lock-backin

Everyone locks out.

During long lectures or study sessions, attention drifts — even when motivation is high. The problem isn’t distraction itself. It’s not knowing what you missed. I built LockBackIn around one idea: Don't burnout, lock in What it does

LockBackIn is a privacy-first focus companion that: -Detects when you drift -Marks what you missed -Generates AI-powered catch-up summaries -Plays your own recorded voice nudge -Tracks long-term burnout patterns -Instead of punishing distraction, it helps you relock immediately.

How we built it: -Next.js + TypeScript -Local computer vision heuristics for focus detection -OpenCV for converting frames to grayscale and computing motion using frame differencing and luminance statistics, all client-side -Web Speech API for transcription -Google Gemini + Featherless AI for structured catch-up cards -Databricks SQL for session analytics All timestamps are session-relative, preventing time drift and keeping segments aligned.

Challenges we ran into -Fixing timestamp drift across sessions -Separating active sessions from historical analytics -Balancing privacy with meaningful data insights -Simplifying the UI into clear Focus / Catch-Up / Insights modes

Accomplishments that we're proud of -Privacy-first architecture -Clean session-based timing system -Dual AI provider integration -Databricks-backed burnout analytics -Personalized voice accountability loop

Built With

• databricks
• featherless
• gemini
• next.js
• opencv
• typescript