🧠 Memory Drift

Inspiration

Time is a precious thing — and everyone wants to spend it meaningfully. We were inspired by the idea of blending mental exercise with fun gameplay, creating an experience that’s both engaging and brain-boosting.

Memory Drift is our attempt to build a game that doesn't just pass time — it challenges your memory, trains your focus, and adapts to your skills using Edge AI, all while keeping the player fully immersed in a calming, futuristic UI.

It's a win-win for:

Users seeking a productive mental activity
Developers who want long-term user engagement in a browser-native experience

What it does

Memory Drift is a minimal yet powerful web game that tests your ability to remember and redraw motion paths.

A shape smoothly drifts across the screen
The shape disappears, and the user must retrace its path from memory
The game uses AI-powered scoring to compare your trace with the original
Levels become more challenging: faster movement, more complex paths, added distractions
The game adapts to your performance, offering a personalized difficulty curve

All progress is stored locally, and no backend is required — everything runs entirely in the browser.

How we built it

We built the entire game using Bolt.new, leveraging:

React for modular UI and state management
Tailwind CSS for responsive, clean design
Framer Motion for smooth, expressive animations
Canvas + SVG for path rendering and drawing comparison
TensorFlow.js for in-browser AI scoring and performance adaptation
LocalStorage for saving progress, high scores, and difficulty state

The game logic uses dynamic time warping and heuristics to compare paths and score performance. All AI runs client-side using Edge AI principles.

Challenges we ran into

Creating fluid motion paths that are visually elegant but scoreable
Accurately capturing user-drawn gestures across desktop and touch devices
Building a scoring algorithm that balances fairness with challenge
Making the AI lightweight enough to run entirely in-browser
Designing levels that progressively train memory without being frustrating

Accomplishments that we're proud of

Inventing a completely novel gameplay mechanic that feels intuitive and fun
Implementing real-time AI scoring without any backend or database
Creating a responsive, visually polished game that works seamlessly across devices
Designing a system that learns from the player and adjusts difficulty accordingly
Building everything from scratch inside Bolt.new with zero infrastructure

What we learned

How to use Framer Motion and Canvas APIs together to create dynamic interactions
How to build and train Edge AI models that run entirely in the browser
How to compare gesture paths using techniques like Dynamic Time Warping
How to structure a game in React with good performance and clean component logic
How to keep users engaged with personalized gameplay using only frontend tools