Inspiration
When we first read the prompt, our team started by listing every kind of barrier we could think of. The one that kept coming up for all of us was access to information. Not information that doesn't exist, but information people can't use at the moment they need it.
We narrowed it down to one situation every one of us had lived through: being in a technical environment with your hands full. Swapping a part inside a laptop, working at a lab bench, or getting stuck in technical software like MATLAB. The help is out there in manuals, forums, and videos, but it never matches what's in front of you, and you can't scroll a tutorial while holding a screwdriver.
AI assistants are powerful, but today they live in a chat box. You have to stop working, type out your problem, and then translate a text answer back into the real world. That gap shuts out anyone without technical experience. It felt like exactly what the prompt was asking for: a system that a meaningful group of people struggle to use, and a chance to expand who can use it. Once we saw it, it was such a big problem for us that we had to build it.
What it does
Reason is a hands-free AI guide for technical tasks. A camera headband sees exactly what you see. You ask for help out loud, and Reason figures out what needs to be done, highlights the exact part on screen, and talks you through it step by step. No experience needed, and your hands stay on the task.
How we built it
We connected a head-mounted camera to an AI vision model that understands what the user is looking at. A language model turns the task into short, plain-language steps, and a visual overlay highlights the exact part to work on. Voice input and output make the whole experience hands-free, so the user never has to stop and type.
Challenges we ran into
The hardest part was moving AI from describing things to pointing at them. Getting the guide to highlight the exact part in a live camera view, not just talk about it, took many iterations. We also had to keep the guidance feeling live even though AI models take time to respond. And we kept catching ourselves using jargon, so we rewrote our instructions until every step sounded like a friend explaining it. With so many possible use cases, the biggest discipline was choosing one task and building it end to end.
What we learned
People don't need more information. They need it at the right moment, in the right place, in words they understand. We learned that accessibility isn't only about disability. Anyone becomes a beginner the moment they face an unfamiliar technical task, and the right guide can turn that beginner into someone who can do it themselves.
What's next
We want to take Reason beyond laptop repair into labs, makerspaces, and technical software like MATLAB, build a library of guided tasks for common devices, and make the headband lighter, cheaper, and available in more languages.
Log in or sign up for Devpost to join the conversation.