Rubber Duck Thursdays!

Hello. Hello everyone. Welcome to today's live stream in the Rubber Duck Thursdays. It's incredible to see so many people already here with us on the chat. I am super happy to see all of you join me today. So for those who are regular viewers of this show, welcome back. Um it's so nice to have you. Uh please drop a quick hi on the chat. Let us know where you're joining us from and uh you can also let us know what you are looking forward to learning today. Uh but for those who are just joining us for the first time, so this is the first rubber duck Thursday session that you're joining. First off, welcome to the show. So happy to have you. And then second, you can also introduce yourself on the chat. Just let us know what you're working on. And what we do here on this live stream is basically have a conversation around GitHub. So we talk about what's new, what's exciting, what's coming up, what's not working, you know, what should be improved, what can be improved. So we talk about GitHub as a whole platform. So GitHub is not just GitHub copilot. It might feel that way because of all the new announcements that you see go out about GitHub copilot, but I promise you there's so much to it. GitHub is a very big platform with so many offerings. So you can just let us know on this chat if you have something specific you'd want to talk about, something specific you want us to demo. And for our new viewers, what you expect to find in this live stream is we may present this in different formats. We might have a demo. So just go through some working examples of different features or we can just have live Q&A. So you'll see some questions coming in on the chat. I'll attempt to answer and I'll answer them to the best of my ability. For some, we may need to also look in some experts to come help us out, but basically we just have conversations around how GitHub is um basically helping you out as developers. So, I'm looking at the chat and we have people joining in from different platforms. We have Twitch, we have YouTube, we have LinkedIn. Welcome. We have representation from different parts of the world. I see India is very well represented. We have France. Um Amelia is asking what's rubber duck? It's a good question. Um I'll admit I'm new to the team here, but I did find rubber duck as an existing brand. So what I'll actually do is ask my teammates exactly why we chose to call it rubber duck and then I'll let you know on the next stream. But yes, that's basically what we do on this show. Um, yeah, I see we have um Nabil from Egypt. Welcome. Uh, very many users from India. Welcome, welcome, welcome. Let us know also what you'd want to talk about today. Right, I have some ideas of stuff we can do, but I'd want to have this heavily focus on what you as the viewers want to learn. So, we have some some room for flexibility here. All right, welcome Lincoln. We have Pande. Welcome, welcome, welcome. Uh, huh, Amelia is asking, talk to us about rubber duck in the specific mode. Not sure what you mean there. Uh, please feel free to expound a little bit more about the question. Um, yeah, someone someone's asking what's going on. So, we're basically just starting our live stream today and we'll basically be talking about all things GitHub. We'll look at what's new and then we'll just have some demos. And if you have any questions, drop them in the chat. Happy to go through your questions as well. All right. Um, K says that he has many doubts about GitHub. That's valid. Uh, please do let me know what are some of the doubts that you have. I'm happy to clarify what I can and I can reroute you to the relevant teams in case you need further explanation. So please um can let us know what are some of the doubts that you have about um GitHub. Okay. Uh so many people joining. Uh welcome day. First time here. I'm a junior developer. I'm really interested to also hear your perspective about all that we'll talk about here today. So welcome to the show Alumi Day. All right. What's this trip about? Briefly, uh, Yash, I think I mentioned that. So, we'll basically just be talking about what's new in the GitHub world and then I have some ideas of some demos I can show, but again, I'm flexible to tailor that to the questions that we see coming in on chat. So, basically, think of this as an ongoing conversation. We'll get to see what's really going on in the world of of GitHub and demo some of the features that we have already available. All right, fantastic. So, keep the love coming. I see so many people joining. Um, all right. So, someone's asking, no, I'm not usually the host. This is actually my first time here. So, I'm also really excited to connect with the community. And so, what I'll do, I'll just kick off. As I said, we talk about everything that's happening in the world of GitHub. And it's not just GitHub copilot even though it will feel like most of the announcements are about copilot. So what we'll do as the first thing and then we can have a conversation from here is basically look at the change log. So we're going to just look at what's new, what's been announced just this week. And if you go to the change log, let me actually just share this link on the chat with all of you. So if you visit that website that will take you to the GitHub change log and you'll find a summary of all the new releas releases, all the new improvements, what's been retired. You'll get to see all that. Uh but for this session, let's just focus on what's new this week. And as you can see, we have so many new things that just got announced this week. Uh just yesterday, we had up four announcements. So let's just briefly um you know mention and talk about what these features mean and then I'm going to go