GitHub Copilot: Your AI Companion for Every Workflow

Hello from beautiful Redmond, Washington, and we are coming at you with a live version of GitHub CPilot Dev Days. So, I'm really pumped and glad that you could join us today because today is going to be a great show. We are going to dive deep into all things GitHub Copilot. We are going to be talking about GitHub Copilot on the CLI, GitHub Copilot with VS Code, GitHub Copilot on Visual Studio, and we're going to wrap up the day with a session where you can get your hands dirty where we're going to walk through with a GitHub co-pilot workshop where you can ask live questions. We'll go back and forth and we will build something and we will be great. But first, let me tell you what this GitHub co-pilot dev days is all about. It's more than just this live stream. It's a worldwide community event. It's a great great celebration. All right, here we go. So what copilot GitHub copilot dev days is was because it started way back in April is 300 different community events across the world where 40,000 different attendees came together to learn about GitHub copilot. We had sessions and we had workshops and sessions included everything from learning about the new agenic workflows co-pilot in the CLI the SDK coding agent all over the place and using it in the ide as well and you can see some photos there of the people who got together at the different workshops and so on. So, a lot of fun and you can actually get involved as well yet because it's not all over because if you go to that short link gh.ioall-the- stuff, you can see all the sessions that were used during the various dev days and the workshops too. So, let's say you just want to go and learn on your own, head over there, view the workshops, do it. If you want to go out and maybe use one for a user group, do that, too. So, please go out to gh.ioall this stuff and pull it down, use it, learn it. It's really, really good. All right. So, let's move on here and start talking about Copilot from back way back when 2021 all the way to now and maybe a little bit in the future because you know what AI changes and it changes quickly. So four years ago, Gen AI kind of came onto the scene in coding agents or development AI development assisted uh codec. And so it was, if you remember way back when, it was kind of just like smart autocomplete. And it was good. It was fine, but things were going to get better. So back then, you could kind of think of it as your pair programmer. It was sitting along with you kind of whispering over your shoulder, this is how you would complete this text. And it eventually went to something called ghost text where you started writing an if statement and looked at your code and kind of came up with this would be the rest of your if statement based on what your code is doing. Eventually, it got to be more. And that's what the Agentic software development life cycle is where we can actually start bringing in agents to start doing things for us. Like, so we can now start typing in some specs and handing it off to do the agenic workflow for us where now we're still sitting in the middle. Humans are still there, but we're now reviewing some code and fixing up the loose ends of it and still sitting in the middle making sure everything is going right. And in the future, we going to have hybrid teams where we're going to have teams of agents doing the work, humans doing the work and working together and and moving things forward. So we are right now in the middle stage of this wave two and moving on to wave three where we're going to have these hybrid teams of us and the agents working together to build great and amazing applications. So we have one GitHub co-pilot and it's everywhere you work. And what I mean by that is that you can have GitHub copilot on github.com and the GitHub mobile apps. So what I'll call the quote unquote first-party applications and then of course you can have it in VS code which is where a lot of folks may have originally encountered GitHub copilot also in various IDE as well such as the Jetrains IDE thirdparty IDE and of course it's going to be in Visual Studio 2 and Xcode and it is all over the place and we'll have a couple sessions on this. I'll talk about it and we'll have right after this another session. It's in the terminal co-pilot CLI. It is amazing. I didn't think I would love the terminal until I started using it and now honestly I can't get away from it. It's in Slack and Teams, right? You could be just talking with your teammates, offload something to Copilot from Teams. Pretty cool. And everywhere where you're tracking your work, whether it's in DevOps, linear or Jira, pop it in. You can have copilot there as well. So, it's all over the place where you can start integrating it and keep your workflow going. So, a little bit about the co-pilot in the CLI or GitHub copilot CLI. And I'm just going to say right now when I say copilot, I mean GitHub copilot. So, I'll just stop saying GitHub copilot all the time. Anyways, so the Copilot CLI kind of gives you what we're calling a TUI, a terminal user interface and it has a ton of core features in there where you can actually see diffs of the files that you change and it knows about multifile context of things that it has changed all the files in your project. So it kind of gets that awareness of what to do. It has um MCP servers available to it. It has skills available to it. You can have different modes where it can plan where it can do autopilot where it just kind of goes off and and implements the code until it's finished. All the different things it it is full featured and I'll take a quick peek at it uh in the demo coming up, but we'll also have a different session on this where we'll dive deep into it. And of course, there is Visual Studio Code. If you don't want to work in the CLI, and I don't blame you, you want to have the full F code editor available to you, there is VS Code, and VS Code brings C-Pilot in with everything that you're used to with VS Code. All the extensions are still there. You still can build and test and everything else. You have all the capabilities of co-pilot, agent mode, planning, running in the background and so on. You have all the toolings, custom agents, MCP, skills, and so on all together. You can see your code though right in front of you. So, it's a little bit easier to work with if you're used to um working with, you know, just coming from VS Code and working with all the files in the in the editor. And then Visual Studio the enterprise development uh integrated develop IDE for uhnet works across you know C++ and so on but