Intercept

Inspiration: Many habit-change tools focus on motivation, tracking, or restriction, yet people still relapse. Through personal experience and research, we realized that failure usually happens during specific, predictable moments — late at night, under stress, or when willpower is low. Existing solutions tend to react after failure or rely on constant self-control.

We were inspired to build Intercept to address this gap: a system that supports users at the moment they are most vulnerable, before relapse occurs. Our goal was to normalize failure as part of behavior change and design a tool that offers timely, empathetic support instead of judgment.

What it does: Intercept is a predictive habit-support application that identifies when a user is most likely to relapse and intervenes before it happens.

The app learns individual patterns based on time of day, recurring behaviors, and self-reported triggers such as stress or boredom. When a high-risk window is detected, Intercept delivers a short, actionable intervention designed to help the user pause, reset, and regain control.

Rather than tracking streaks or penalizing slips, Intercept measures progress through resisted urges, reinforcing positive behavior even during difficult moments.

How we built it: We built Intercept around a privacy-first, pattern-based prediction system.

The application uses time-series analysis to identify recurring high-risk windows by examining historical behavior patterns and contextual signals. Each time window is assigned a confidence score based on frequency, recency, and trigger overlap. Interventions are only triggered when the confidence threshold is met and a real-time confirmation signal (such as a phone unlock) is detected.

All pattern recognition is designed to run on-device to minimize latency and protect user privacy. The system is lightweight, explainable, and improves accuracy as more user-specific data becomes available.

Challenges we ran into: One of our main challenges was balancing effectiveness with user experience. Over-intervention can quickly lead to notification fatigue and disengagement. To address this, we implemented conservative confidence thresholds and kept interventions short and optional.

Another challenge was designing a system that motivates users without reinforcing guilt or shame. Traditional streak-based systems often discourage users after a single slip, so we restructured progress tracking around resilience and recovery instead of perfection.

Accomplishments that we’re proud of: We are proud of building a complete, end-to-end system that connects onboarding, prediction, real-time intervention, and progress tracking into a cohesive experience.

We are especially proud of our focus on empathy and trust. Intercept does not punish users for failure, does not rely on restrictive blocking, and does not monetize personal data. These design decisions were intentional and central to our vision.

What we learned: This project taught us that effective behavior change tools must address emotional and contextual factors, not just logic or information.

We also learned the importance of designing explainable systems that users can trust. In many cases, simpler and more transparent solutions are more effective than complex or opaque ones, especially in sensitive domains like mental health and habit formation.

What’s next for Intercept: Our next step is validating Intercept with real users through a controlled beta to measure prediction accuracy, engagement during high-risk moments, and relapse reduction.

Following validation, we plan to expand into institutional pilots, beginning with universities and wellness programs. Long-term, Intercept will evolve into a habit-agnostic platform that can support multiple behaviors using the same predictive and intervention framework.