straight into the questions. I can see a few questions coming in. So let's spend just a few minutes talking through what's been announced this week and then we'll go straight into the questions. So the first one which is an improvement we saw this roll out yesterday just yesterday March 25th is for users who are on Jira. So, Jira is the workspace that you use for your project management and your tooling. GitHub copilot is now having some improvements for the Jira offering. So, as you can see, we'll read through what this announcement is all about, what's basically new. So for those who've been using GitHub Copilot in Jira, we've had some feedback come in and as of yesterday, the integration now provides better guidance including clearer error messages and steps to resolve common configuration issues. So feedback that we got from early adopters was that the usability of copilot in Jira wasn't the best. So just yesterday we had a release that aims to fix this. So if you are using Jira, be sure to try it out. The experience has been improved a little bit. As you can see, you can now choose which AI model you want the coding agent to use for your tasks directly on Jira. So you can see that we have quite a few updates when it comes to the overall experience with Copilot agent um with the copilot coding agent in Jira. So definitely check that out. That should be exciting for users on Jira. Then number two is uh just yesterday I saw this. We have some updates to our privacy statement and terms of service. Again let's look at what that's all about. Okay. So as of yesterday we had an update on our privacy statement. So what does this mean? This basically means that um we've updated the affiliates section to expand the purpose for which we share personal data with GitHub affiliates. So what this basically means is that for you as a user within Copilot free copilot pro or copilot pro plus GitHub um GitHub will use your data to train the AI models unless you opt out. Okay. So, I'll repeat that just so it's clear and I'll actually just read it as it is. Notably, from April 24th, that's next month, we will begin using interaction data. That is specifically the inputs, outputs, code snippets, and associated context from users within these three plans. That's copilot free, pro, and pro plus to train and improve our AI models unless you opt out. So that last part is pretty key. You do have the option to opt out from this. So if you go into your settings, you can easily just opt out and GitHub will not be using your interaction or your associated contextual data. So that's something I wanted to briefly mention. If you're interested in learning more about what this privacy update means, we do have that properly documented on the change log as you can see. So and again we also have a section at the bottom which talks about okay so what's new the terms of service has been there for some time so what exactly is new you can get to read more about that and then what you need to know in terms of what's being used to train the model so of course it's not everything GitHub will still try as much as possible to um to avoid using any personal personal identifiable information to use as m as little of your data as possible. So if you want to really understand what's happening there, then please give this a read. That will be one that can answer most of your questions then. All right, let me just hop into the chat a little bit. See some questions coming in. Okay. All right. Alumday is asking please can you briefly explain the basic step to create GitHub and push. Yeah. So Olumi to get started with um GitHub just go to github.com and then you'll have an interactive step um process to create an account and then from there you'll be able to basically add projects push your code repositories there. But if you want to have like a full explanation on how you can get started with GitHub and Git, there is an intro um series that's on that's on that's on GitHub and I'm trying to get that link for you. Just give me a minute. All right. can't get it right now, but I'll I'll share that link before the end of the stream. It's basically a YouTube series that covers all things introduction to G. So, it should get you started and set up for you to start working with it. So, just remind me on the chat before we leave. I'm happy to pop that link back up. Um, all right. So, I see some questions that I'll probably answer once we get to the demo part of the live stream. So do stay tuned for that. Um I made an account on GitHub first time. So will I get to know about GitHub in detail? Um yes, that's a good question. For this live stream, maybe we won't cover the very basics of GitHub. But as I said earlier, I'll share a link to a YouTube series that you can follow on your own self-paced that will get you up to speed and up to date with everything you need to know to set up successfully on GitHub. All right. Um, David is saying, I'm curious about that keys and secrets would be avoided. Right. So David, I'm assuming you're talking about this second announcement here, the updates to our privacy um statement in terms of use. So yes, to answer your question, yes, your keys, your secrets will be avoided. And this is basically going back to how copilot itself works, right? So if you if you look at the overall architecture of GitHub copilot, there's how you interface with it. So how you you create your prompts copilot itself will gather some context. So think about the files that you have open and anything you choose to pass along as context. All this doesn't get sent into the model directly but rather it goes through a proxy service. Right? So before your prompts leave your IDE or whichever interface you're using it a copilot and before