UX of todo lists

This is based on the work today, which is quite funny - and very frustrating.

I gave a prompt to Cursor, and explained in a format of:

"We are building X, by doing why:

  • In Z see example how to do this
  • In Z1 how to do that
  • Etc
    "

And Cursor converted it to to-dos. I would say quite accurate, but I saw already that second, out of 4 to-dos, is wrong. And there is no real way to stop it and make it replan it, because the only way is to click stop, write another prompt, which Cursor then adds as a fifth line of todo, and goes more and more into the rabbit hole of halucinating.

On top of it, when this starts happening, it gets stuck at “Generating…”. Then I need again to stop it, tell it it is stuck, it says you are absolutely right, let me do what you asked me to, and goes back to the rabbit hole.

And requests are wasted, time is wasted, and I am closer and closer to look at the alternatives. Using Sonnet 4, non thinking, used Auto, was even less useful, and not sure anymore what to think about this big experiment.

And before someone asks - this is happening with a single chat, and in the new chat(s). Whole day today.