RDT: Exploring the GitHub Copilot App

Hi. Okay, let me make sure that you can see me. Hello. Hello. Oh, hola. Ola. I can see some people already in the chat. Welcome, welcome, welcome, welcome to the stream today. Ah, my camera is just slightly off for some reason today, but it's okay. Um, welcome to the stream. Please say hello in the chat. Let me know if you can hear me. Let me know where you're watching from. Hola, Andrea. It's nice to see you here in the chat. Welcome to Rabbit Duck. Uh, welcome to Rockbuck Thursdays. I can see some people saying, "What is this?" As usual, I don't know everyone. We always have this stream every Thursday and so this might be your first time seeing this stream and that's okay. Please tell me where you are watching from. I like to know where everyone's watching from and so would be great to say hello hi to Ense I want to say from South Africa. I love the South African flag. Love South Africa. I love Cape Town specifically. It's a great city. Uh how's everyone doing? Where are you watching from? What are you building this week? Are you building anything interesting? Let me know in the chat. Okay, I can see Ozod Beck uh letting me know that PHP is forever. What's your favorite coding language? If you have one, you can drop that as well in the chat. My name is Marene and so today in the stream today, we are going to be just chatting about a bunch of different things and hopefully this should be a fun stream. We are going to be doing perfect. Great, great, great, great, great. I can see people on the stream. Jason is saying hello from Atlanta. It's great to see you, Jason. Uh, where's everyone calling in from? I can see Adu Adiku from Nigeria. It's nice to see you as well. Hello, Shivam from India. Nice to see you. Hi everyone. Hi everyone. Say hi in the chat if you're joining. Some people's first time today. Hi Bai. Good to see you. Um hello. Hello Syninja. Glad you can hear me and see me. What's everyone working on? Hi Hajger. Nice to see you as well. I'm just going to make sure that the stream is clear. Is my video clear? Is another thing that I'm a little bit concerned about. I want to make sure my video is clear. But uh I think we should be having a good stream today. And the main things we have a couple of things that we want to chat about today. The main thing we're going to be talking about today is the new GitHub co-pilot app. I don't know if you've seen this. Hi Asher from Pakistan. Hey Musa from Italy. Good to see you. Hello most South African. No. Okay. No. No. We already said hi to you from South Africa. Vicash from India, good to see you. Muhammad as well, good to see you. Uh from Dubai, nice to see you as well. Sinjit Betty from Scotland. Hi. I'm in London, so like a couple of hours away from you. I'm also a couple of ways away hours away from Melinda who is in Portugal. So good to see you all. Um what's everyone building? I can see Basit is saying he is building an ecommerce web application. That is very cool. I love this. Um is building an ecosystem. Oh my gosh. Hi Stephen from Zimbabwe. I'm originally from Zimbabwe. So it's nice to see other Zimbabweans on the stream. Hello. Hello. Hello Christian. Good to see you. Vladimir as well from Alicante which is also kind of close to me. It's good to see everyone. Uh we're working on things like MCP servers. I can see someone working on an MCP server uh project and Parfait is saying hello to GitHub I think. But today everyone we are going to be looking at the GitHub copilot app. So this is mainly what we are going to be uh talking about today. So if you're interested in this, this is what we are going to be doing. So, I would like to see if I can hang on. I want to see if I can improve my streaming quality because I feel like my it's a bit fuzzy for me. I don't know why it's a little bit fuzzy. Let me switch over to a different network and see. Switching over to another Wi-Fi connection. I am trying out different Wi-Fi streams today. So hopefully that'll be fine. Great. It's working now. But has everyone heard about the Okay, let's hear what other people are saying. Betty is saying she is working on a personal project which is the timecost.com measuring purchases as time also classroom and couples app. I love that. Very cool. Hello to Elvis from Florida. Nice to see you. Um great to see Andrea who's a software architect as well. Um so good to see everyone today. What we are going to be looking at today is going to be the GitHub co-pilot application. This is what we're gonna do for the stream. Um, and if you have not heard of the app, the GitHub app, I'm gonna switch over and we are going to be chatting about that. Hey, Betty again says left England, London a couple of weeks ago. Um, went over to see the rainbow. It's beautiful here and it's Yeah, so many things to see. Musa says, "I'm a teacher and I teach GitHub every day." I'm so happy to hear that. Ismail saying hi from Egypt. So just good to see everyone. Hi from Korea, from South Korea I guess as well from Kangho. Um and I also saw Yasin from Ethiopia as well. So today we are actually going to be talking about the GitHub copilot app and we are going to be trying our best to go ahead and build with this application. I don't know how many of you have heard of the app before. So, I'm going to switch over here. Uh, no, let me switch over here. So, this is a brand new app uh that GitHub has released recently. This is quite new. And so, today we are going to actually be testing it out. We're going to be seeing if we like