Using Plan Mode, Model Selection, and Voice-Input
00:00 So one point that I also want to drive home in this overview of Cursor is that you are still in control, even though this has a lot of power, and you will see in just a bit, you can still do stuff by yourself.
00:11 You can write code by yourself, and you can also write terminal commands by yourself. And so let’s start off by doing that. One good practice for software development is that you are going to start off a Git repository right at the beginning.
00:22
So I’m gonna say here, git init. And I’m just typing this normally, as I would if I did not develop with AI right now.
00:31 Let’s see, git status.. on branch, no commits yet because there is nothing in there. But now I have a Git repository in here and I can both track and commit files myself, or I can also ask the AI to do these steps for me.
00:45 And now let’s jump a little bit into viewing the power that some of these tools now present you with. So I’m hopping over to the Agents view. And another thing that I always like to start with is not start with the agent right away, but go into Plan mode first.
00:59 So I’m gonna select Plan mode, and you can also select a different model if you want to. Auto might be a good choice. In this case, I’m gonna go with Sonnet 4.5.
01:08 I like to use this model for planning, and I’m gonna create a prompt that’s gonna instruct the LM what I want to build here. But in plan mode, it’s not actually going to start building anything. First it’s just going to present me with the plan of what it is planning to build, right?
01:23 One thing that I also like to do is to use voice input. You can click on this little symbol here, or use the shortcut. On my Mac, it is Command + Shift + Space.
01:32 And then you can just speak the words in here instead of needing to type them. So let’s try that out: “Hi. I would like to create a FastAPI app
01:43 that can give me back randomized items. I should also be able to add items, update items, delete items. So I want the basic CRUD functionality on a REST API that is built with Python’s FastAPI I’d like you to use Pydantic to make sure for data validation.
02:10 And yeah, I want the basic CRUD endpoints. And then also another endpoint that gives me back all the string items that I’ve added to the
02:22 app in a randomized order.” All right, so I stopped that now. You can see that the voice is not 100% great yet, so I might want to go in here and edit it a bit.
02:36 Depends also how fluently you speak probably. But again, even if you submit this by voice, you still have a chance to edit it.
02:47 Okay, “I’d like to use Pydantic to make sure that there is validation.” Alright, so this is a bit me conversationally talking, and maybe it’s not entirely great, you know, what it recorded in here.
02:58 There are also external tools that you can use for this voice input, but it’s nice that you have this included already in Cursor. Now let me add some more information.
03:08
“Please manage the project using uv, uv init, uv add, etc., and use a modern pyproject.toml format.
03:24 Also, please use modern Python type hints,
03:31
avoiding from __future__ import,
03:36 and follow PEP 8 everywhere.
03:41 Alright, so I’ve added a bit more project information by text. So it’s either you can speak it, or you can also type it in there. And now I could go off and create this plan, but before you’ll actually watch that plan getting made, I want to briefly tell you about the option to use multiple models for a task.
Become a Member to join the conversation.
