I stopped writing detailed tickets. Then I started using Claude Code.

I spent more time in Jira than I care to remember during my time as Tech Lead on my last team. Shuffling things from one column to the next, agonizing over the burndown chart, and most importantly, writing tickets. My tickets were solid. Anyone on the team (or more importantly, not on the team) could pick them up and understand what it would look like to accomplish them. Engineers could expect edge cases and useful pointers to parts of the codebase. Product managers could stop reading after the user story and confidently understand the work. The only issue was the amount of time they took.

My new team couldn’t have been more different. Tickets took the form of GitHub issues with a single sentence or two describing the intended behavior. When I first joined, I continued my detailed ticket writing habit. As time went on, however, I slowly started getting lazy. When I realized I was the only person reading these tickets, I started writing the bare minimum to capture the key details. Edge cases stopped being explicit and started existing only in my mind.

I’d been using Claude Code for months at this point. I’d mostly crank out a prompt without much thought for smaller features. Plan mode was my weapon of choice for more complex tasks. Then, one day I had the idea to ask Claude to implement a feature based on an old GitHub issue I had written when I still took ticket-writing seriously – acceptance criteria, edge cases, even notes on which components to touch. The end result was stunning. It looked like code I’d have written and something clicked.

Plan mode is spec-driven development and tickets are another form of the same thing. Giving the LLM more context on how I’d implement a solution meant I was happier with the end result because it felt like my work.

What comes next? As models continue to improve, what does this skill become? Does the structure stay user story + acceptance criteria + edge cases + engineering considerations? Or does it become something entirely different now that an LLM might be the first reader?