How do you use these different “modes”?
I just use agent and sonnet for everything. If I’m chatting, I just say it should not execute, just plan or analyze. That way I don’t have to switch context.
Am I missing something or?
Pro plan user.
Agent, when I run out of premmium requests, the slow requests breaks my flow, i like to progress iteratively and fast. To work on simpler tasks it is too much. So, I resort to it for tasks chat gives me a hard time to acomplish.
Normal I dont use because it prompts the user for confirmations agent would just continue till finished.
So chat is where I try to spend most of my time in, it requires more precise instructions though, I understand its context is not as automated as agent’s (I might be wrong but I feels this way to me), so writing instructions templates I can reuse in chat is the way to go for me.
If I’m focused in a specific line, property, function, etc, I just use CMD K to keep myself in the editor and stay in the flow.
In summary as resources are priced, including slow requests which are free but the price here is your time waiting, I try to use the right tool for the right task to move as fast as possible.
Regarding models, I use
agent: sonnet;
chat: deepseek v3;
cmd k: any fast model, the mini ones, these are for simpler tasks.
1 Like