ParaPal
What inspired us
We asked a simple question: what happens when someone with ALS wants to turn off their fan at midnight and no one is home? They wait. Every night. 10 million people worldwide are fully conscious and cognitively intact but completely paralyzed. They own smart devices. They have Google Home sitting in their rooms. And they cannot use any of it — because every interface ever built assumes you have a voice or a hand. That gap is what ParaPal is built to close.
What we learned
We learned that the hard problem isn't the technology — it's the design. The constraint of a single binary signal forces you to think completely differently about interface design. Every extra step costs real time and real effort for this user. We also learned that prediction is everything. Without intelligent letter ordering, free-form typing through a scanning interface takes minutes. That realization changed how we thought about the whole product.
How we built our prototype
We built a working auto-scanning web interface where the screen cycles through options automatically and the user selects with a single keypress — simulating what a neural signal would do in the real product. The interface has two pages: a home screen with preset actions, and a query builder with a full letter grid modeled after the Hector Salamanca communication board. We implemented bigram-based character reordering so the most probable next letter always appears first. For the signal input we used the spacebar as a stand-in for the actual EEG or EMG hardware, with a WebSocket layer already built to accept real hardware signals when connected.
Challenges we faced
Designing for one input was the core challenge. Every UI decision looks different when the user has exactly one button. Navigation, error recovery, speed — all of it had to be rethought from scratch. We also struggled with scanning speed: too fast and users miss their target, too slow and the interface becomes frustrating. Finding the right default and making it configurable was more nuanced than we expected. Finally we had to resist the temptation to build the full product and focus on proving the core interaction model works first.
Log in or sign up for Devpost to join the conversation.