that hits the underlying model there is a proxy service in between and one of the functions is that it's actually going to filter out um such information. So you do not expect your secrets, your keys to basically be used as part of the training data. So David, I hope that clarifies um that question that you had which is a really good one. But again to also go further, I believe that these um release notes should also have some additional context based on your question. All right. Um Kazim says, I'm curious. I think opt out should be the default not the reverse meaning users are forced to opt in. Um valid point Kazim but from my understanding how I've read it is that yes by default you are opted in for those three subscriptions but again as a very first step that you can take is to opt out of that particular um experience. So yes, you do have to go through your settings and then just opt out and select exactly what you'd want to be contributing to their improvement of the models. All right. Um, so yeah, I'll share that link shortly. So there's a question here to know what if we can make huh independent LM that has all the updated features apart being a copiloted. Sorry um Vastava I don't know if I read that correctly but um I'm not understanding your question. Maybe try and refresh it just a little bit just so it's clear. All right. GitHub is good but not open source and not a safe model where you can write some code. Emilia, what do you mean by some code? I'd love to know what exactly you mean by that and then I can help clarifying. All right, David, you're welcome. Hope that's clear. Okay, so let's uh proceed a little bit. I see we have we actually have so many announcements this week. won't get through them all but you can visit the change log to see what's happening. Um so one that's interesting here for enterprise and organizations is that copilot usage metrics now identify active co-pilot coding agent users. So what does this mean? I believe that before you'd be able to get an understanding of who within your organization is using which feature but it it was not possible to tell apart who's using the coding agents specifically. So as of yesterday an improvement was rolled out where right now copilot usage metrics will clearly indicate which users have copilot um coding agent activity. So within your enterprise, within your organization, you will be able to tell um from the API you should be able to tell who has been using the copilot coding agent. And this goes a long way in helping you understand um you know the investment that you made in GitHub copilot. Does it make sense? Are your developers adopting the use of these tools? Is the return on investment worth it? So this is a step further into giving you some more insights on how copilot is basically being adopted within your organization. So that's a good note to have which is pretty pretty interesting. All right so let's see as I said again we won't go through everything but this is quite interesting. Uh there's one here that I'll probably mention. So for those who've been using copilot coding agents, uh I believe this is an experience that is actually quite helpful. So before if you're having a coding agent work on a certain issue, it would open a pull request or you'd have a pull request open and then if you use the at copilot notation here. So if you just at mention copilot and have it fix let's say a particular part of the changes. So for example, you can atmention copilot in a pull request and then have it address a comment or fix something that's failing. But the before experience was that copilot would then go off and create its own pull request with your current pull request as the base, right? So you'd have to be merging a pull request from copilot into your own pull request and then work on merging that into main. So what this announcement does is that right now at copilot we'll directly um we'll directly modify your existing pull request. So we've eliminated that uh eliminated that um additional step of having every art mention of copilot create its own individual pull request. So this will just make the process a bit more efficient. So that's another announcement that I thought to mention briefly. But yeah, as you can see and as I said earlier, we have so many things rolling out. So yeah, always join this stream just to have conversations about what this means. All right, so I'll go back to the chat. I see lots of interactions. Thank you for the questions. Um, Emilia is asking, "I am a several thinker about to create something and don't h Yeah, Emilio, I'm not sure I quite understand your your question. Sorry if I'm missing something." GitHub is the definition of a safe development environment. David, I do agree with you. Um, okay. So, you mean GitHub acting like Claude? It's a good Okay, it's it's an observation. I'll give you that, but I don't think it's accurate. You saying that GitHub acts like cloud thing is these are two separate offerings. GitHub was uh was there way before we got into this AI rush. So the thing is GitHub is a holistic platform with so many offerings. One of them being the availability of different models and different coding agents. So the reason as to why you might feel that way is because co cloud is actually one of the coding agents that's readily available on GitHub. All right. So you can use GitHub but not use these coding agents. But if you choose to use coding agents, we have brought in so many options for you to work with the agents you prefer to work with. Right? And I'm actually right here on VS Code. I love using VS Code. So I can just use this to briefly illustrate what I mean. So if you go on VS Code, you'll see that I myself being a Copilot user and I have a GitHub