Inspiration
Our DoS challenged us to keep screen time under 30 minutes a day. We wanted something with commercial potential and a mischievous twist, so we built “benevolent malware” that treats procrastination like a bug to quarantine. It’s a virtual DoS: persistent, hard to ignore, and (intentionally) a bit funny.
What it does
Lock3d1n monitors window/tab focus (OS + Chrome), keyboard activity, and lightweight camera cues (eyes open/closed, away/down) to infer attention. It classifies activity (whitelist/blacklist/unclassified) and triggers escalating interventions, such as forced Anki reviews, popups that re-open if dismissed, alert sounds, and occasional screen flashes, until we’re back on task. On approved apps, it stays quiet unless we idle too long.
How we built it
- Inputs: Window focus switching (Windows/macOS), Chrome tab switching (Windows/macOS), Keyboard activity, and simple on-device camera analysis.
- Classifier: Uses a pre-trained ConvNet and facial landmark detection (see here) to check if your eyes are open/closed for certain durations. Also determines tab/window switches and uses those to see if you're drifting away from work.
- Policy Engine: Maps states to actions (Anki redirect, persistent popups, sounds) and re-prompts you if you lose focus on Anki
- UX: Popups, minimal controls, local processing, opt-in toggle.
Challenges we ran into
Cross-platform compatibility. Accelerating and augmenting the eye-tracking algorithm for better correctness/detection rates.
Accomplishments that we're proud of
Making a virus! Implementing/upgrading an ML eye-tracker. Sleeping for this one...
What we learned
The value of patience in creativity/ideation.
What's next for Lock3d1n
Integration into KuDoS. Dynamic learning of 'focus' indicators per-student, and more advanced feature analysis.
Log in or sign up for Devpost to join the conversation.