Why Cursor hits different than ChatGPT projects

What I felt as I've been using Cursor is the ingredients are the same, but they're arranged slightly differently. There's slightly different UX because it was built for software development. These small changes add up to very big phenomena in how I work.

If anybody's experienced El Niño, you know that some very small changes in the speed of the wind in a particular part of the ocean changes a little bit, and then the whole globe is completely different.

One small change with Cursor: the project knowledge is editable. The enormous impact is that it gets really smart, really fast. When I was using Claude projects, I had my project knowledge and lots of chats, then I had outputs, and then I would decide which of those becomes project knowledge. In a Cursor environment, the project knowledge just keeps updating in a very tight loop.

Another small change: the initiative doesn't live in a thread like it would in a project. It lives in a doc that just keeps getting better. I've noticed that my chat threads in Cursor become super disposable, super throwaway. In an LLM project, the history matters. But in Cursor, the history is stored in a document. These chats are just little tools that you use to do something and throw them out. So you never really hit a limit.

Another really cool thing: you can use initiatives as context when you're talking about another initiative if you want. They can know about each other. Even the hierarchy in which you organize folders is context to Cursor. That also tells it how things relate to each other.

**➡️ Cursor accidentally built an excellent PM thinking partner. Your knowledge evolves in real-time, initiatives cross-pollinate, and chat becomes disposable because everything important lives in your documents.**