the app. Um, there's lots of new things for us to explore in this application. So, just let me know. Have you heard about this app? Let me know in the chat if you have. It's a very cool application. If you were here for last week's stream, you'll know that we built a game in this application. And this is kind of what I'm going to be building on today to show you some other new features. The app is actually GA, so it's generally available for you to use. So if you would like to use this app, um, and you want to test it out, I have left a link in the comments for where you can access the app. So I'm going to scroll up and I'm also going to show this on the screen. This is the app. if you want to download it. This is going to be just a new way to be able to use GitHub with AI. So, there's so many things you can build with this application. If you were here for last week's stream, you'll probably have seen that we built this rubber duck game. And there's a number of things you can do with the game. Let me show you. We literally built this on the screen on the stream, by the way. So, and we were able to like have a duck. Oh, and we also had like a someone was on the stream and we were able to actually have um a a sort of uh Easter egg where if you press the duck arrows twice and to go up, it just works. So, oh, there's also going to be some ah, okay, I'm just going to play a little bit. You know, I'm terrible at games. And this is what I said last week as well because I'm not very good at gaming. But this is actually what we built all entirely in the app last week. And so, we are going to this week be attempting to improve this game while we explore new features of the app. One other thing I want to show you before we get started is we, you know, there's been a ton of updates. One of the main things that people really liked in last week's stream and that I actually really like about the app is this animation. So, we have uh we have Mona, who is the GitHub Octo. You can move Mona around in the app, and you can play a game with Mona. So, you can go like this. Last time I actually got the game wrong, like I was like trying to play it and then I didn't know how to jump and stuff, but it's actually not difficult to jump. You just have to like time it and Oh. Oh, wait. Okay. Okay. Yeah. Yeah. There we go. There we go. And this is a really fun way if you're building something. Um I don't know what this green thing is. Oh. Oh. Oh my gosh. Wow. I loved the green thing. No. No. Oh, no. All right, I have fallen. But that was a new best score for me. I have 33 points now. That was incredible. But there's so many little Easter eggs in this app that just make it really fun to use and a really fun experience. Like even if Mona falls, she's gonna come back. She's gonna land in some water and she's gonna come back. So, it's really fun. What has anyone tried the app before? Let me go over to the comments to see. Have we actually given the app a uh a try? What are our thoughts? Hello to Sirly Dev, uh, who I usually see on this stream. Nice to see you over from Twitch. And Sirly is saying, "I missed the notification. How did you get to the game?" So, we are going to be showing how we actually can build stuff in this game. So, last week, if you missed the last one, the the stream last week, we No, it was a week before last week where we built this rubber duck game in the app. And what we're using today is someone is saying, okay, okay, okay. I gotta say I gotta say uh M man Verna said that it's similar to the snake app the snake game on Nokia. Yes, the game that we built is similar to that. Someone says I'm not great at at playing games. I've already said I'm not I've never said that I was good. Um Christian who has tried How many of us have actually tried the app before? because this app is so fun. One of the most addictive features is for sure this game with Mona. And so I did post I'm going to show it again. If you want to try the app right now, go ahead and go to that link. It's github.com github and then you can go ahead and try the game yourself. So, uh, this is definitely one of the funnest features I would say is this being able to play with Mono when you're bored and you can like go and like I'm just terrible. I just I'm just not great at games. I got zero on this app. But like it's such an addictive like feature is like playing with the game once you're building stuff. So, what does everyone say? Shirley is telling us that I mostly use the CLI but I could be convinced to use the app. I think you should use the app. Let's give it a try. This is the the app is new. So if you have not heard about the app, it's because it was announced for the first time at Microsoft Build and today it just went G. So actually in the evening stream we will be having Cassidy and a bunch of people from the VS Code and GitHub team are going to be on to talk about the game to talk about the app and how you can build with this app. But today we are going to go back we last time uh we built this snake this rubber duck game which is supposed to be like the snake game but it's obviously for our stream today. Uh, Serie is saying, "The sick game on an Nokia was so addictive." One of my former team leaders says, "You could always tell it was me and one of the cubicles because of the clicking rapidly." Yes, this is the thing. We want to make a game that people can enjoy and will have fun playing. And so, we were actually on the last stream we did, we used the app to build a version of this game. So, if you