DraftProse

Cursor for authors. The AI-native writing studio for long-form storytelling, supercharging the next generation of human authors.

Comment

Project Description

DraftProse is an AI-native writing studio designed specifically for long-form content (novels, theses, screenplays). While standard editors treat text as a linear stream, DraftProse treats it as a structured project. It combines the organizational power of industry standards like Scrivener with the generative intelligence of Google's Gemini models.

Key features include:

  • Binder: A drag-and-drop tree for organizing chapters and scenes.
  • Corkboard: A spatial view to visualize and rearrange the narrative structure.
  • Muse: An integrated AI writing partner that understands your entire project context.
  • Focus Mode: A distraction-free interface for deep work.
  • Compiler: A powerful export engine to turn the project into a formatted manuscript.

How we built it

The entire application was architected and coded using Antigravity, an advanced agentic coding assistant powered by Gemini 3 Pro.

  • Frontend: Next.js (App Router), TypeScript, Tailwind CSS, Shadcn/UI for a premium component library.
  • State Management: Zustand for complex client-side project state (Binder/Corkboard sync).
  • Drag & Drop: @dnd-kit for the intricate nesting logic in the Binder and Corkboard.
  • AI Integration: Gemini 3 Flash is integrated directly into the "Muse" chat interface. We chose Flash for its exceptional speed and 1M token context window, allowing it to hold the entire manuscript in context for continuity checks without latency.

Challenges we ran into

  • Complex State Management: Syncing a recursive tree structure (Binder) with a flat spatial view (Corkboard) and handling drag-and-drop operations between them was technically demanding.
  • Recursive File Exports: Building a "Compiler" that can flatten a nested tree of documents into a linear, formatted DOCX/PDF required implementing a custom traversal engine (CompileEngine.ts).

Accomplishments that we're proud of

  • Antigravity Workflow: We built a complex, full-stack application significantly faster by pair-programming with Gemini 3 Pro agentically.
  • The Compiler: Successfully building a client-side export engine that handles formatting and file generation.
  • Visual Polish: The app feels incredibly smooth, with "Typewriter Scrolling" and glassmorphism UI elements that rival native desktop apps.

What we learned

  • Agentic Coding: Leveraging a high-reasoning model like Gemini 3 Pro allows for architectural-level discussions, not just code completion.
  • Context is King: For AI writing assistants, passing the right context (character notes, previous chapters) is more important than the prompt itself.

What's next for DraftProse

  • Cloud Sync: Moving from local browser persistence to a real-time collaborative cloud database.
  • Multi-Modal Muse: allowing users to upload sketches of maps or characters for Gemini to analyze.
  • Marketplace: Sharing and selling plot templates and character sheets.

Built With

Share this project:

Updates