copilot subscription, I do have access to additional coding agents. So I can use the claude agent, I can use codeex, I can use GitHub copilot. So probably that's what um Alan you mean by having that sentiment around GitHub acting like claude. So it's not really acting like cloud but it's a matter of GitHub being a very big platform with various offerings all for different purposes and then having these coding agents as options that you can basically use as part of your workflows. So, I hope that clarifies um what it is I believe you've you you meant by that comment. All right. Um Okay. So, just let me know if I've answered that question. Okay. Good. Good comments. I save everything in my memory. All right. Okay. Amelia is asking here that if I create some new code, I don't want uh anyone else to see. Amelia, that's possible. You can create a repository and then set it to private. So when you're creating a new repo on GitHub, you can choose the um privacy level. So it can either be a public repo. This way everyone can have access to it or a private repo. So that means you're the only one who will have access to that code. And you can choose to invite people to that repository as you wish. All right. Good. So yeah. So that's it about the change log. So what we'll do is um I had an idea of something we can go over. It's a it's some conversation that I've actually seen um coming up. So yesterday on X, let me just share this. Yesterday on X there was a discussion uh there was actually a discussion around vibe coding uh a vibe coding debate AI hype meets real limits and I'm just going to zoom in a little bit here to make this more visible for everyone. Okay so I hope this is visible and yeah so I actually missed this conversation. It's one I wanted to join but I had a conflicting um meeting but the idea here and I had a chance to go through the conversation here is that people are really concerned that um VIP coding is now uh the effect of VIP coding is now being felt and a similar conversation that I read uh again on X where I spend so much time um is that someone just inherited code from a VIP coder who exited the company and So the experience is that you can clearly tell code that has been written without following the traditional um development practices and frameworks. So code that has just been received from a model and then just made its way somehow to a production environment. So those are some of the conversations that I have been picking a lot on over the internet and I believe that so many of you have seen similar concerns. So I just wanted to um have a conversation about this and I'm curious to hear what everyone else thinks about this and if you just go back just take a step back into the you know the whole idea of vibe coding right so this is an experience where you as the user you have an idea there's something I want to see there's something I want to build and then you craft that in form of a prompt and then you send that into an AI model so in In this case, it doesn't matter which model, which tool, you just send that prompt into the model. The model just basically um generates some code to satisfy the prompt that it gets. And what mostly happens is that you'll get an output, you look at it, and you'll be like, okay, this doesn't really um represent what I wanted. So you end up editing that prompt and then resending it back to the model and then you're kind of stuck in that loop where you have you're getting some AI generated code. You're looking at the output. There's something you like. There's something you don't like. There's some changes you want to see. So it's a loop of you editing your prompts and getting some output from the model. And then at some point you will get to a state where you are happy with the output that you're seeing. So you'll choose to work with whatever tool it is you're building. So that's that's the workflow when you're vibe coding. And I'm not against vibe coding. I just feel like it has its own place. There are scenarios where say you will definitely default to VIP coding. The first scenario being uh whoever is now building this tool doesn't have an engineering background. And we've seen that a lot. Right now we're not just having developers build software. We're having builders. So this is anyone from any background from any discipline just getting their hands on technology that can actually build this software. So for them this is the workflow that makes sense. This is the only workflow that they know, right? Uh don't get me wrong, I also feel like developers also have work um have scenarios that call for vibe coding. So an example that I'll give is at the beginning you saw me going through this page. So this is actually a tool that I use to manage my presentations. So I find myself doing so many presentations in the course of a week and um trying to just prepare for them would need me to work with different powerpoints to have some notes somewhere maybe notepad for my notes. So I decided, hey, what if I just had a tool that will allow me to stay within my flow on VS Code, which is my editor of choice. So I'm on VS Code and I can easily just walk through my presentation. It's planned, it's strategic. So I can easily do that. But something that I'll share with you all is that I have not seen any single line of code that renders this to that renders this experience. I was basically working on a certain on on another project and then I just thought of hey we have AI everywhere right now. So I basically just went on to GitHub copilot CLI and I happened to record this. So I'm going to show you exactly what I did to get this tool that I'm using. So as you can see I basically just switched to GitHub Copilot in the CLI. I switched to