Member-only story
I’ve Paid for Copilot, Cursor, Claude Code, and Codex. I Still Haven’t Picked One.
What each one actually does in my week, what I spend, and why an impressive demo no longer makes me switch.
Probably around $70 to $100. That’s what I spend on AI coding tools in the months when I’m actively using several paid ones. It varies because I don’t keep every subscription running all the time, and I haven’t found a combination I’m happy to pay for indefinitely.
English isn’t my first language, so I used AI to help translate, organize, and write the English sentences in this piece. The experiences and opinions are my own.
I started with GitHub Copilot, then spent more time with Cursor, followed by Claude Code and ChatGPT/Codex. Initially, I assumed I’d try a few and settle on one. That hasn’t really happened. There’s plenty of overlap, but I’ve found myself reaching for different tools depending on the work.
No clean breaks
I haven’t made many clean breaks. Usually, I start using one tool more and another less.
Copilot was the first to become part of my everyday workflow because it was right there in the editor. It helped with things I already knew how to write: DTO mappings, validation, LINQ queries, test setup. I could start a method and let it fill in some of the repetitive code.
I started looking beyond it when I needed help with changes that crossed several layers of an application. In the way I was using…
