MIT GrandHacks 2025

"73% of patients forget to mention symptoms during visits. 795,000 Americans suffer death or permanent disability from diagnostic errors every year. 56% of those errors trace to poor patient history. We built Pulse."

Patients forget symptoms, suffer recency bias in 15-minute doctor visits, and pain scores are unreliable. Pulse lets patients capture symptoms in the moment — a quick gesture, a spoken description, logged with timestamp and wearable context — then generates a readable summary you can pull up on your phone in the waiting room.

The result: you walk into your doctor's office and say "I've had 10 episodes in 14 days, severity trending from 3 to 9, with new visual disturbances starting last week" instead of "yeah, some headaches."

Read the full story: STORY.md

• Miniconda or Anaconda
• Git

# 1. Clone the repo
git clone [email protected]:Sylindril/Pulse.git
cd Pulse

# 2. Create and activate the conda environment
conda create -n pulse python=3.12 nodejs -y
conda activate pulse

# 3. Install dependencies
npm install

# 4. Start the dev server
npm run dev -- --host

Open http://localhost:5173 in Chrome or Edge (needed for voice capture).

WSL users: localhost forwards automatically from Windows to WSL. Just open the URL in your Windows browser.

Voice capture note: The Web Speech API works in Chrome and Edge. Firefox/Safari will automatically fall back to text input mode.

• Long-press the pulsing circle for 1.2 seconds (or just tap "or tap to record")
• Speak your symptom — words appear in real-time as you talk
• Or tap "Type instead" to type it out
• Tap Done → rate severity (1-10 slider, skippable)
• If you've logged a similar symptom recently: tap how it compares to last time (Better / Same / Worse / Worst ever)
• Green checkmark — done!

• Tap "My timeline" on the capture page, or navigate to /doctor
• Summary button (top right) → readable narrative of your symptom history, grouped by body area, with trends and worst episodes highlighted. Read this in the waiting room before your visit.
• Stats panel → entry count, avg severity, most common area, trend direction
• Heart rate chart → 14-day wearable data with symptom markers overlaid
• Entry cards → each shows transcription, severity badge, comparison indicator, body area, and vitals (HR + SpO2) frozen at recording time
• Edit/delete → hover over any entry card to reveal pencil/trash icons
• Share → generates a unique read-only link for your doctor
• Reset → circular arrow icon reloads the demo dataset

• Hit Share on the timeline → get a unique link
• Your doctor opens it → sees your full timeline (read-only, no account needed)

The app comes preloaded with a 2-week patient story showing escalating headaches — the kind of pattern that gets missed in real visits but is obvious in the timeline view. Hit the reset button (circular arrow on the timeline page) to reload it anytime.

The dataset lives in src/data/demo-entries.ts. Each entry:

{
  id: 'demo-1',
  userId: 'demo-user',
  transcription: 'Your symptom description here',
  severity: 6,                     // 1-10 scale
  comparison: 'worse',             // 'better' | 'same' | 'worse' | 'worst' (optional)
  bodyArea: 'head',                // head, stomach, chest, back, neck, legs, arms, throat, general
  vitals: { heartRate: 88, bloodOxygen: 97.1 },
  recordedAt: day(5, 14),          // day(daysAgo, hour) — 5 days ago at 2pm
  createdAt: day(5, 14),
}

The day(daysAgo, hour) helper generates relative timestamps so the demo always looks fresh.

After editing, reset in the browser: click the reset button on /doctor, or clear pulse-symptom-entries from localStorage.

symptom-snap/
├── STORY.md                          # Why this matters (data, evidence, market)
├── CLAUDE_IDEAS.md                   # AI-generated feature ideas
├── HUMAN_IDEAS.md                    # Team ideas
├── FEATURES.md                       # MVP feature documentation
├── CLAUDE.md                         # Instructions for Claude Code
├── .env                              # Supabase keys (not committed)
└── src/
    ├── main.tsx                      # Entry point
    ├── App.tsx                       # Router
    ├── index.css                     # Tailwind + theme
    ├── lib/
    │   ├── speech.ts                 # Web Speech API wrapper
    │   ├── summary.ts                # Narrative summary generator
    │   ├── supabase.ts               # Supabase client (optional)
    │   └── mock-wearable.ts          # Wearable data utilities
    ├── hooks/
    │   ├── useGesture.ts             # Long-press detection + progress ring
    │   ├── useSpeechRecognition.ts   # Voice capture hook
    │   └── useSymptomEntries.ts      # Entry CRUD + localStorage
    ├── components/
    │   ├── capture/
    │   │   ├── CaptureOverlay.tsx    # Full-screen recording UI
    │   │   ├── VoiceVisualizer.tsx   # Pulsing waveform animation
    │   │   └── SeveritySlider.tsx    # Severity + comparison input
    │   └── timeline/
    │       ├── TimelineView.tsx      # Timeline container
    │       ├── TimelineEntry.tsx     # Individual symptom card
    │       ├── SummaryPanel.tsx      # Stats panel
    │       └── WearableStrip.tsx     # HR sparkline chart
    ├── pages/
    │   ├── PatientHome.tsx           # Capture page (/)
    │   ├── DoctorTimeline.tsx        # Timeline page (/doctor)
    │   └── SharedView.tsx            # Shared view (/shared/:token)
    └── data/
        ├── demo-entries.ts           # ← EDIT THIS for demo data
        └── demo-wearable.ts          # Auto-generated wearable trends

The app works fully offline with localStorage — no backend needed for the demo. To enable cross-device sync:

1. Create a project at supabase.com
2. Create a .env file in the project root:

   VITE_SUPABASE_URL=https://your-project.supabase.co
   VITE_SUPABASE_ANON_KEY=your-anon-key-here
   

conda activate pulse
npm run build    # outputs to dist/

Deploy dist/ to Vercel, Netlify, or any static host. HTTPS is required for the microphone and motion APIs to work on mobile.

This was built at MIT GrandHacks 2025. If you're a teammate:

1. Clone the repo and follow Quick Start above
2. Edit demo data in src/data/demo-entries.ts
3. Add your ideas to HUMAN_IDEAS.md
4. Read STORY.md for the pitch narrative and clinical evidence

MIT