HopperFocus

AI-powered productivity tool that beats ADHD time blindness using visual timers, loss aversion mechanics, and voice narration. Break down assignments into 5-15 minute micro-tasks with real stakes.

Comment

In case youtube is having issues this it the direct link to the video presentation + demo:

https://www.youtube.com/watch?v=gvWHvEnOm9s

https://docs.google.com/presentation/d/1va-BaTlpHCgCi5liM0kkJ2xle0vNVWx7VkfbsIWzgJU/

Inspiration

Traditional productivity tools are built for neurotypical brains. They rely on abstract calendars and distant deadlines, which completely ignore a core symptom of ADHD: time blindness. To an ADHD brain, a project due in two weeks feels exactly the same as one due in two years. It doesn't exist until the panic sets in the night before.

Seeing brilliant students completely shut down because an assignment felt too overwhelming made the mission clear. We needed a tool that didn't just list tasks, but visually and psychologically enforced the passage of time. Named after Grace Hopper, HopperFocus turns the ADHD trait of "hyper-focus" from a random occurrence into a reliable tool.

What it does

HopperFocus is a gamified productivity dashboard that cures task paralysis by breaking massive assignments into high-stakes, 5-minute micro-quests.

  • AI Task Breakdown: Paste a terrifying 10-page syllabus into the app. The Gemini API acts as an "Odds Maker," instantly slicing it into 10 to 20 actionable micro-tasks.
  • The Wizard's Wager: To trigger immediate motivation (loss aversion), users must stake their virtual "Mana" currency to start a task.
  • Hyper-Focus Mode: The screen clears all distractions, displaying only a visual countdown timer. As time drains, the UI pulses from calm blue to urgent red.
  • Audio Body-Doubling: Using the ElevenLabs API, a digital "Familiar" speaks to the user, announcing task starts and time warnings to keep them anchored.
  • Instant Dopamine: Complete the task before the timer hits zero to win back your stake plus a bounty. Fail, and the Mana is burned.

How we built it

We architected a robust, async full-stack application to handle the AI parsing and real-time timers:

  • Frontend: React 18, Vite, and TypeScript, styled with Tailwind CSS for a highly accessible, custom "magical" theme.
  • Backend: Python and FastAPI for high-performance, non-blocking routing.
  • Database: MongoDB Atlas (via Motor) to act as the ledger for user stats and Mana balances.
  • AI & Audio: Google Gemini 2.5 Flash for the core parsing engine, and the ElevenLabs REST API for high-fidelity voice synthesis.

Challenges we ran into

  • Taming the AI Output: Early on, Gemini would return conversational paragraphs instead of usable data. We had to heavily engineer the prompt and lock the model into a strict JSON-mode response using Pydantic schemas to ensure the React frontend wouldn't crash when parsing the tasks.
  • The Double-Mount Audio Bug: React 18's Strict Mode caused our ElevenLabs audio triggers to fire twice, resulting in overlapping voices. We solved this by implementing useRef hooks to meticulously track the playback state.
  • Timer Drift: Relying on standard setInterval for the Hyper-Focus countdown caused the timer to desync from real time. We had to refactor the logic to calculate remaining time based on absolute UNIX timestamps.

Accomplishments that we're proud of

We didn't just build a flashy to-do list; we successfully translated peer-reviewed cognitive psychology (loss aversion and executive dysfunction) into a working software architecture in under 24 hours. Getting Gemini, FastAPI, MongoDB, and ElevenLabs to seamlessly pass data back and forth without dropping a single task or lagging the UI was a massive full-stack victory for the team.

What we learned

Prompt engineering is truly an art form. We learned that to get consistent data for a UI, you must be ruthlessly explicit with your AI constraints. Furthermore, we realized that accessibility isn't just about screen readers; true cognitive accessibility means giving the user total autonomy over how they view and interact with their workload.

What's next for HopperFocus

Our immediate next step is building out the social accountability layer. We plan to introduce a "Study Buddy" system where friends can link their accounts, wager Mana in cooperative focus sessions, and share their AI-generated assignment breakdowns with classmates.

Built With

  • and-typescript
  • custom-"magical"-theme.-backend:-python-and-fastapi-for-high-performance
  • elevenlabs
  • fastapi
  • google-gemini
  • mongodb
  • motor
  • non-blocking-routing.-database:-mongodb-atlas-(via-motor)-to-act-as-the-ledger-for-user-stats-and-mana-balances.-ai-&-audio:-google-gemini-2.5-flash-for-the-core-parsing-engine
  • python
  • react
  • styled-with-tailwind-css-for-a-highly-accessible
  • tailwindcss
  • typescript
  • vite
Share this project:

Updates