MicroSpark AI

MicroSpark AI turns overwhelming assignments into one tiny next step at a time, helping neurodivergent students build momentum and do their own work without AI doing it for them.

Comment

Tagline

“I can’t start.” → “I can do this one thing.”

Inspiration

For many neurodivergent students, the hardest part of an assignment is not understanding it. It is starting.

A long rubric, open-ended prompt, or blank page can quickly become overwhelming. Most tools either add more pressure with timers and checklists or let AI do the work for the student.

MicroSpark AI takes a different approach: it helps students start without taking over.

What it does

MicroSpark turns overwhelming assignments into one tiny next step at a time.

Each step has a clear action and a clear finish line. Students can choose their energy level, and if they get stuck, Unstick Me adapts the current step based on the problem: too big, too vague, missing information, not knowing what to write, or lack of interest.

Most importantly, MicroSpark never writes the assignment for them.

The AI creates momentum. The student does the work.

How we built it

We built MicroSpark with Next.js, React, TypeScript, Cloudflare, and Google Gemini.

Gemini reads complex assignment instructions and turns them into small, structured actions while preserving the original requirements. We also added server-side validation so malformed AI output does not reach the learner.

The interface is intentionally calm and accessible, with keyboard support, reduced motion, no countdown timers, no autoplay sounds, and no tracking.

Challenges we ran into

The biggest challenge was making AI helpful without turning it into a homework generator.

We had to carefully design the prompts so the model helps students start, organize, and clarify instead of solving graded work.

Another challenge was keeping the product simple. It was easy to add more features, but that would defeat the purpose.

Our design rule became:

Show only what the learner needs right now.

What we learned

We learned that better educational AI does not always mean generating more.

Sometimes, one clear action is more useful than a perfect 20-step plan.

We also learned that accessibility is not just about colors, fonts, and keyboard navigation. It is also about how much information is shown, when it appears, and how much control the learner keeps.

Accomplishments that we're proud of

We built:

  • One-step-at-a-time scaffolding
  • Energy-adaptive actions
  • Five Unstick Me strategies
  • Clear completion signals
  • Academic-integrity safeguards
  • Privacy-first, accessible design

Our favorite transformation is still the simplest:

“I can’t start.” → “I can do this one thing.”

What's next for MicroSpark AI

Next, we want to add photo uploads, text-to-speech, multilingual support, and optional learner-support summaries.

Our goal is not to build an AI that does more schoolwork.

It is to help more students believe they can start.

Built With

Share this project:

Updates

Submission history