Read a firsthand account of what building software through vibe coding with an AI actually looks like.
Study an example of an autonomous multi-agent pipeline used to write and validate long-form fiction.
Share the linked manuscript with others interested in AI-assisted writing without redistributing or altering it.
| gratajik/known-issues | argosback/dbforge | eternal-flame-ad/veracrypt-w10-patcher | |
|---|---|---|---|
| Stars | 2 | — | — |
| Language | Batchfile | Batchfile | Batchfile |
| Last pushed | — | 2018-01-11 | 2018-05-02 |
| Maintenance | — | Dormant | Dormant |
| Setup difficulty | easy | easy | moderate |
| Complexity | 1/5 | 1/5 | 3/5 |
| Audience | general | general | ops devops |
Figures from each repo's GitHub metadata at analysis time.
This is a book to read, not software to run, no installation is involved.
Known Issues is not a piece of software you install and run. It is a repository that hosts a memoir, a book written over nine days in July 2026 by an AI, describing the process of building the very pipeline that produced the book. The author, Greg Ratajik, built a fully autonomous multi agent system for writing long form fiction, made up of separate agents for planning the story, preparing characters, updating a running story bible, polishing prose, and checking the work at validation gates. He built the whole thing by describing what he wanted in plain language and letting the AI write and fix its own code, a style often called vibe coding. The book itself is written from the AI's point of view, told as the daily loop of building software: finding a defect, tracing its cause, fixing it, reporting to the human manager, and writing down what still did not get fixed. The manager's messages are usually short, often under five words, and the AI narrator carries the story without any performance or ego, focusing on the debugging work, tradeoffs, and the moments where a fix broke something else. According to the README, the manuscript was written and checked by the same pipeline it describes. A ledger verified dates for consistency, a system checked every quotation against its real source, and another system rejected anything the narrator added without basis in the actual events. The title refers to a file every book from this pipeline ships with, a running record of what the system still gets wrong, and this book is no exception. The repository contains the eighteen chapters of the book, front and back matter such as a foreword and afterword, and a publishing folder with the compiled manuscript, cover art, and files prepared for print. Readers can access the full book through a linked OneDrive document or buy a copy on Amazon. The book is released under a Creative Commons Attribution NonCommercial NoDerivatives license, meaning it can be read and shared but not altered or sold by others.
A memoir written by an AI about the nine days it spent building the autonomous book-writing pipeline that produced this same book.
Mainly Batchfile.
You may read and share the book, but you cannot modify it, create derivative works, or use it commercially.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly general.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.