same deal copilot is in there we can do the same life cycle where you're used to when you're coding by yourself but now you can get copilot in there to help you out it has you know all the agents in there as well you hook MCP servers, you have all the tools. So, you're bringing all the goodness of Visual Studio over into the Aentic development AI assisted world as well. So, all the goodness of Visual Studio plus the goodness of Copilot pushed together. And of course there is Copilot in the third party IDEES Eclipse Jet Brains and Xcode which is really exciting because you still have um things like the custom agents available. You can still do plan mode. You can still uh have instruction files available to it. So all of these things that you might have used in VS Code that is CLI are available in the idees that are not originated from Microsoft. So that's that's actually something I want to drive home that Copilot is available everywhere which is actually super super neat. And then this final slide which is really cool and there's going to be a nice little animation on this one coming up is that there is GitHub mobile right and so nothing from nothing makes your family happier is when you're start uh unloading GitHub copilot agents on a mobile session. So you can do it on the go. You see an issue or a co pull request come through and then you can send an agent after it to um say, you know what, I want to check for any security vulnerabilities with this with this session. So you're out walking the dog and you can steer the agent or kick off a whole agent session and see it work. So and you can see all the other agent sessions that you might have launched from VS Code or from the CLI there as well. So, one co-pilot everywhere that you go. All right, I have done enough talking. So, I'm going to jump into a demo now to um show you what we are looking at between the CLI and VS code all together. All right. So, this is what we are dealing with. It is a very very simple. Of course, I vibed at the blog. And what it does lists the latest post. You can go in, you can see the post. And the way you add a new post is you create a new issue on the repo for the blog itself. Add a title, write a description, and then when it comes through of the issue, what I would do is All I need to do is do a slash approve and it creates the post itself. So that's what we're dealing with today is that we're going to take this blog and make it better by sending across some Aentic AI development with it. So first thing first, let's take a peek at the CLI. Now, as I mentioned, the CLI has a couple different things put into it. Um I just want to show off that we do have MCP servers available to us. So in here I don't have too many uh different ones loaded up. I have the uh Aspire one here. So if I was doing some .NET development, .NET Aspire development, I would have this all there. I actually have a Slack and a Work IQ MCP server and so on. So I can use this just as any other MCP host client as well. There's also things called skills. And so skills are kind of sort of similar to MCP servers. I think they're kind of taking they will take over from MCP servers in that they give you a way to do specialized things and include instructions on how to do the specialized thing that you want to do and even maybe some code files to run. So without getting too deep on what skills are, you can see I have some here like working with word files, working with whoops itself and also want to suggest other skills from awesome copilot. So, what I can just do here um is suggest some skills from awesome copilot, which is a repo that we have that lets you customize the AI output from GitHub Copilot that are pertinent to this project. Just suggest one. All right. So, I'm going to let this skill go up and run and just let it come up. And it should say once it thinks about it after the working that it's going to invoke the skill and then once it says it does, we'll jump in to Yep. It's going to run the skill recommener awesome copilot suggest skills. And so while it goes off and thinks about that, the next thing I want to implement here is or show off is the plan mode of this co-pilot CLI. So if I hit shift tab, I can go between different modes. So this first mode, and you can tell it's just doesn't have any extra coloring at the bottom. This is just regular mode. What I'll call I'm going to do something, talk back and forth to it. It's what I did with the skill. just do something for me. When I go to plan mode, what I'm going to do is I'm going to have Copilot put together a plan to accomplish something for me before it takes any action. So, I'm going to say here, let's implement some page navigation. do it as I have in and if I do an at symbol I can start referencing files that I have in my application. So I already put in some navigation specs that I want copilot to follow. So I'm going to say let's implement some page navigation and follow what I have in the navigation specs file. Read it thoroughly first. and then come up with the plan. Don't bother to ask me any questions. Just do it. All right. So, it's going to go off and figure out the plan to do it. And when it's done, it's going to say, "Here it is. What do you want me to do with it?" Cool. So, as it works with that, I'm going to start to do yet another thing. I am really going to challenge all my context that I have in my head right now. And so this third thing I want to show off is that we are running on autopilot. You can see that has, excuse me, that has that light green at the bottom and it says autopilot. So what autopilot does, it just takes off and runs until the task I give it is finished. And so what I want to implement now on my website is a light and dark mode. So implement light and dark mode as per outlined in the light and dark mode spec. Read that spec thoroughly. I like to tell it to make sure and finish up. Exclamation point. And off we go. All right. Excuse me. So, it's going to go off. Read everything and it should be implementing the light and dark mode. So, I'm going to jump back to our skills that we had it go up and suggest. And it says it went through, looked at everything, and it because this is an Astro site, it's suggesting there's a skill out there called Chrome DevTools. And it thinks that would be the one skill that we should invol invoke here to help us out going forward. Cool. And let's see here. All right. So, next, my plan. It has come up with everything. And now it's asking me, do we want to accept a plan? Do we want to accept a plan or exit? And I'll prompt myself. What do we want to do? I'm going to exit plan