Cursor On The Side Is The Bottom Now

If you only use Cursor a bit with a chat now and then, you are probably at the bottom of AI adoption right now.

When we look at Senior Product Engineer profiles, almost every CV says "agentic workflows". But when you talk to people about it, you realize how wide that range is.

On the low end, people watch a single Cursor session build a feature for them, half-done, and then do a quick review and test it by hand.

At the high end, a person lets a strong model sort issues by size, gives the small ones to a cheap model to run in parallel, and then lets the strong model find the bugs in the result. They built agent loops whose only job is to find what the coding agent got wrong and improve the process on their own.

Both groups think they are AI adopters, but the gap between them is much bigger than most people think. If you have not even tried what the low end does, you are already out.

To stay relevant as an engineer, using a bit of AI on the side is not enough. Give agents tasks, learn where they fail, and build the checks around them, which takes real hours every week for a long time. Once that setup works, you ship a lot more.