autopilot which is one of the modes. So when you're working with copilot, you'll see we have different modes. You can be in plan mode, you can be in the normal agent mode, and then we have this recently announced autopilot mode. And the idea here is that this agent will go off to build whatever you've told it to build without necessarily coming back to you every single step of the way asking for your clarification, for your input. It's going to just go on and build something. So that's exactly what I did. I just gave the agent in autopilot mode a rough idea of the problem that I had and what I envisioned to be a good solution and then I had it build out the entire thing. So it figures out whatever stack it's going to use. Up to this date I've actually never opened that folder. So as you can see here um it finished the first iteration and then I decided to test it. So this is me planning for a talk. I'll add the links that I show and then after that I add some steps. So these are the steps. This is how I want to go through my presentation and then I hit an issue. So fixing that was as simple as just taking a screenshot of the issue, pasting it in copilot and then telling it that hey I'm hitting this issue but before you even fix it ensure that you use I'm giving it an MCP server that will allow it to spin up its own browser. So I'm telling it ensure that you just test this on your own. What I want is a working version. Right? So this is a mode where I don't want copilot to keep asking me, hey Julia, do I do this? Do I use this? Just figure it out. Give me something that's that's working. Um so as you can see here, this is an this is a browser instance that's being operated by Copilot. So no, I'm not the one who was working on this um browser. So this is Copilot. um doing the browser interactions trying to test out and then at last it told me that hey I fixed this this was the issue and then when I came back I had a working tool and the reason why I'm sharing this is and I'll go back to to this diagram is as I'm saying there are scenarios where VIP coding can be allowed in this case this is a local tool I'm just using it on my on my local computer I don't intend on pushing that code anywhere so I can be a bit more lenient and not really care so much about the code, but I'm caring so much about the experience. This is the second time I'm using this tool and it's actually solving a problem for me. So, we do have scenarios for vibe coding. But again, a comment that I actually picked up uh as well as part of that discussion was that using AI to code as a dev and vibe coding aren't the same thing. These are two distinct things and I completely 100% agree with this comment that um if you think of a typical vibe coder this is someone who either doesn't care so much about the code or doesn't even know you know how to approach this coding um this entire coding experience. So for them they are from a nontechnical background they just want to get a working solution but how an actual software developer uses AI those are not the same things right there is a difference and so to sort of try and answer this question and have a contribution to this discussion I like to think of GitHub copilot and how it fits into the tried and tested software development life cycle stages okay so for the developers on the stream. I'm sure that you will definitely um relate to what you're seeing on the screen right now. This is a tried and tested framework where software development is not just about pushing prompts to AI, getting your code and publishing or deploying that. It's about a set of stages that basically need to build on top of each other. Okay? And if we can break this down a little bit, you start with your planning. So you get your ideas on the table, you get in a room with your teammates, you have some sticky notes, you know that has evolved over the years, but the idea is you consolidate your thoughts, your ideas and put together a plan. And then from that plan, you then transition into a design stage where you basically now build designs. You have a visual feel of what your plan looks like if implemented. And again, this design phase also has its own review steps. So you'll have people come in, test out wireframes, test out um designs, and basically give their ideas. So you'll also have some form of loop there just to validate that from the problem you're trying to solve and the plan that we created, do these designs actually meet the solutions that we're looking for? And then assuming your designs pass, then we have the implement stage. Now, this of course is where you get to write your code. This is where you get to build software. But we've seen that VIP coding basically just skips all this and goes straight into having coding agents, having models just build out a version of your idea. I say a version of your idea because in most cases, these prompts that we use with vibe coding are not very um comprehensive. So it's just a small representation of exactly what it is that you wanted to build. And then once you get into this building phase, this implementation stage, you now have a loop between the, you know, our friends that we love from the QA and testing department where they'll come in, they'll ensure that your code is up to their company standards, you have your testing already added. So basically, it's a very thorough process uh before you even get to deployment. Okay. And then once you're in deployment, the next conversation is about maintenance of that code. So what I'll try and do in this stream and we may not cover everything in this one live stream, but we'll of course have I plan on having this as a continuous process. So next week for example in this stream, you can come and pick