Superwavy

Inspiration

Twitter began as SMS status sharing - a simple way to know what friends were up to. It's now often more focused on sharing opinions, for better and for worse. We spotted a fun opportunity to revive the old Twitter vibe using Chrome's new on-device AI capabilities, by filtering tweets in the feed in real time, either only status or only opinions, or any other vibe-based filter. Also we added some other metric based filtering to find underrated tweets and people.

What it does

SuperWavy filters Twitter feeds using Chrome's semantic AI APIs to surface authentic status updates over opinions, or vice versa. It features instant Japanese/Spanish translation and custom filtering powered by on-device models.

How we built it

Frontend: Tailwind CSS, React + Storybook for scalable UI AI Integration: Chrome's Prompt API for semantic analysis Translation: Chrome's Translation API for real-time processing Architecture: Designed for cross-platform expansion (Twitter, Threads, BlueSky)

Challenges we ran into

Current prompt API response times (2-3s) required creative optimization. Solved by embracing constraints and focusing on high-precision filtering over recall.

Accomplishments that we're proud of

Built responsive semantic filtering using experimental Chrome AI APIs Achieved near-instant translation for Japanese developer content Created extensible architecture for future platform integration

What we learned

On-device AI models offer compelling tradeoffs between latency and capability. The constraints of frontend processing pushed us toward more elegant solutions. Constraints are important!

What's next for Superwavy

Open-source v1 with extensible filter system Expansion to additional platforms Community-driven "vibe packs" for custom filtering Integration with Threads, Bluesky and Mastodon