Transcription
There are so many ridiculously new, amazing AI features packed into the latest release of Visual Studio Code 1.103. I'm going to walk you through some of my favorites that you can get your hands on today in the mainline VS Code. Let's get into it.
[Music]
All right, first things first is inside of agent mode, we have GBT 5. But we can see it right here in my model selector. And of course, if you select on this model, you'll see all your different models available to you. So here we have GBT5, which is the full GBT model. I'll also point out that now there is GBT5 mini, which is also in preview here at a 0x multiplier. So that is free unlimited requests based on your subscription that you have to GitHub Copilot.
Now, I want to point out another feature that was added now under manage models. This is a feature that's been around for a while to bring your own key. So if you had an Azure AI subscription, Anthropic, maybe you're using Olama locally or you had your own OpenAI or Open Router to pick on anything, you could add your own key and then use those models. Now if you select Copilot, which is a provider, so the GitHub copilot itself, what we'll see is that we have the ability now to manage models that show up when using GitHub Copilot. So if you want to unselect for example GBD4 or you want to maybe get rid of 40 or 03 and the old cloud models you can do this here. You can just simply select these hit okay and now in our model picker we can see that I only have a few models left out here which is awesome. So you can customize exactly you want to see and those will show up of course in agent ask and in edit. So we can see them here as well. So now we have all the other ones available. And if I select back to agent, boom, I'm good to go.
Now, while we're down here talking about agent mode, I want to point out a few other things, which is that we can see that I have the MCP servers and I have this little refresh to start my servers. But I can also hover over this and see that there's automatic start MCP servers when sending a chat message. You can simply tap on that and now the MCP servers will automatically be booted up, which is rad. You can also do that by going into your settings and say MCP and you can see that there is the MCP auto start there. You can do never only new or new and outdated. So those will automatically boot it there which is awesome.
Now on top of that if we select a little tool we can say that the configure tools has also been refactored. So we can see all the built-in tools. We can easily drop these down. They're automatically just dropped down really nice. We see I have GitHub pull requests uh GitHub copilot for Azure. I can see how many are selected here and easily toggle these on or off and hit okay. I can also with one click add an MCP server. I can also add extensions that have MCP servers and also configure modes and tool sets which is RAD. Also point out if you go into your extensions, we can see I have my extensions installed, but we also have MCP servers. And I can tap on this little icon to actually browse MCP servers as well. This will show you MCP servers installed globally, but also ones that are installed inside of just your local configuration, too. So, that's a huge nice improvement there.
Now, another feature that I love is figuring out how much AI usage I've actually been doing on and off every single day. There's a new feature. If you turn on AI stats, you can go ahead and turn this on. AI stats enabled. Once you do that, on the bottom right, you can hover over this little bar, and this is going to show you your AI usage statistics over the last 24 hours. So, AI versus typing averages. So, I've been averaging 84% usage of AI over typing. And I can also see my inline suggestions that I have been using, which is awesome. This is something it's like a mini game for myself is can I fill up my bars every single day inside of here, which I love. So, I love that.
All right. Now, let's talk about actually using models and doing different things. So, the first thing that I'll point out here is that there's a new way of working with the models, which is to give them a to-do or a task list. So, if you come up here and you do to-do, you're going to see uh this one, which is the to-do list enabled. It might be renamed in the future as task. So, you might want to search for task, but it is the to-do list today, at least as far as this stable release goes. And this allows you to basically use to-do lists in the chat automatically. So it will automatically generate to-do lists and then the model will figure out how to implement and follow those as well. So let's try this out.
