Personal Project · 2026

Designing for Neurodiversity

My role: Product Manager, Designer, and Researcher

6: Build-measure-learn cycles, solo as PM, designer, and researcher
2: Daily testers: my own neurodivergent kids, not a panel

Tools

FigmaLovableClaudeCursorUser Research

Teams

Personal ProjectEnd Users (Family)

Bottom line

My role: Product Manager, Designer, and Researcher. I owned the product loop end to end: problem definition, success metrics, scope, accessibility-first UX, user research, and a shipped live prototype.

The problem: EdTech apps optimized for attention economy tactics created the opposite outcome for neurodivergent learners: overstimulation, rigid flows, and feedback that eroded motivation.

Defined North Star metrics around healthy engagement, as a PM would, before any screen existed: time-on-task, completion rate, and zero mid-task drop-offs.
Scoped and shipped the smallest testable loop first, then let six rounds of real use with my own kids decide what got built next.
Validated against daily use with my two neurodivergent kids, not a claim about a broader population or a different market.

Challenge

Job to be done: a neurodivergent child needs to start a learning session and actually finish it without shutting down, so a session that starts is a session that gets completed, not abandoned mid-task.

Mainstream EdTech relies on gamification, pop-ups, and visual overload, patterns that create overstimulation and drop-off for neurodivergent children with ADHD.

Rigid task flows and harsh error feedback erode psychological safety, making it harder to sustain attention and complete learning sessions.

Approach

Defined success like a PM before touching a screen: healthy engagement metrics (time-on-task, task completion, zero mid-task drop-offs), not feature count.

Scoped like a PM too: picked the smallest task loop that could prove the engagement hypothesis, one reading exercise end-to-end, instead of building a full curriculum first, so real use with my kids could tell me what to build next.

V1, started in the mother tongue: Hebrew interface — V1 · Hebrew

V2, the interface becomes the lesson: English, loud gamified palette — V2 · English, default gamification

Overrode the AI builder's default gamification (loud colors, XP bars, stacking badges) with a coin economy simple enough to explain in one sentence, and a calmer palette.

V3 home screen, calmer palette and coin economy — V3 · Home

V4 quest-style home screen — V4 · Quest home

Reading passage list, each with coin reward and level — V3 · Reading

Listening lessons list with en-US audio — V3 · Listening

Speaking practice: pronunciation, dialogue, songs — V3 · Speaking

Vocabulary games grouped by type — V3 · Games

Achievements and class competition ranked without names — V3 · Class board

Built transparent navigation and in-line accessibility (tap-to-translate, text-to-speech) so no one has to leave the app to get unstuck.

V4 comfort settings: Minimal Mode and Calm Mode toggles — V4 · Comfort settings

V6 home screen, safe and quietly bilingual — V6 · Home

V6 non-diagnostic personalize-your-learning quiz — V6 · Personalize your learning

Shipped a non-diagnostic 'personalize your learning' quiz and direct comfort toggles, grounded in Cognitive Load Theory and ADHD executive-function research.

Impact

No usage-at-scale data. The validation was hands-on: my own neurodivergent kids as daily testers from day one.

As PM, I decided what got built each cycle and why; as designer, I built it: set the metric, ship the smallest testable version, watch it fail in a specific way, fix that one thing. Six times.

This stays a personal project: built for my own two kids and, by extension, for other children their age carrying the same ADHD-driven limitations, overstimulation, rigid flows that break focus, harsh feedback that triggers shutdown. Not positioned as a proven pattern for any other market.

What I'd do differently

Retrospective

This never left daily use with my own two kids. Before I'd call the coin economy or the engagement metrics validated for a general neurodivergent population, it needs testing with families outside mine.
Six iteration cycles were fast because I was the PM, designer, researcher, and the parent of the only two users, in a team setting that loop would need a lighter-weight way to keep everyone that close to the failure signal.