missed when I was playing it earlier, basically it's a rubber duck and Oh, okay. No, we're going to try and actually get at least one point so that I demonstrate what the game is like. Okay, good. We got it. So, there's uh there's this piece of bread that we you're supposed to eat as the duck. And then there's also different things that come up. Oh, okay. At least I got some scores here. But um basically what we're going to do is we're going to explore different ways to use this this app. So you don't it's not only of course it's not only for playing games. It's also for working on things. So one of the reasons why I love this app is that there's a number of different things you can do. So something that you could do for example that we talked about last week is that you can go ahead and schedule different automations. So one of the automations uh that I have that I let's how do we wait let's how can I take a look at the automation this is new so I'm also going to be uh running it with you. So if I go to the automations and I run this, you can decide which automations you want to run at which times. And here for example um I have a weekly notification running where what happens is that the weekly notification will look through my Slack. So I've connected the Slack MCP server. So, what I'm just making sure there's nothing on the screen that you're going to be like, okay, anyway, it's fine. Um, but what will happen is that you schedule a prompt. So, my prompt is generate a weekly priority brief for the upcoming work week and pull data from these sources. So, I've connected different MCP servers. So I have the Slack MCP server and then I have my email connected with work IQ and my calendar also connected and also my team's chat. So I work for both Microsoft and GitHub and the interesting thing is that a lot of the times with the GitHub or we do a lot in Slack and then at Microsoft on the Microsoft side of things I do a lot in Microsoft Teams. And so I have two MC servers that I connect. when I connect to Slack and I ask the agent, can you look through my Slack messages to see what's important for the week for me to do? And I also connect to something called work IQ. And work IQ is also going to look through, it's going to run some commands to pull in my emails. It's going to pull in my team's chats and it's going to look through all of these and then ultimately will generate a report of like what are my core priorities to do during the week. So this is running like you can see in the background, but we can see that it's searching Slack and it's searching Work IQ at the same time. If we go to the automations, we can also go ahead and browse templates. So there's a couple of things we can do. For example, I really like this issue triaging template. And when we use the issue triaging template, for example, we can it's going to we choose a time. So we can say, you know, maybe we want it to be manual or we want it to run daily or like I said with my work priorities one we have we can set this to run weekly and so every week it's going to either triage issues or set my priorities. So for this one for example I'm going to set it to manual so that we can just run it and take an example. So this is an issue triage automation. It's going to run this prompt to say review the latest GitHub issues and propose priorities and owners. So I'm going to select a project. One of the projects that I'm helping on right now is lang chain Azure. I'm working on that a lot as a maintainer. And so I'm going to create an automation for this. And then when I press run, it's going to run that automation on its own. Again, like I said, I can go ahead and schedule this automation. So I can manually click the run button or I can go ahead and schedule it so that the automation is going to run every week. And while it's running, for example, here with the Langchain Azure repository, it's going to look at all of the issues with the issue triag. It's going to look at all of my issues. It's going to fetch the latest issues and then it's going to help me prioritize those issues and assign help me also assign people to uh do those issues. So, while that's running, let's see what people are saying about the app. So, um, uh, Sirly Dev is saying, "I installed the app last week when I saw you grabbing the octagots and seeing the anding the mouse." Yes, very true. It's really fun. I can really recommend. Hi to Perth, I think. Al, nice to see you. Hello to Alaii from Nigeria. Um, yeah, Andrea is saying that it can only be addictive without self-control. The concept is brilliant. Yes, we need some self-control to be able to not play the games while we work, but because we're waiting for agents so much when we're using the app. Sometimes it's good to have a game to play. Usuz is talking about our app that we built and said, "Did you use rubber duck debugging with the rubber duck game?" We did use rubber duck debugging. And if you haven't heard about rubber duck debugging, this is like automata automatic debugging that happens um when the app is if like there's a big change that the agent wants to make to the app. Salami is saying this is really helpful. Thank you. No worries. So let's see what has happened. So say so for example we just ran that automation and with this automation we said review the last GitHub issues and proposed priorities and owners. So you can see these are my highest priority issues and we also already have assigned you know suggested owners who should be the one to fix this issue. So we