Pass • Break-Tested
Foundation: Neo4j database • Lowenfeld’s stages (1947) • Graph lookup ~90ms on my machine
The Guardrail, in CodeThe research sets the order, the AI writes the words
1. The Order Check
(:Stage)-[:PRECEDES]->(:Stage)
Each stage must point to the next — checked in the database
2. Not Sure? Show Nothing
Confidence < 0.75 → null
No badge is better than a fake badge
3. AI Stays in Its Lane
5-Step Workflow
It can only write from stages that checked out
✦See the Guardrail Code (milestone_guard.ts)▼
// The research sets the order; the AI only writes the words
const result = await neo4jDriver.executeRead(async (tx) => {
  return tx.run(`
    MATCH (current:Stage {id: $stageId})-[:PRECEDES]->(next:Stage)
    RETURN current, next
  `, { stageId });
});

// If the check fails, show nothing — no badge is better than a fake badge
if (!result.records.length || confidence < 0.75) return null;

// The 5-step workflow only writes from stages that checked out
const guide = await langGraphDAG.invoke({
  stage: result.records[0].get('current'),
  next: result.records[0].get('next'),
  drawingPrimitives: visionFeatures
});
Tested: 50 of 50 stage-order tests passSee the Full Test Setup in /lab →
How Fast It Runs (Measured on My Machine)
Graph lookup: ~90ms typical across 50 runs (the slowest 10% took up to ~660ms; raw log: latency_runs.jsonl). A full scan: ~800ms typical. The free database sleeps after 72 hours idle, so the first scan after that takes ~30 seconds unless the keep-alive is running. No production monitoring yet.
The Privacy Approach
The photo is processed in the browser (HTML5 Canvas): location data is stripped before upload, and the image is shrunk to 1024x768, under 80KB. No face or biometric data is stored. Built with children’s privacy rules (COPPA) in mind — not certified compliant, and I don’t claim otherwise.
Looking Back — What It Cost and What I Learned
Live • No Fake User Counts
50/50
Stage-Order Tests
Break-tested: deleting a stage link on purpose made 12 tests fail. Restoring it brought back 50/50.
About a Minute
From Photo to Something Real to Say
A parent goes from “that’s nice!” to a specific question about the drawing, with no prep. The scan itself takes under a second.
0
Families Enrolled — Yet
It’s a live pilot and recruiting now. Until families use it, I’m not claiming results for kids.
What Building This Alone Taught Me
  • Graphs beat searches when order matters: A search matches a 4-year-old’s tadpole person to a 10-year-old’s comic because the words are similar. Development is a sequence — storing the stages as a chain enforces the real order, not a similar one.
  • I over-built the agent team: 11 named agents felt clever, but 2 of the steps do most of the work. If I’d had a PM, they would have cut the rest on day 2. Don’t name a step unless you can justify it to a tired parent.
  • Shame is the real competitor: Not other ed-tech apps. A parent feeling judged by a clinical UI and defaulting to “that's nice!”. The No-Line Rule was a direct response to shame.
  • What I'd do differently: Move the vision step onto the phone itself (Wasm/ONNX), so scans work in basements and cabins with no signal.
Where It Stands
  • Pilot status, honestly: The app is live and the stage-order checks are tested. I’m recruiting 10–15 pilot families (kids ages 3–8) for real kitchen-table trials — 0 enrolled yet, so no claims about results for kids.
  • Still writing: The short follow-up activities — 15-minute activities across all 6 stages, about 15–20 milestones in total.
06 / LOOKING BACK — WHAT’S REAL AND WHAT’S NEXT

Shipped, Verified, and Honest About What's Next

I built Artstache for real kitchen tables. There are no user counts to report yet — no families have enrolled. What I can report: the stages live in a graph that keeps them in order, the tests prove the order holds even when I try to break it, and the design feels like a family archive instead of a medical chart.

The Design
A warm paper look instead of a clinical report: no harsh lines, three soft layers of paper, two typefaces (one for reflection, one for bedtime), big thumb-friendly targets, and the guide in a pull-up drawer.
The Build
Neo4j holds Lowenfeld’s stages (1947) as a chain, so the order can’t be skipped. A 5-step LangGraph workflow writes the guide from those stages only. Order tests pass 50/50, including the break test. The photo’s location data is stripped in the browser before upload.