Three
A social app built around one constraint: say how you're feeling in exactly three words. The limit makes posting effortless and honest, and an LLM tags each post's emotion, so the app can show you, literally in color, how your friends are collectively doing. I designed the experience and ran the handoff to a ten-developer team. Three took first place at the end-of-semester demo.
Why another social app
01 / 03
Because the existing ones optimize for consumption, not honesty: Instagram is a highlight reel, BeReal is spontaneous but emotionally shallow. Across 10+ interviews and 50+ survey responses, people asked for the same three things: authentic interaction, a genuinely simple way to express themselves, and a way to actually know how their friends are doing.
Three words is the answer to all three at once. It's too short to perform in, quick enough to do daily, and specific enough to mean something.
The product
02 / 03A daily prompt asks for your three words. An LLM reads each post and labels its emotion; the post about progress gets tagged Hopeful without the author doing anything. Those labels feed the Mood Wheel, a rotating visualization of the collective mood of your circle. A FaceID-protected journal holds the longer entries, timestamped and searchable. And Fusion turns the emotion model into play: pick three posts to match a target emotion, get scored by the model, compete with friends.
Takeaways
03 / 03- Handoff at 10-developer scale
- Specs, documentation, and standing communication with the dev team so what shipped matched what was designed. The design was only half the job.
- Designing with AI
- The LLM had to enhance authenticity rather than intrude on it: emotion tagging works because it's ambient, not because it's a feature you operate.
- Making emotion visible
- Color, motion, and interaction can carry intangible things. The Mood Wheel is the project in miniature: a feeling, made legible at a glance.