have the these are very high priority that I need to assign right away. These are medium priority. We have low priority as well. What I can do as well is I can go to what can we do? We can actually open this as a session. And what I'm going to also do today is I'm going to show you something which is a brand new feature that's very cool in this app. And I can say create canvas. And what this is going to do is going to help me work with the agent to be able to see the agent's progress when I'm doing something. So say for example, I think this is this very high priority issue. It's the number one priority issue that the agent is seeing. So what I want is for the agent I want to say um create canvas make a branch on the repo and then um what we want to do is we want to fix uh and then create a PR no create the code in this new branch to fix the issue 714 and we will review first and um and then create a PR. I want to see exactly what you are doing when you are fixing the issue. So make a dashboard. So I can see what is being done and understand the code changes. So what what is nice about this create canvas feature? This is a new feature I'm going to show you just now that we just run. So what we did is we started by running an automation to be able to get our top priority issues. After we ran the automation, the next thing that we're doing is we're going to create a canvas to be able to have that canvas. Then in real time, we want to see exactly what the agent is updating and for us to be able to understand which changes it's making to fix an issue. And for this, I'm using create canvas, which is a way for you to interact or see what an agent is doing without you actually having to um o only read the text or you know just type things out. Create canvas actually creates for you like an app an inapp thing um where you can see a dashboard or things like that of what the agent is doing. So you can see here it's saying it's setting up the dashboard canvas and then we will get a look of exactly what the agent is going to do to help us with uh with this task. Another thing I really like is that you can see all of the things all of the prompts that I've run in this session. So you can see that I've asked to review and you can see this create canvas one. And we're going to get back to the game example just now in a second. But I just want to show you the canvas when it's created. While we're waiting, maybe let's go back to the chat and see what everyone is saying. Uh for the the game mana is saying, "How about changing the duck to a hen and after certain points evolve it ancestor T-Rex breathing fire?" Maybe we should do that. Maybe a bit later we're going to do that. I'll save that as a potential prompt. Um, we can see this is a great question. Muhammad is asking normally in the prompt you schedule your doing automation. I would like to understand what does it do in addition. So okay wait before we answer that question here is our canvas that has been created by our agent. And basically what the canvas is, it's a place for us to see in real time what the agent is doing. So that's one of the ways we can use the canvas. There's a number of different ways you can use it, but we can see here we want to see exactly what the agent is going to do to fix the issue. And so the agent on its own has created this nice dashboard where we can see its progress as it fixes this issue. So we can see it said this is the issue it's fixing. It's telling me that there's six tasks that it wants to complete and these are the different tasks. It's in progress right now. Is doing this task and then it's going to run these other file these other tasks as well. It's also got this session where it's going to show me which files it's changed. And all of this is happening in real time. We can see the agent is going to be updating us as it completes the task. So, it's actually pretty much almost already started. Um, you know, it's already completed these three tasks here. And so, this is a really nice way instead of just reading what your agent is doing in the chat, you can use this canvas to actually see the progress of what your agent is doing in this way here. So, that's another way to use the command that I used for this was create canvas. So this is something that I would recommend that if you are someone that is using agents a lot to have this kind of interactive feature. Let's actually see if it's going to run finish this in the time that we're here. We can see right now it's writing unit tests for the new methods and then it's going to run the llinter and the tests and it's already showed us these are the files that have been changed. So really this is a good way for you to be able to understand exactly what your agent has done. We now are running the llinter and the tests. So yeah, one of the things I would say people complain about all the time is not understanding what their agent has done. And so in this case with canvas, we can see the progress our agent has made. It's running the llinter. it's done these things and it's changed these files for us. Um, and you can have this, you can actually save the prompt or the canvas and you can export this canvas if you want other people to be able to access it as well, which is super nice. I think this is going to I wanted to actually finish the task. Um, it's running the test right now and linting, which is what we can see here as the final task. But this is I think very helpful for developers that are wanting to work on different issues with um