Cursor is better than copilot. On’y negative for me that it sometimes it goes off the rails, not offering to change, only show code snippets saying “it will be best way to do this way” without suggesting to change specific files.
When I got a difficult bug, it told me to check value of variable. I don’t know vscode’s debugger. So I asked how to see the value, it didn’t tell me how but inserted debug print on every other line. That reminded me how many developer will use print statements instead of actual debugging. I suppose it can’t use IDE’a features beyond inserting the text?
-
Which model is better, in particular for Flutter? Does it matter, different AIs pre-trained for different languages? Or do they differ in their general strength?
-
When to ask with context and without? With context the first time? Or every time? Is without context for follow-ups?
-
Asking with context will cost much more AI credits? Or is the cost per API call, and asking with context generates the same one call?
-
Is it okay to change model in the middle of doing something, or will it lose context? Will it reload context then, or do you have to do that yourself?
-
Is it possible to click accept for the entire answer and not approve every block of change in every file? Or at least accept for each file without having to accept every block.
-
What to do for emulation in Android development? I don’t want to install Genymotion.
-
In Flutter, it’s an Android project and an iOS project in the same repository. Does Cursor know how to deal with that?
-
Should I start a new conversation when, for example, starting work on a new feature? or keep one long conversation for everything?