Structured, semantic, and reusable — introducing Inkwell through a focused, minimal landing built for clarity and flow.
Inkwell Docs: Displaying Inkwell’s document editor with modular, AI-aware writing blocks.
Inkwell Spreadsheets: Showcasing Inkwell’s intelligent spreadsheet integration for dynamic data linking.
Featuring Inkwell’s unified folder view for cross-document navigation and context.

Inkwell – The Future of Long-Form Documents

Inspiration

Documents are everywhere, but document tools haven’t evolved.
Whether you're writing a contract, a research paper, a policy, or a strategic memo—you're stuck with tools made for letter writing.

We were inspired by a simple realization:
Writing is now structured, collaborative, and dynamic—but our editors are still static, manual, and fragile.

Inkwell was created to bring modern structure, semantic intelligence, and effortless reuse into long-form content.

What it does

Inkwell is a structure-first, AI-powered editor for complex documents.

It understands and transforms your writing into modular, intelligent blocks:
– Forecasts, clauses, insights, summaries, tables
– Easily moved between documents and spreadsheets
– Automatically styled, linked, and reusable
– Instantly exportable in DOCX, PDF, or Markdown

It’s like having an IDE for serious writing—one that understands not just the words, but the meaning and purpose behind them.

How we built it

Built with React, GPT-4, and hosted via Netlify
Developed a semantic block engine capable of parsing natural language into labeled, typed elements
Built a real-time AI assistant that works across documents, spreadsheets, and notes
Designed a system for modular reuse and version control — every block can be tracked, reused, and updated everywhere
Backed by LaTeX as the single source of truth (rawLatex)
Real-time WYSIWYG with live-rendered math (KaTeX) and parsed LaTeX commands
Hosted a live prototype using React + GPT backend, now available at:
link

Challenges we ran into

Token usage is significant: managing multiple files and context-aware prompts under a single GPT session strains limits and cost
Structural syncing: maintaining accurate updates across reused blocks (especially in long chains) required careful architecture
Styling + consistency: building a unified visual system for all output formats proved more complex than expected
Performance limitations: current demo is hosted on a personal GPT account, so performance may vary under load

Accomplishments that we're proud of

Created a no-code LaTeX editor that seamlessly communicates with Spreadsheets.
Converted a narrative forecast into structured spreadsheet data with one prompt
Reversed the process: selected metrics → auto-generated narrative summaries
Created an end-to-end system where every doc, clause, and number is reusable and traceable
Built a platform that matches the real-world structure of professional writing, not just casual text generation

What we learned

Writers don’t just want AI to “generate”—they want tools that remember, reference, and evolve
Structure is the missing layer in modern productivity
AI is powerful, but only when it’s aligned with intent, memory, and formatting context
A modular model creates new paths for automation, collaboration, and scale