with AI. Let me go back to the chat to see what people are saying while we wait. So um Sirly. Okay. No, no, no. Let me go to the newer one uh to see. I love this comment from Surirly Deb. We're saying you should also have a boss button key combination you can hit that immediately hides the game. Even if the agents are busy and you can't work on anything, the boss won't understand. So, you put a static image of a spreadsheet or some code or something perhaps with a spinning uh or busy cursor on top. I love that. That's a great suggestion. Maybe we're going to do that next. Igor is saying, "It's always interesting to see how the guys test their new tool on the fly without polish presentation, but with real what if questions. I'm looking forward to funny bugs and unexpected features." Yes, this is going to be all of this is happening real time. So, if this is not going to work, then you're going to see that it's not going to work. But this again, this is perfect here. All of the all of the things were done. We get a good summary. So basically what I did is I told the agent create a canvas for me to see in real time how you're fixing this issue. I also told it to fix the issue in a different branch. And later on we're going to submit a PR for this if we're happy with it. And you can see this is really a nice way to interact with the agent. We can actually see the activity log of what the agent did and which files it changed as well. So this is really good in terms of like automating some of the development process for you but also still understanding what your agent is doing. Arthur is saying very interesting demo. Copilot really changes the approach to daily development and automated task. It does. It's it's very cool. I can just I can recommend if you want a new way to test things out. So that was one way we can use this feature that I wanted to show you called create canvas. Another thing is that if we go back to our game. So, let's do this. How do I go? How do I go back to the game? Okay. Okay. Okay. Good. Let's go here. And if we go back to our duck game, what we had done last time is that we want to improve our game. So, we want to add different things to make our game more interesting and we want to use all of the features that Copilot offers us to be able to do that. Do we have some suggestions to in terms of what we should add to the game? And then I also think we want to create the game as a canvas as well. So, one of the things that we can do, um, I like the game how it is so far. I'm going to go and show you, but I think I want to add like a maze. I want it to look like, you know, the I want there to be like a bit of a maze where it can look like a Tetris sort of game. So, let's go ahead and tell the agent, can you uh can you add a maze so it looks more interesting? Um, and then let's also say, let's say create, let's use that same create canvas. Uh, wait. Let's use the same create canvas command. Wait, I don't know if it's going to work. Create canvas. Ow. Hang on. I've now forgotten how you Wait, wait, wait, wait, wait. Let's see. Maybe you have to start from zero and then do that. Yeah, you've got to start with nothing on this screen. And then we're going to create canvas. I'm gonna say, "Can you add a maze so it looks more interesting? Show me what you are updating in the canvas." So, we're going to have our agent use a canvas and then we are going to try to make the game more interesting because right now I like the game as it is, but we want to be able to I just feel like right now I don't like that there's just the black screen. I want a maze so we can see. So, it's more interesting if the duck is going to be Oh, what's that? What's that? Oh my god. Okay, I saw a new thing, but it's okay. It's okay. It's fine. We'll come back. We'll come back. So, it's going to create a canvas so we can see exactly what are the changes. Let's see what other people are saying. Arthur is saying, "Awesome session. Thanks for sharing. No worries." So, Nam is using Copilot CLI, which I also love. Copit CLI is fantastic. I can recommend using that as well if you want to, but really it's up to you what you want to use. Sirly dev is saying copilot in Excel is fantastic at comparing two tabs of similar data and producing a column that has details the differences between them. For instance, if you run a report multiple times a day, I love that idea. And you can actually probably make that an automation and have that run to compare the changes in an Excel file. So if someone's doing something related to data, you can definitely do that. Um Kesh is saying great co-pilot. I'm happy that you like copilot. Pos is saying get up copilot is great with your explanation. Thank you. I'm glad we're enjoying the explanation. Wow. What do we think about the maze? Do we like the maze that's on the screen? I do. We feel like this is a good maze. I don't think I like it. Wait, hang on. I don't know if I like this. Wait. Uh, okay. Maze. What's a h what's happening? We got to figure out what's happening right now. It's saying click cells to toggle wools. The agent updates this as it edits the game. Okay. So, wait. I think the agent misunderstood what I'm asking for because it's saying now let me read the current game file to understand the maze integration points but build the canvas. Now I have a clear picture. Let me build the maze editor canvas that visually shows the 20x 20 grid. Let's the agent update wall placements and also update the game file. I'll implement it all in the