[androng]

I think I can obtain a triple benefit by using Cursor "ask" mode instead of "agent" mode.

1) I don't over-rely on the AI so I don't accidentally commit bugs

2) I can just put in a OpenAI API key pay-as-you-go instead of subscribing to Cursor Pro monthly and getting screwed by SaaS fee I don't use

3) I actually learn what the AI says and add it to my long-term memory instead of just having it write code for me in Agent mode

admittedly this only works for small tasks, for bigger edits I think trying to learn everything the AI says is not really scalable or at least it takes me much longer.

[brailsafe]

> for bigger edits I think trying to learn everything the AI says is not really scalable or at least it takes me much longer.

Seems like this is the inherent difficulty in being a skillful developer. Atleast in the context of non-trivial collaborative projects, big edits that the person commiting doesn't understand might as well be a diceroll, and imo those big edits should really only be applied if the intent was to save the time in writing it.

[gs17]

The biggest benefit to me (using Copilot instead of Cursor) is that you can make sure that it understands the problem and the solution is what you expected. If I want to see if it can make a big change, I usually flow through "I need X, what are our options" -> "Discuss option N more" -> "Ok, now you can do it".