F.OS

Inspiration

F.OS started from a personal problem: deep work kept getting interrupted by “quick checks” that turned into distraction spirals. We wanted a system that could intervene in real time, not just report after the fact.

What it does

F.OS is a macOS focus copilot that:

• Tracks active apps/tabs, typing cadence, mouse behavior, idle time, and app switches.
• Classifies behavior as productive vs distracting.
• Learns personal baselines.
• Detects distraction loops and delivers real-time nudges via a menu bar UI.
• Stores data locally (with optional Snowflake analytics sync).

How we built it

• Built Python-based trackers for window, keystroke, mouse, system, and clipboard signals.
• Stored events/sessions in SQLite with structured writes.
• Added a behavior engine for profile learning, pattern detection, and nudge generation.
• Built a macOS menu bar + overlay UX for low-friction interventions.
• Packaged with py2app and created a landing page to present the concept.

Challenges we ran into

• macOS permissions and reliable system event capture.
• Balancing low latency nudges with low CPU overhead.
• Reducing false positives so nudges feel helpful.
• Handling sensitive behavioral data with privacy-first defaults.
• Synchronizing multi-threaded trackers and flush cycles safely.

Accomplishments that we're proud of

• End-to-end pipeline from raw activity signals to real-time intervention.
• Fast distraction detection and immediate nudge triggering.
• Adaptive, user-specific baselines instead of static rules.
• Polished menu bar/overlay experience.
• Local-first architecture with optional cloud analytics.

What we learned

• Timing and trust matter more than raw model complexity.
• UX decisions (tone/frequency of nudges) are as important as algorithms.
• Event-driven systems need careful batching and thread safety.
• Privacy design must be a core product choice, not an add-on.

What's next for F.OS

• Stronger personalization and relapse prediction.
• More user controls (snooze, stricter focus modes, custom rules).
• Better weekly insights/coaching summaries.
• Cross-platform support beyond macOS.
• Broader user testing to tune intervention quality and reduce alert fatigue.

Built With

• appkit
• applescript
• azure
• chromadb
• ioreg
• nsworkspace
• python
• quartz
• snowflake
• snowflake-ai
• snowflake-connecter
• snowflake-ml
• snowflake-sql
• snowflake-streamlit
• sqlite
• window-apis