extension. Extension running. The maze is open. Now, let me say it. A fun maze layout. You should see the maze. Now, let me I don't know if I like that. Wait, it's saying okay wool soap duck start. The duck is going to start here. Let's see if it's also updating the actual game. That's one thing I don't know if it's doing. Okay, it is. It is updating it. So, what's happening here? I think I'm not sure but I think what's happening is that it's updating in the background to say exactly where it's going to put the different parts of the maze and then it's also and then so so that we can see exactly what it's doing in the background here but at the same time it's also updated here. So, um, oh, so our maze in the real app or show is showing. And the maze is actually going to be Oh, wait. Where did the maze go? I feel like it didn't show the maze now. Hang on. Let's see. I don't know why it's reverted some of the changes. Maybe it's because I needed to update. But what we want basically, so it's saying to us that the made editor canvas is open in the side panel showing the 20x20 grid layout. You can see blue cells is the soap bar walls the duck must avoid. The golden circle is the position of the duck and the dark cells are the open space. The maze has a symmetrical pattern with wall clusters in the corners and centering. You can click the cells in the editor to toggle uh the walls on or off. Okay, so I can actually change what the walls would look like. Refresh the game browser to see the maze in action. I actually thought that was very cool. So instead of like a proper maze, we have blue cells are soap bars that the rubber duck needs to avoid. So I think this is really nice actually. Um hang on before I continue. Let me check to see what does everyone think of these changes. Um, yeah, exactly. Certainly is like it's amazing. I agree. It's it's it's very cool. Sasha was saying it's amaze editor. Yes. So, okay. So, this is why it it it understood what I was saying. it was saying, "Okay, this is going to be an easy way for you to actually edit the maze that it's going to build in the game." So, this is a backend sort of way for me to interact with the app in a fun way. And I think this is actually very cool. So, you can see if we go back to the game, this is a new the actual maze stuff is new. So, we've got to I do think there's too many things because as we know, I'm not a fantastic gamer and so I want to reduce the amount of maze stuff around. So, I don't think we need this stuff. Let's reduce these so that we have it a bit clearer and it's a bit easier for me to maybe win the game because I'm not a fantastic gamer. So, we're going to remove those. And then if we reload, we should see those changes taking effect, which it it is clearer. It is still a lot of maze. Oh, no. I don't know if it updated. Did it update? Because that's still a lot of maze. I feel like that's still a lot of maze. Ah, okay. I'm not going to win because there's there's too much maze going on right now and it's going to be difficult. What are we saying? Um, so Pio just is saying Claude Opus is too good for development but with unlimited tokens. Yes, the Opus models. What models is everyone using? Like do we like the Opus models? Do we feel like we like the GPT models? One of the things that a lot of people are talking about right now, I'm interested to hear what everyone thinks, but a lot of people are talking right now about how, you know, it's it's quite expensive these days to um it's quite expensive to use the, you know, agents now or different LLMs. The price has gone up. we know with tokenbased billing and so it's probably a good idea to be a bit flexible with the type of model that we're going to use and so I could actually try on the stream to see if we can use a different model and the models cost different prices so you don't have to use claude opus for example because that's a very expensive model to use but you could also switch this model out and you could different models. Like for example, I really like GBT uh 5. I think the GBT 5 series is very good. And for most use cases, it should just work even if you're not using the latest model. So, it just depends what models is everyone using, what do we want to try? Maybe we could I'm going to hide your comment and let's try and make a different change, but let's switch out the model instead of Opus. Uh so auto is also another way to save money because when you're using auto, it's going to use the best um model for the use case and the most like the cheapest one for the use case as well. We can try and see if we can use let's try and use GPT5 mini which is one of the cheapest models. available today. And let's see if we can say, can you change the duck? So, it's a uh princess duck. I like I love princesses. So, let's see if GBT 5 Mini is able to do the job. This is a very cheap model. And so, if it's able to do the job, we'll see that. Okay, this is a great alternative for us to be able to use. Um, let's see what other people are saying. Okay. Okay. I see uh a n develop n developer is saying Maya to see how it works. Maya would be great. Have we heard about the Maya models? The Maya models are from Microsoft. These are new um and very interesting to try. I can see. I'm actually not sure if I have access to Maya models here. Let me check if I do. I think I do, but let me check. Let's see. Maya. Okay, there is Maya code flash. So, that's a model that's very cheap. It's built by Microsoft as well. And so, it's going to be very affordable for us