I let AI build a tool to help me figure out what was waking me up at night

Faced with persistent nighttime disturbances in a noisy city and an inability to pinpoint their source, the author, Martin, embarked on a mission to demystify his sleep interruptions. He conceived of an AI-assisted system to correlate external sounds with his sleep data, transforming a previously daunting diagnostic problem into a feasible weekend engineering endeavor.

• Problem Identification: Martin frequently woke up at 3 AM without knowing the cause, rendering effective solutions impossible, even with sleep tracker data indicating disturbances.
• System Architecture: The solution integrated existing smart home components (Home Assistant, various sensors) with new additions: two cheap USB microphones (one inside, one outside), a Raspberry Pi for audio capture, and sleep data from a Garmin watch.
• Custom Web Application: A self-hosted web app visualizes sleep stages, heart rate, sensor events, and noise incidents on a synced timeline, allowing Martin to easily identify and investigate highlighted wake-up moments.
• AI's Transformative Role: AI tools were crucial for rapid development, enabling the entire project to be completed in approximately 8 hours. AI assisted by generating code, facilitating testing, and even gaining SSH access to the Raspberry Pi for direct iteration and setup.
• Human-in-the-Loop Analysis: While AI built the tool, Martin manually reviewed audio clips to identify specific sounds; AI is not yet used for sound classification.
• Key Findings & Solutions: Analysis quickly revealed common culprits: slamming doors, clattering dishes, and various street noises like motorbikes and trash trucks. Armed with concrete data, Martin implemented targeted fixes, including acoustic panels, improved insulation, and direct conversations.
• Technical Underpinnings: The Raspberry Pi continuously records into an in-memory buffer, applying noise suppression and saving timestamped audio snippets when a volume threshold is crossed. A custom Home Assistant integration controls recording, and open-source libraries fetch Garmin sleep data. The web app stitches these diverse data sources onto a synced timeline.
• Future Enhancements: Potential future features include smarter notifications (only for disturbed nights), AI-driven sound clustering, and automated sound classification.

Martin concludes that AI tooling has significantly reduced the cost and effort involved in building personal, custom solutions. His experience underscores the value of data-driven problem-solving ('measure before you fix') and highlights how integrating various data streams provides crucial context, transforming 'would be nice, not worth building' ideas into achievable weekend projects. This project serves as a compelling case study for leveraging modern AI tools to improve personal quality of life.