Hoppy

Stay consistent with your health, without the guilt: see the healthier you take shape, one ordinary day at a time.

Comment

Inspiration

Most people over 40 in India don't struggle to start getting healthy. They struggle to keep going. Diet apps feel like exams: count everything, get a red warning when you slip, give up by week three. They're also built for salads and smoothies, not dal, roti and chai. I wanted something that feels like encouragement instead of judgement, and that makes slow progress visible, because the real changes take weeks to show in a mirror.

What it does

Hoppy turns your selfie into a 3D health twin that lives your day with you. You snap your plate, say "do roti aur dal khaya", or type it, and Hoppy understands Indian home food in katoris and rotis. Your meals, water, movement and sleep fill six simple daily stats. Strong days, sustained over weeks, move your twin up through body stages, and slipping up only pauses progress, never takes it away.

Around that, Hoppy helps you stay consistent:

  • a water tumbler that fills
  • streaks and badges
  • reminders for meals, water and medicines
  • a weekly leaderboard with friends that ranks consistency, not kilos
  • a food-sticker history of every day
  • consults with dietitians and trainers
  • a clean report you can show your doctor

How we built it

I built Hoppy with Claude Code, plan-first: a scope, a product spec and a technical spec before any code, then the build in small verified steps. The app is a Next.js 16 PWA on Vercel with Supabase (Postgres, auth, storage, scheduled reminders). The health targets come from research rather than guesses:

  • calories: the Mifflin–St Jeor formula
  • protein and fibre: ICMR-NIN 2020 for Indians
  • activity: WHO 2020 and MET values from the Compendium of Physical Activities
  • water: EFSA guidance

OpenAI's gpt-image-2.5 turns a selfie into the twin, using one style reference so every user's twin shares the same look. A fast local parser understands Hinglish logs, so most entries need no AI call at all.

Challenges we ran into

The design took several tries. The first version felt childish, and game ideas like a Pokémon GO-style world and an endless runner were fun but wrong for adults over 40. We landed on the health twin, then rebuilt every screen to match mockups I made. Getting the twin right was its own challenge: it had to look like you while matching one consistent 3D style, and stay the same person across every pose. We also had to move to new image models mid-build when OpenAI announced it was retiring the old ones, and cut twin images from 1.9 MB to under 60 KB so they load fast on phones.

Accomplishments that we're proud of

  • A selfie becomes a recognisable 3D twin in about 45 seconds, live.
  • Logging that matches how Indians actually eat and talk: "pachis minute walk aur do glass paani piya" logs a workout and two glasses of water in one go.
  • A tone that never shames: no red "over budget" warnings, and stages that never go backwards.
  • Every target is grounded in published guidance, with 83 automated tests behind the rules.

What we learned

Planning before building saved us: every pivot only changed the look, while the research-based rules and data held up. I also learned that for this audience, how something feels matters as much as what it does. The same feature can feel motivating or patronising depending on tone. And on cost: making images only when a screepend to cents per user.

What's next for Hoppy

  • Test it with real people over 40 for a few weeks, and learn what actually keeps them coming back
  • Add Hindi and regional-language voice, and more regional dishes.
  • Connect the expert consults to real dietitians.
  • Sync steps and sleep from phones and watches.
  • Make it an installable app on iPhone too.

Built With

Share this project:

Updates

Submission history