to use. So, let's do the Maya model next. Let's just see what GBT 5 Mini did. Uh, right now, let's reload and see. Do we like Okay. Yes, it made. Okay, let's reload. This is the princess. Did we see the princess one? okay. I'm going to try because I don't want it to. It's still too many mazes for me right now. I feel like it hasn't updated, but I don't love the duck that GPT5 Minute Mini created. Let's ask the new Maya Flash model to create a better duck. Can you improve the duck? So, it's a magical princess duck cuz the one that the vaia model and also you know what I also wanted to wait when uh princess duck when you are done please show it here on in the chat as well. Okay. So, we want it to try again. I personally don't love the duck that the GPT 5 Mini created. It's okay. It tried, but I don't love it. Um, I also feel like the maze editor is not working for me because the maze editor is I need to update it and see like how it can actually affect the game because I feel like it's not it's not affecting the game enough. But we're going to try out the Maya models. We're going to try out Maya code one flash which is supposed to be a model. Let me reduce this uh I mean let me remove the comments so you can see um here you can see that I'm now going to try Maya code one flash and this is a brand new model from uh GitHub and we see it here. It is supposed to be very good at coding. So, if you've used Claude Haiku, that's a very affordable model and it's also very good at coding. And this model should be as good. Let's see what it did for the duck. Uh if it created a better one. Okay. Um I don't think that's the duck that it created because that looks about the same. Done. Duck. Uh refresh the preview tab. Uh, I don't feel like that looks like a duck to me. Let's say, can you show the SVG here because I can't see it in the game and I want to see how what it looks like. Uh, Tana is saying that GBT 5 is not good in my opinion. It's still in 2025. I don't know if that's true. Do we does do other people feel like that's true? I don't think that's true. Maybe it is. Maybe it is. Toana. I'm not really sure though. But let me know. Do who's using GBT5? Like it doesn't have to be like mini, but it could be any of the other ones. I'd love to know what your experience is. Do you like GBT5 versus uh the clawed models? It really depends on the user this is giving me. Okay, let's say create Show me an SVG of the duck in the canvas. So the problem is for some reason it's not wanting to render the duck in SVG here in the chat. So we are going to use the create canvas feature to be able to see the duck that's being created by the different models um so that we can compare. So again right now I'm using Maya code one flash and the main thing is that we want to compare to see do we like the Maya models. I like uh I've heard very good things. They are very coste effective. So, if you are someone that is worried about the cost of your models, you can try out this new model to see if it's a going to be a good fit for you. Um, again, uh, I did see Arthur saying, "Copilot CLI is an absolute lifesaver for my daily CIS admin scripting. Using Copilot for PowerShell and Bash gave saves me so much time as a CIS admin. Copilot chat makes decoding legacy script so much easier. Exactly. If you have stuff in your job that is stuff that's not super creative, I feel like this is the perfect thing for Copilot to be able to do for you. It's going to, you know, take a lot of the work out of having to manually do things on your own. Okay. Wow. I'm going to take this off so we can see. This is what This is our princess duck. I just don't know. Do we feel I just feel like this is this is cute. I'm not saying it's not cute. I like the duck to be fair. I like the duck. I don't know if this is this would be necessarily what I would make for the duck as a princess duck, but it's pretty decent. Let's see if we comp. I I don't know if we should compare it, though. Maybe let's try auto. Can you let's try auto and say can you update the duck and make it better and show me in the So one thing that I really like I know I know someone said uh Mr. Bob says a snowman that thinks it's a duck. To be honest, this is a snowman. I'm so sorry, but this is a snowman. I really love I really love ducks, but it's not it's not great. And then Sir De saying, "Is that crown upside down?" It is upside down. The crown is upside down. I love it though. It's not bad. It looks kind of pointy and painful. Yeah, I'm not saying it's the duck is bad. To be fair, this is not bad. Then Surirly De is also saying a snowman with yellow snow fe snowmen with yellow snow. Yes, not a good idea, but I just I just also feel like this is kind of giving snowmen right now, but it tried and it's not terrible. I feel like this is it first go. It also started from what GBT Mini gave us and I didn't feel like GBT 5 Mini gave us a good starting point. So, my thing is, okay, okay, we have some improvement with auto. Are you seeing this? Oh my gosh, I love that. Can you see it's now sparkling? I really like that. Wow. I like the sparkles, but the problem is the crown is upside down. I wonder why the crown is upside down. It's just like a bit weird that it is upside down. Let's change it. say, "Can you update the crown so it's the right way up?" I feel like that would be good if the crown could be the right way up. But generally speaking, I think it's decent. I think that the Maya model did a good job and I think it's pretty decent in terms of like the general