Fragz

Inspiration

We built Fragz because the modern workspace is broken. Did you know the average worker wastes 40% of their day just switching contexts? Every time you get interrupted, it takes 23 minutes to regain deep focus. In the US alone, this results in an $11 Billion productivity loss annually. We realized we have fantastic tools for planning work (Calendars, Task Managers, Knowledge Bases), but we have absolutely nothing that orchestrates the actual doing of work. We wanted to build the "last mile" of productivity—a tool that forces you into a flow state.

What it does

Fragz is an Intelligent Work Session Orchestrator. It doesn't just list what you need to do; it guides you through doing it.

1. Unified Context: It pulls your tasks, calendar events, and documentation into a single view, removing the need to alt-tab between 5 different apps.
2. Native Focus Mode: Built as a native desktop app, it pins essential tools (like your chat and notes) to the side of your screen while blocking out the noise.
3. Session Management: It tracks your "flow" time, managing breaks and deep work intervals intelligently.

How we built it

We prioritized performance and a "native" feel, powered by state-of-the-art AI. The core application is a hybrid powerhouse:

1. Frontend: We used React with TypeScript for a reactive, type-safe UI.
2. Desktop Engine: We chose Tauri (v2) over Electron. This gives us an incredibly small bundle size and native performance by leveraging the OS's underlying webview.
3. Styling: TailwindCSS for rapid, consistent styling, ensuring a "pinned" layout where sidebars stay fixed while content scrolls independently.
4. AI Intelligence: The brain of Fragz is built on Google's Gemini models. We leverage Gemini 3 Flash for instant, low-latency interactions to keep the UI snappy, while Gemini 3 Pro powers our deep context analysis and complex session orchestration capabilities.

Challenges we ran into

The biggest challenge was bridging the gap between a "Web App" and a "Native App."

1. Layout Constraints: We had to fight the browser's natural tendency to scroll the whole page. We implemented complex CSS flexbox patterns (h-full, overflow-hidden, min-h-0) to ensure that only specific panels scroll, giving the app a solid, native application feel.
2. Type Safety: Implementing a robust Drag-and-Drop system for the Task board required strict TypeScript generic narrowing to handle different data types (Columns vs. Cards) without runtime errors.
3. Cross-Platform Consistency: Ensuring the Tauri window behaved exactly like the Web Dashboard required careful responsive design.

Accomplishments that we're proud of

1. Zero-Latency Drag & Drop: The task board feels incredibly physical and responsive.
2. The "Pinned" Layout: We successfully created a UI that feels like a native OS utility rather than a website in a box.
3. The Metric: We aren't just a todo list; we are a solution to the "23-minute refocus" problem.
4. Intelligent Orchestration: We're incredibly proud of how Gemini 3 Flash instantly analyzes your notes and tasks to suggest the perfect work session without waiting, creating a seamless, lag-free user experience.

What we learned

We learned that Gemini 3 Flash is an absolute game-changer for UI interactivity. Traditionally, AI features felt "bolted on" because users had to wait for responses. With Flash, we could make suggestions feel instantaneous, almost like a native autocomplete. Conversely, integrating Gemini 3 Pro taught us the power of large context windows. Balancing these two models, Flash for speed, Pro for depth, was key to making Fragz feel both fast and intelligent.

What's next for Fragz

We plan to introduce Adaptive Deep Work Sessions. Imagine Fragz not just blocking notifications, but using Gemini 3 massive context window to read your project documentation, summarize relevant emails, and pre-fill your draft before your work block even starts. We want to move from "scheduling tasks" to "preparing the mental environment" for you.

Built With

• fastapi
• gemini-api
• langchain
• langgraph
• nextjs
• pgvector
• postgresql
• python
• rust
• tauri
• typescript