Inspiration

When we set out to build a productivity companion this weekend, we wanted to challenge the way traditional software handles burnout and focus. Most productivity apps act like anxious middle managers—flashing red badges, aggressive Pomodoro timers, and streak counters that punish you the moment your attention slips.

Our project, For Frog's Sake 🐸, was inspired by three very different sources:

  1. The Anti-Desktop Goose: We loved the chaotic charm of virtual desktop companions like Desktop Goose, but wanted to flip the premise. Instead of a goose hijacking your cursor to cause chaos, what if a tiny desktop creature watched your cursor for you, noticing when you are spiraling and gently offering to share the weight?
  2. NASA-TLX & Implicit Load: In human factors engineering, the NASA Task Load Index measures subjective workload across dimensions like mental demand and frustration. Measuring this usually requires interrupting users with surveys. We realized frustration leaks directly into motor behavior—frantic typing, backspace loops, rapid tab-switching, and frozen cursors. By tracking these behavior patterns passively and privately, we could measure cognitive load without asking a single question.
  3. Felipe the Feeling Frog: At the emotional core is a childhood story we found online about Felipe, a small green frog who felt everything deeply. When the pond grew too loud, Felipe didn't fight the current—he paused on his lily pad, took one big belly breath, and let the ripples settle. That became our app's soul: a companion on a translucent lily pad in the bottom-right corner of your screen that reminds you to breathe.

How We Built It

We engineered For Frog's Sake as a lightweight, macOS-first desktop overlay using Tauri v2, combining a high-performance Rust telemetry backend with a React 19 + TypeScript kinetic frontend.

  • Privacy-First Telemetry (Rust): The Rust daemon runs locally using bounded 30-second ring buffers via rdev and active-win-pos-rs. Zero key contents, screenshots, or clipboard data are ever recorded. It tracks pure timing and kinematics across seven triggers, including frantic typing, backspace loops, cursor hesitation, oscillation, tab thrashing, idle freezes, and doomscrolling. Every 5 seconds, a normalized Pond-Ripple Stress Score ($S \in [0, 1]$) is computed from these motor signals to determine intervention severity.

  • Context-Aware AI (Gemini): When a trigger crosses threshold, an evaluation loop queries Gemini (gemini-2.5-flash) to generate a single, warm, frog-flavored sentence naming the exact friction signal, with deterministic fallbacks if offline.

  • Kinetic Overlay & Audio: Built using Tauri's frameless, transparent window with a click-through hitbox. Interactions launch guided resets—like belly breathing, visual grounding, and mindful pauses—accompanied by native Web Audio API synthesized chimes with zero bundled audio files.

  • Selection Summaries: An optional feature where the focused app's selected text is read via macOS Accessibility (never the clipboard) to offer a Gemini summary for text highlights over 400 characters.


Challenges We Faced

  • Transparent Click-Through Hitboxes on macOS: An always-on-top overlay covering the screen corner can easily block clicks meant for underlying apps. We solved this with a 120ms cursor-tracking loop in React that dynamically toggles Tauri's setIgnoreCursorEvents based on pointer intersection with the frog, lily pad, or speech bubble.

  • Preventing Notification Irony: An anti-burnout tool that spams you causes more stress than it relieves. We built a strict priority hierarchy, 10-second auto-dismissal TTLs, and mode-aware cooldowns (Focus 25s, Fun 30s, Chill 120s with doomscrolling muted).

  • Cross-App Accessibility Quirks: Reading selected text without touching the clipboard required handling macOS Accessibility APIs (AXSelectedText) across native Cocoa apps, Chromium browsers, and Electron editors like VS Code, implementing retry logic while accessibility trees initialize.


What We Learned

  • Subtlety beats gamification: Users experiencing executive dysfunction or ADHD paralysis don't need points, streaks, or guilt—they need permission to pause. Framing interventions around physical empathy rather than productivity metrics changes everything.
  • Kinematics tell a clear story safely: You don't need to read a word someone types to know they are stuck in a perfectionism loop. Velocity, hesitation, and backspace rhythm are remarkably expressive signals on their own.

  • Rust + Tauri v2 is a powerhouse for desktop tools: Keeping the telemetry daemon in Rust maintained virtually invisible CPU and memory overhead while enabling smooth 60fps Framer Motion animations in the webview.

Built With

Share this project:

Updates

Submission history