vibe of the duck. The only thing is the crown was upside down, but I actually like the sparkles. So, I think this is I don't know how are we rating this this model. This was the Maya model that came up with this and I think it's pretty decent. It's just Okay, there we go. Now, our crown is the right way up and it's actually not bad. I actually think this is pretty decent and I I would use that. So, we are hearing Tana is saying GBT models are perfect for academic research. Claude is perfect for programming. I think it depends on the use case and what the user prefers. What do you prefer when you're coding? Do you agree with Tana? Let me know in the chat. Do you like GBT models for coding or just for academic research? Is this just a vibe that we are feeling? What do we think here? Uh P is saying I think everything is good but depends on our smart prompt because I made share market forecasting within an hour and it works perfectly. I I'm assuming that's with GBT5 or some of the GBT models. So I do think that the DPT models are very good. It just yes I would agree with P push here because it's one of those things where I do think you need to experiment with the prompting and I also think that using things the different uh features like if we go to if we go back I want us to auto go back to Maya code one flash and I want Maya to help us plan so we can say or or another thing that I will say that I've been seeing is let me go ahead and remove the comment. Something I've been seeing people doing and saying that this works effectively is to start out by planning with one of the more expensive models. So maybe you'll plan with Claude Opus 4.7 or maybe you'll plan with something like GBT5 and then after you have a plan you go ahead and switch to a cheaper model. So, for example, let's say what do we want to do? Let's say let's uh make the princess duck uh what should we do? Let's make the princess duck a swan and add more sparkle. And then what we can do is we can get Opus 4.7 to actually write a plan for us. And the plan should be as detailed as possible. And then once we have the plan from the more expensive model, we can then switch over to a cheaper model and have the cheaper model actually go ahead and write things for us. So, while this is going, we're almost at time for our stream today. I do want to mention, like I said, I would like to recommend checking out go ahead and check out automations. This is probably my favorite feature. I I actually went ahead and I got a full list of some priorities, weekly priorities that I need to do. I won't actually show all of these here because I don't want to get into trouble with work. Um, so I'm not going to show all of those, but for example, you can connect to your Slack or to your teams and see your weekly priorities. You can go ahead and do issue triaging and have it priority prioritize your different issues. You can also go ahead and uh see those in a canvas. But I would definitely recommend going ahead downloading the app, browsing the different uh templates that are going to be available. You can even get cost tips, a repo audit like we said last time, different things here, performance audit as well. So this is with automations. And then after that, I would also recommend going ahead and trying out the app yourself. Like in our case here, okay, we've got a full plan. uh, okay, we said suggest changes to plan, edit plan mode. Um, let's say approve the plan, but uh, let me choose the model. I don't know if it's gonna because I want to say I want to use that plan and I want to Oh, okay. It's opened it. So, I've chosen the model, I think. What is happening? Waiting on you to review the plan. Okay, let's say approve and then we're going to switch over. Uh how do we do this? Uh we've approved the plan. We can see the plan here. Um and what we want is for the Maya model to run. Okay, let's say take the approved plan and use that to implement the changes. So I think that what what people were saying is helpful in this case is that you can get your model to look at a plan that was created from one of these the more expensive models and then to actually implement the plan and to implement the features. You can then get a cheaper model like the Maya model to be able to then actually implement those changes on your behalf. So I really like that and I think it actually works out to have a much better um result. So this is the GitHub copilot app. Um this is probably we are at time now. So this is all the time we have for our session today. Does anyone have any last minute questions? If you don't, one of the things that I would recommend I will go back and I will put the link to the to the app so you can try it yourself. Go ahead and give the GitHub copilot app a try uh at home wherever you are. I would also recommend something else that I would recommend is joining the session later on today. We have a stream that is going to be the GitHub engineering team, the some people from the VS Code team. Cassidy will be on the stream and a number of different people will be there. So go ahead and join that stream and that will also be a deeper dive into Copilot app. I'd also recommend trying it out yourself. Wow. Can you see what was it? Can you see the magical swan that was created? It's amazing. But thank you so much everyone for joining me on this stream today. I hope you had fun. I really have fun chatting you to you as well. So uh I will hopefully see you next week and yeah, it was great to see you everyone. Have a good day. Bye.