The bug that only happened 0.3% of the time (And how AI solved it in 30 seconds)
As an independent software engineer, I spend a lot of time debugging systems. Most bugs are straightforward. Some are difficult. A few are the kind that consume entire weekends.
The worst are the bugs that only happen occasionally.
I was searching for the bug. And noticed a book called “50 AI Workflows for Engineers”. A chapter of that book in particular stuck with me. It described a service that worked correctly 99.7% of the time. No crashes. No stack traces. No obvious errors. Just a tiny percentage of requests quietly returning incorrect results.
If you’ve ever dealt with a bug like that, you know how painful it can be. Traditional debugging often turns into hours of scrolling through logs, comparing timestamps, and chasing theories that go nowhere.
Get Alan Hsu’s stories in your inbox
Join Medium for free to get updates from this writer.
What I found interesting was the workflow the book proposed. Instead of manually reading logs line by line, it suggested gathering examples of known-good and known-bad requests, extracting the surrounding log windows, and using AI to compare them systematically. The goal isn’t to have AI magically solve the problem, but to help surface patterns and correlations that humans might never notice when reading logs sequentially.
That idea resonated with me because it highlights what AI does best in engineering: pattern discovery. Not replacing engineering judgment, but helping us find signals hidden inside enormous amounts of data.
I’ve read plenty of books about AI for developers, but what I appreciated about this one is that it’s focused on practical workflows rather than generic prompting advice. Every chapter is grounded in real engineering problems and offers concrete ways to use AI as part of your daily work.
What I appreciated about the book is that it focuses on practical workflows like this rather than generic prompt advice. The examples are grounded in real engineering work: debugging production issues, analyzing logs, reviewing code, designing systems, and building reliable AI-assisted development habits. As someone who has read far too many books that promise to teach AI for developers but mostly discuss theory, this was a refreshing change.
If you’re a software engineer looking for concrete ways to integrate AI into your daily workflow, I found this book to be one of the more useful resources I’ve come across recently:
50 AI Workflows for Engineers: From Debugging to System Design, Code Review & Engineering Automation
The chapter on AI-assisted log analysis alone made me rethink how I approach intermittent production issues. And if you’ve ever spent three days chasing a bug that only appears 0.3% of the time, you’ll probably appreciate it too.
