Skip to content

DEVPOST.md

Latest commit

 

History

History
57 lines (43 loc) · 6.63 KB

DEVPOST.md

File metadata and controls

57 lines (43 loc) · 6.63 KB

Inspiration

95% of college students regularly procrastinate (American Psychological Association, 2021). We've all been there — staring at a massive assignment, feeling so overwhelmed that we don't even know where to start. The result? We put it off, pull an all-nighter the day before, and burn out. Existing to-do list apps and AI assistants can break tasks down, but they just hand you another overwhelming list. We wanted something different: a coach that walks you through one step at a time, understands why you're stuck, and actually holds you accountable with proof verification.

What it does

Doable is an AI-powered productivity coach that turns overwhelming tasks into bite-sized, doable steps — and makes sure you actually do them.

  • Tell it why you're stuck. Before breaking down your task, Doable asks why you're procrastinating — is it too big? Too boring? Scary? No idea where to start? The AI tailors its approach based on your specific blocker.
  • One step at a time. Instead of dumping a long to-do list on you, Doable guides you through one micro-task at a time with a progress bar, acting as a real coach rather than a task manager.
  • Proof verification. Before you can move to the next step, you upload a photo proving you completed it. Our AI uses image recognition to verify it's legitimate — no cheating allowed.
  • Smart reminders. Schedule reminders with quick presets or a custom time picker. Notifications are delivered via Web Push, so they work even when the app is closed.
  • Compete with friends. Connect with friends using unique codes, earn points for completing steps, and climb the leaderboard. A little healthy competition goes a long way.
  • Analytics & streaks. Track your productivity trends, maintain daily streaks, and visualise your progress over time.
  • Google Calendar integration. Don't want to manually input tasks? Pull them directly from your Google Calendar.
  • Spotify integration. Prefer to work with music? Play your playlist right inside the app.
  • Reward system. Complete a task and get celebrated with confetti, points, and a flip-through diary of all your proof photos showing how far you've come.

How we built it

  • Frontend: Next.js 16 with React 19, TypeScript, and Tailwind CSS 4 for a clean, responsive mobile-first UI. Framer Motion handles animations, and Recharts powers the analytics visualisations.
  • AI Engine: Anthropic's Claude API powers the intelligent task breakdown, adapting its coaching style based on the user's specific procrastination blocker. It also handles proof-of-completion image verification.
  • Backend: Next.js App Router API routes with MongoDB Atlas for persistent storage of tasks, user profiles, reminders, friend connections, and leaderboard data.
  • Authentication: NextAuth.js supporting both Google OAuth and email/password sign-up with bcrypt password hashing.
  • Notifications: Full Web Push implementation with VAPID authentication, server-side reminder scheduling, and a service worker for background push delivery.
  • PWA: Installable as a progressive web app on mobile and desktop with offline service worker caching.

Challenges we ran into

  • Data isolation between accounts. Early on, we discovered that when a second user logged in on the same browser, they could see the previous user's tasks because localStorage is browser-scoped, not user-scoped. We solved this by making the database the single source of truth and adding a user-scope check that clears local data on account switch.
  • Build-time crashes with Web Push. The web-push library's VAPID configuration was running at module load time during Next.js static analysis, crashing the build when environment variables weren't available. We had to refactor to lazy initialization inside request handlers.
  • Balancing coach vs. task list. We went through several iterations of the UX — our first version just showed a list of broken-down tasks, which felt no different from ChatGPT. We pivoted to a step-by-step guided flow with a progress bar that reveals one task at a time, which felt much more like having an actual coach.
  • Proof verification accuracy. Getting the AI to reliably distinguish between legitimate proof photos and random images required careful prompt engineering and iterating on the verification criteria.

Accomplishments that we're proud of

  • The app genuinely feels different from existing productivity tools — the combination of personalised blocker-based coaching, one-step-at-a-time guidance, and proof verification creates accountability that a regular to-do list can't match.
  • A polished, mobile-first UI with smooth animations and a cohesive design system that looks and feels like a real product, not a hackathon prototype.
  • Full-stack Web Push notifications that work even when the app is closed — a real technical achievement for a hackathon timeframe.
  • The competitive leaderboard and friend system adds a social layer that keeps users coming back.
  • Our hand-drawn rabbit mascot that reacts to different situations (burned out, scared, distracted, locked in) gives the app personality and makes the coaching feel warm rather than robotic.

What we learned

  • How Web Push notifications actually work end-to-end — VAPID keys, service workers, push subscriptions, and the entire lifecycle of a background notification.
  • The importance of making the database the single source of truth in a multi-device/multi-account world, rather than relying on client-side storage.
  • That UX matters more than features — showing one step at a time instead of a full list was a small change that completely transformed how the app feels to use.
  • How to build secure API routes from the ground up — input validation, auth checks, mass assignment prevention, and error message sanitisation.

What's next for Doable

  • Vercel Cron integration for reminders that fire reliably even when no users have the app open, ensuring no reminder is ever missed.
  • Group accountability — create study groups where members can see each other's progress in real time and cheer each other on.
  • Habit tracking — extend beyond one-off tasks to recurring habits, helping users build long-term productive routines.
  • App blocking integration — partner with focus apps to automatically block distracting apps while a user is working through their steps.
  • Richer analytics — deeper insights into procrastination patterns, peak productivity hours, and personalised tips based on historical data.
  • Offline mode — full offline support with background sync, so users can complete steps and upload proofs even without internet, with everything syncing when they reconnect.