Back to browse
I generated 235 system docs in a day using GPT-5.5

I generated 235 system docs in a day using GPT-5.5

by robeym·Jun 10, 2026·4 points·0 comments

AI Analysis

MidShip It

AI-generated docs for their own ERP, not a tool others can use.

Strengths
  • llms.txt and llms-full.txt files optimized for AI tool consumption
  • Best-practices template could help other teams replicate the workflow
Weaknesses
  • No actual tool shipped—just a case study about internal documentation process
  • Docs are for their proprietary ERP, not a reusable solution
Target Audience

ERP users and teams documenting complex systems

Post Description

I generated 235 system documentation pages for a startup in about 8 hours using GPT-5.5 in Codex.

We've been putting off the chore of writing technical docs since our team could handle all customer questions directly. However, since releasing an AI tool in the system, users have started asking it system questions as opposed to just data questions like it was designed for. It couldn't answer how-to questions, so, users got confused.

We didn't have a direct way to teach the Ai about our system since our internal docs were thin and the codebase large.

We ended up defining some good documentation practices in best-doc-practices.md, passed it to GPT-5.5, and asked it to write system docs for every module in the system (hundreds), handling 5-10 at a time for context limits.

To our surprise, it worked very well. The tool had read-access to all relevant server/client/database files, so the output was accurate and aligned to what we defined in the best-doc-practices.md.

Similar Projects