Undoc

Overview

Undoc AI was an early attempt at solving one of software development's oldest problems: documentation. Users could upload a repository directly from their machine or provide a Git repository URL. The system would analyze the codebase and generate an editable documentation website similar to what developers were already familiar with from modern libraries and frameworks. The core question behind the product was simple:

How can we turn source code into knowledge?

The Challenge

In 2023, AI-generated documentation was still a relatively unfamiliar concept. There were no established patterns for products like this, and very few direct competitors to learn from. One of the main UX challenges was helping users understand what the AI was doing, what kind of output they should expect, and how much they could trust the generated content.

To reduce this uncertainty, the experience was designed around a familiar workflow:

Repository → Analysis → Generated Documentation → Review & Edit

Product Design

The generated output was presented as a fully editable documentation experience with an inline editor inspired by tools like Notion. Instead of treating AI output as a final artifact, the product treated it as a starting point for collaboration between developers and AI. The visual language intentionally borrowed familiar patterns from modern documentation platforms so that users could immediately understand how to navigate and work with the generated content.

Looking Back

Today, AI-assisted documentation feels almost expected. In 2023, it felt much closer to experimentation.

Although the product was eventually discontinued as AI capabilities became integrated directly into developer tools and coding environments, working on Undoc AI offered an early opportunity to explore how AI could fit naturally into developer workflows.