Now over here inside my application, I've been messing around with a few things. Let me open up the home here. And I've been wanting to add in uh a little banner basically to to to show a survey. So I'm just going to go ahead and paste in my prompt here, which is to add a link to a survey, make it discreet, you know, basically track it. Here I'm going to say use existing uh existing um user settings to store the information about this. And then there's a link to the survey. So I'm going to send this off. Now this is going to go off and it's going to do some working. It's going to do analysis here obviously of what it needs to do and understand what the feature is to implement. And then it's going to come with this task list here. Obviously this is pretty big because I have it really zoomed in. You can kind of see it here. This task list. So, it's going to inspect the user settings, add a survey flag. It's going to extend the API. It's going to add a UI. It's going to add CSS. It's going to wire click handler. It's going to build and run test. And then it's going to do quality gates as well. So now it's going to do this. And now every single time it makes a change to the code, it's now going to go ahead and start to update this. So we can see it go and update the task list as it starts to execute it, which is pretty awesome to see here as well. So it's going through and it's doing all of this work for us. So, as it's figuring out the task, it's going to go ahead and just go to town basically at this point, which is pretty rad. So, I really like this sort of flow here that is showing me what it's going to do. I have this progress of how long it's working for. And I can also kind of analyze the plan ahead of time and say, all right, do I need to follow up with this after it's done? Should I stop it before it even goes any further? And what else maybe do I need to do in this instance to get this done?
All right, cool. So, it finished everything and I can go ahead and collapse this down and kind of get a review of everything here. Now, I want to point out something though is I was kind of waiting around for it to kind of finish this and aimone kind of reviewing all the changes that I was doing. But at the same time, I could be using background agents to do a bunch of these tasks. So, even if I was working on something, I'm not ready to commit right now, I could actually use a background agent to sort of get this working. So, if I go into my settings and I type in agent sessions here. So, agent sessions. Let's go and drop this down. There's a few uh things I can enable. First is that I want to do the UI integration for the GitHub pull requests. And then here I can also say agent sessions view location. and I'm going to say view. What that's going to allow me to do is open this here and see local chat sessions and any of the GitHub copilot coding agents. So these are sessions that are live and then I can also of course see pull requests that were open on my behalf.
Now the difference here is that if I actually open one of these some accessibility features this is opening the codeen coding agent session that has happened on GitHub. I'm reviewing all of it and I can actually follow up with it and review it and look at the code changes. I can tap on it and actually see that code change right here of what it actually did to the files in that coding agent session. So, it's really integrated which is cool. I don't have to go to GitHub. I don't have to do anything like that. I can review all the parts of everything that happened here in that session. But what's really neat is that I can now come in and because I in did this UI integration, I also get this little icon to delegate to the coding agent. So for example, I could say this is a good start. I would like to add the survey to more areas of the application that make sense like configuring a report or if the user has shared history reports or shares history analysis. So now instead of just sending it off maybe I got to hit the hit the road or I logging off for the day I can just delegate this to the coding agent. This will analyze the chat history. And here it's going to ask me if I would like for the context of the chat and my changes to be included in a new pull request. So, let's do it. This is now going to delegate everything off and it's going to change me over to the branch as well. And of course, I can swap back to main if I want to.
Now, the cool part here is that now we can see that the chat, this is all happening in real time. A new pull request has been created for me. I can see that in line here to have the survey link functionality. I can tap on it, go to GitHub, but I also have a live chat session here. So, I can still be working on my current changes here in Maine or whatever branch I'm working on. And then now this is a brand new session that's going to happen. I can start continue work and come back to this at any time or I could pause it. I could stop it and let it do its thing.
So, those are just some of the features. Now I want you to go of course into like the help menu over here if you can see it and then there is show release notes. There is so much more in this release to explore from different MCP integrations, the virtual tools, checkpoints which I didn't even mention. You can go back basically at any point there like checkpoints that you can enable which are super cool. There's tool picker optimizations, tool grouping, a whole bunch of cool stuff. So check out all of these amazing features. And of course, check out the Visual Studio Code YouTube for even more deep dives and the release party as well. So these are just some of my new favorite features. I've been loving this feature of actually seeing all of my coding agents happen in the background as I continue to work on my mainline 2 in this multi- aent environment.
Right, there you have. Those are just some of the features packed into VS Code version 1.103. Now, some of these are previews, some of them are experiments, some are mainline, but they're all in the main branch there. So, you can go ahead and grab and install and update the latest VS Code to get these features when you're using Get Copilot and agent mode and all that good stuff. Well, let me know what other features you love in this release of VS Code because there is tons packed into this release browser release notes and dive in. Put comments below what your favorite are or other things you want me to see covered on this channel. That's going to do it for this one. So, until next time, don't forget to do the things that I always say, like, subscribe. Thanks for watching.
[Music]