mode. And I'll prompt myself. And what I'm going to say here now, I could just go ahead and say, you know what, go ahead and build it. But I'm going to say delegate this. Delegate and implement the plan. And what delegate does is it's going to actually send it up to the coding agent in the cloud and let um GitHub copilot ask for it. So, yep, let's go that. Yes, sounds good. And now it's going to go off create this new branch. Summarize the conversation that we had here. So, it's essentially taking the whole conversation that we had in this particular tab and sending it off to GitHub in the cloud to do everything. And so, what we'll take a peek at this and see it running on github.com in just a second when it runs through here once it starts its work. All right. And let's finally now go see our light and dark mode task is complete. Cool. All right. So, let me just quick jump out to here. I'll refresh. And I see theme light theme dark. Neat. We got it. But great. And so now I'm going to jump back out to this. This is the repo. If I go to agents and I can see now it's implementing the implementation plan. So that was the thing that we just sent off. All right. So now we are moving fast. We are continuing to move fast. And then now we are going to go into our code. And so one thing I want to show off here is I am going to bring up the agent window. And so you can see all the um code that we were or the agent sessions that we're working on. We have implement light and dark mode um implement page navigation and so on. What I am going to do is create a new one down here. And I'm also going to go to view browser. All right. So, this is actually going to be really neat. So, when I go to view browser, I'm going to then bring up my pocket post place here. So, this is integrated within Visual Studio Code, which is actually really, really cool. So now I'm kind of reducing the burden of having me go out and back to see the changes I want to make. And then this button on the URL bar, I can do share with agent. So it's asking me, do I want to allow that? And then it has this really cool highlighting around it. And now my agent has access to that browser. So what I can say here then is change the theme to dark mode. And because the agent has access to the browser, it's going to be able to see. All right, cool. I should have said change it as click on the dark mode button. How about that? Dark SLlight mode button there. Now it should know that it wants I want it to do the clicking there. So it went ahead and did the clicking for me. And so that's actually kind of cool. So now I can actually start iterating on things like this. So let's say I wanted to get rid of the top level header here. That doesn't do anything for me. Remove the top level header that says thoughts published with blah blah blah. Remove the whole section. So, I'm actually now kind of just going one back and forth and then seeing how the um resulting page looks as as we move through it. And there gone. Pretty cool. So, then navigate into a post. Pick any one of them. And so yeah, this is actually really cool. So it kind of gives you the feeling of of um I don't know, it's neat just be able to go back and forth and iterate while watching it in a live mode, live reload mode itself. So very very cool. All right. So the next thing what I want to do is let's see if our page navigation is done here. Still processing. So once that gets done, I can bring it back in. And what I'm going to do then while we wait for that is one of the things that we have um both within the CLI and within um VS Code of course is that we can pick various models that we want to run. Right now I'm using a GPT53 codeex. Works great. But if I wanted to boost things up and have it really think I can run Opus 47 or GPT54. I'm super happy with 53 codeex as you saw as I was going back and forth. Just works great. It also has um MCP servers as well. So for here I look and I have a uh GitHub MCP server within a VS code and so I can actually start looking at all of the all of the PRs and lists all of the issues and so on. So I'm going to say give me a status on any open pull requests for this repo. And so now it's going to should be using the MCP server to see about any poll requests that we have going going on. And another thing I can do is I can change the way the approvals work. I can use it has the default approvals which has it prompt me anytime it wants to do something uh that could be destructive like change some things like run a terminal command or I can go to bypass approval where all tool calls just get approved and so that's what I'm going to do enable that and it's saying came back with my PR and I'll make this bigger so everybody could see a little bit easier is We got one PR number eight and it's working on the back previous navigation. So nothing else has been found quite yet. Cool. So let's see if it has made it up. All right. Very good. It has finished and I'm just going to believe it worked because that is the way I roll. And instead of merging it in the main, I'm going to merge it into my current branch which is called Dev Day Show. Oh, we should be good. All right. So, what I'm going to pull down then is the branch called hostile rodent. And then let's see if it worked. I'm going to bring the agent over. Open this back up. I'm going to start a new session. Click into a post. And I'm not seeing my back and forth. So, let's just make sure. Didn't get it. All right. So, instead of trying to debug it here, since we're running almost out of time, I'm going to jump back to the slides real quick and I will do some debugging. on the on the repo and I'll put it on the readme. All right. So, you have all the latest models that are available, right? So, we took a peek at it. We we looked at uh was running everything uh codeex 53, but you have everything from Anthropic there uh Gemini and you can run everything locally too if you wanted to with Olama or Foundry logo as well. And your one GitHub subscription is going to work everywhere. Right. So we showed it working in github.com. We showed it on the CLI. We showed it in VS Code. It's the same subscription. You don't have to subscribe to different ones. It's just going to work everywhere. And even things that you are so um uh as you would think as normal idees, SQL Server Management School, Studio, and Raycast as well, they work in two. So with that, thank you. That was a high-level quick shot of the CLI and VS Code working together to mess with a mess with a really simple blog. And I'll get that page navigation working and pop it up on the repo and we'll put all that in the show notes.