Independent developer Artur Sapek has launched Revise, an AI-powered word processor that combines document editing with visual revision history and an integrated AI agent capable of reading, understanding, and modifying documents. The tool, announced via Hacker News on March 22, was built over ten months — seven of them full-time — using agentic coding tools for much of the implementation.
Revise’s core differentiator is its approach to AI-assisted editing. Rather than offering isolated AI suggestions in a sidebar or chat panel, the editor embeds an AI agent directly into the document workflow. The agent can read the full document context, make targeted edits across multiple sections simultaneously, and produce changes that appear in the same visual revision history as human edits. Users can review, accept, or revert AI modifications using the same interface they use for tracking manual changes.
The technical stack uses Y.js, a conflict-free replicated data type (CRDT) library, for its collaborative editing foundation. CRDTs enable real-time collaboration without a central server arbitrating edits — a design choice that allows both human users and AI agents to make concurrent modifications without conflicts. This architecture treats the AI agent as another collaborator rather than an external tool, which simplifies the revision history into a single unified timeline.
The product enters a crowded document editing market where AI features have become standard. Google Docs added Gemini integration in 2025, Microsoft Copilot is embedded throughout Word, and Notion AI provides document-level assistance. Revise’s bet is that a purpose-built editor designed around AI collaboration from the ground up can offer a more coherent experience than AI features bolted onto existing document platforms. The visual revision history — showing exactly what the AI changed and when — addresses a trust problem that plagues AI editing tools: users want to see and control what was modified.
Revise is built by Migraine, Inc. and is available as a web application. The developer noted on Hacker News that the project demonstrates the viability of solo developers building complex productivity tools using AI coding assistants — the editor itself was substantially built with agentic coding tools, making it both a product of and a tool for AI-assisted work. Pricing and availability details for teams and enterprise use have not yet been announced.
