Inspiration
75% of high school students report feeling constantly stressed. 1 in 3 college students struggles with anxiety or depression. 366 million adults worldwide live with ADHD — losing productivity not because of lack of ability, but lack of support.
We realized something: Software helps you work faster, but not think better.
Most tools react after mistakes happen — after you send the angry message, after you burn out, after you lose focus.
We wanted to build something that helps in the moment that matters.
That’s how Luke was born.
What it does
Luke is an AI-powered desktop companion that lives on your screen.
He monitors your cognitive load in real time, detects emotional signals, and gently intervenes before mistakes happen — without disrupting your workflow.
He’s not an app you open. He’s always there. Quiet, aware, and helpful.
Key Features Live Cognitive Load Score (0–100) Updates every 10 seconds using behavior signals like app switching, activity, and break patterns Chat Coach (Real-time Intervention) Detects emotionally charged messages from your clipboard and pauses you before you hit send, guiding a quick reset Voice Assistant ("Hey Luke") Hands-free interaction using voice input/output — designed for accessibility and focus Animated Desktop Presence A lightweight, always-on companion that feels alive without being distracting How we built it
Luke is built as a real-time, local-first desktop system:
Frontend & Desktop
Electron + React + Vite Tailwind CSS + Framer Motion
State & Storage
Zustand SQLite (local-only for privacy)
AI Stack
Claude (reasoning + conversation) OpenAI Whisper (voice input) ElevenLabs (voice output) face-api.js (facial emotion detection, local) Sentiment.js (tone analysis)
We focused heavily on low latency and privacy-first architecture, ensuring Luke feels instant and trustworthy.
Challenges we ran into Intervening without being annoying We had to design a system that steps in only when necessary — not constantly interrupting the user Real-time signal processing Combining multiple inputs (emotion, activity, behavior) into a meaningful cognitive score Latency in voice interaction Making the voice pipeline feel natural and responsive Privacy concerns Ensuring sensitive user data stays local while still delivering intelligent insights Accomplishments that we're proud of Built a fully functional desktop AI companion, not just a chatbot Created a real-time cognitive load scoring system Designed a 3-second intervention system that prevents impulsive actions Achieved a privacy-first, local processing approach Focused on underserved users — students, ADHD, and mental health support What we learned Timing is more important than intelligence — when you intervene matters most Users value non-intrusive support over aggressive automation Emotional awareness is a missing layer in modern computing Local-first systems significantly increase user trust What’s next for Luke
We’re expanding Luke into a full cognitive support system:
Facial emotion detection (real-time, local) Smart reminders based on stress levels Study visualizations (“Explain this visually”) Weekly behavioral reports for self-reflection ADHD-focused interaction mode Counselor dashboard (privacy-safe insights only)
Our vision: Luke becomes a cognitive layer for your operating system — helping you think clearly, stay focused, and act intentionally.
Log in or sign up for Devpost to join the conversation.