Rewire

Rewire: Doomscrolling for good, a tool that’s designed to help you learn and reset your attention span

Comment

Inspiration

We live in a weird paradox: people want to learn, but the same apps that broke our attention spans are the ones we reach for when we're bored. Every "solution" asks you to cold turkey your way out, delete TikTok, do a dopamine detox or download a blocker. In reality, nobody wants to do this. We wanted to find a solution for doomscrolling that was actually feasible. The insight was simple. If chaos is what got people hooked, chaos is what you use to bring them back. You don't yank the rug on them; you slowly slide it.

What it does

We live in a weird paradox: people want to learn, but the same apps that broke our attention spans are the ones we reach for when we're bored. Every "solution" asks you to cold turkey your way out, delete TikTok, do a dopamine detox or download a blocker. In reality, nobody wants to do this. We wanted to find a solution for doomscrolling that was actually feasible. The insight was simple. If chaos is what got people hooked, chaos is what you use to bring them back. You don't yank the rug on them; you slowly slide it.

How we built it

Claude API generates the educational script for each video. We pass it the topic and a chaos level (1–5) and it returns a script in the right register, punchy and fragmented for level 1, calm and narrative for level 5, along with approximate word timings for caption sync. ElevenLabs API converts the script to voiceover audio. We use an energetic voice for early chaos-level videos and a measured, calm voice for later ones. Generation takes ~3 seconds per clip. Pre-downloaded background clips (Subway Surfers, Minecraft parkour, Minecraft build, nature footage, plain background) loop silently as tags while the audio plays on top. The videos are hosted on Cloudinary for easy streaming access and high-quality connections

The ui design was done in Figma make, from conceptualizing our idea to create our v1 of our app. Figma was the go-to tool to take our ideas into visual representations and seamlessly collaborate as a team.

Challenges we ran into

We faced a wide range of challenges these past 36 hours. From API’s not connecting and features not working as intended to figuring out how to fine-tune and calibrate an algorithm that would work for users How fast should the chaos drop? Too fast and users notice and feel tricked. Too slow, and the arc doesn't land in a single session. We iterated on the curve several times: it's now a gentle exponential decay across videos, with the steepest drop On the technical side, getting caption timing to feel natural without a dedicated sync library took more work than expected. Claude's word-timing estimates are approximations, and ElevenLabs' pacing varies slightly by voice. We ended up building a lightweight correction layer that stretches or compresses caption display windows to match actual audio duration. The other challenge was prompt engineering the chaos gradient. A "level 1" script needs to feel genuinely unhinged while still being accurate. A hallucinating, chaotic script is worse than useless. We spent a lot of time on the system prompt to keep educational integrity at every level of chaos.

Overall, the last 36 hours taught us how to take an idea to a product, from navigating challenges to building something we were truly proud of

What we learned

Overall, the last 36 hours taught us how to take an idea to a product, from navigating challenges to building something we were truly proud of. It taught us how to think on the fly, how to debug issues and keep innovating and improving the product we are working in. We were able to learn how to collaborate together, especially as a team with minimal hackathon experience and composed of people who met 30 minutes before the hackathon. This project showed us how anyone can come together to build something that you can end up being proud of.

What's next for Rewire

What's next is to continue innovating, building out a curated for you page of not only content that you prompt, building out more customization in voice settings and options, and improving the algorithm to make the process of aligning your attention more seamless

Built With

Share this project:

Updates