up from where we we leave off today. But the idea is talking about GitHub copilot but in the sense of how it fits into this particular workflow and um I won't be naive to just imply that this workflow will never change especially right now with uh the evolution of coding agents. This framework might evolve more to accommodate the coding agents but I do believe that it will still remain around the same idea. we will still have this similar workflow be implemented or being relevant for a very long time. So that's how I plan on tailoring this conversation. So again, I'm just going to pause, go to the chat. I see so many comments. So let me try reading through. Okay. Uh how many stars can we get on a public repo on default? H I'm not sure. I'm not sure. There's a limit. I'm not sure if you can go um you can go as as high as you can. I'm not sure to be honest if there is a limit on the number of stars, but it's an interesting question. I mean, if anyone knows probably you can let us know in the chat. Okay. Um, Nabil, I see your question, but please do me a favor and um specify exactly what you mean. So, could you please let us know or tell us what's the biggest advantage or benefit uh that you gain from this new update that isn't available on any other platform. So, I'm sorry I've read through so many updates. If you could just let me know which exact update you're referring to, I can come back to this comment. So your question is the biggest advantage over other platforms. So if you can just clarify that update that you're referring to, I'm happy to give my response on that. All right. So Zane is asking, "How does the auto option in model selection select models when using Copilot?" Okay. Uh that's a good question. I'm just going to show you briefly. So for those who are not familiar with auto option uh when you're working with copilot in this case on VS code we do have the model picker and from this model picker you can see have a variety of models that you have as part of your subscription but the very first option at the top here is the auto model or the auto option. So what this does is that GitHub copilot will receive your prompt. So it's going to you don't preset the model that you want to use but copilot on its own is going to look at your prompt basically analyze what this prompt needs to to be achieved and then based on its understanding of the different strengths and weaknesses of the models you have access to is going to pick the model that is best suited for that task. So again, you're not having to come in and decide I want to use this exact model because a you might have so many options to choose from and then b you don't know which are the strengths, which are the weaknesses for each models and as we know different models perform best at different tasks. We may not have a model that does very well across the board. So in this case, you change the model based on the task that you want to complete. So for example, if you're just doing a coding activity, a coding task, GPT54 for instance has proven to be really really good at writing code. But if you're working with, let's say, designing some um some some UIs, the same model doesn't really do a good job compared to something like Opus 4.6. So if you don't want to have to think about um you know what's the best model for the best for for a certain job then you can switch to auto mode which will analyze your prompt and then based on what you want to achieve is going to match that to the best model. So that's how the auto option works. You're not having to think about which model to select but rather you're going to basically have it pick the model that best suits your prompt. So Zane, let me know if that clarifies your question. Okay. Um, there's a comment here from Mira. How VIP coding can lead you to big tech giants. Will they consider the vibe coders? Um I can attempt to give a response based on you know my own thinking here is that at the end of the day we still need professionals who understand how software works basically if something were to go wrong and you are a vibe coder who's in charge you wouldn't know where to start right so I do believe that there is still a need to have people as part of the team who really can understand who can just pop the hood and look at what's what's happening behind the scenes. So instead of just accepting um suggestions from from AI, you basically need to have people who can understand what's working behind the scenes. Now a joke that I've heard being said is if you can only code when you have access to the internet. I mean, if everything goes off, for some reason, coding agents are not in business anymore, you have no access to models, and you can't produce software, then you may need to re-evaluate your learning approach. So, that's my personal take there. Um, it depends on intent. If you're just building for the sake of having something that's working, then fine, you can take that path. But if your intent is to build and build mastery and have an understanding of how is this piece of software working, what's rendering this experience, how can this be made more efficient, then to some extent you still need to have some fundamental or some understanding of the basic computer science um fundamental. So that's that's my personal take there. Uh but you can let us know what you think. Um TA that was a good comment. What's the demanded skill all developers should learn to fit with the 2030 tech industry? Um, so yeah, that's a good question. Really, really good question. And I believe so many people are having the exact same question I included. So please on the chat, we do have experienced people. So let us know what what your thoughts are here. But um for our first I believe that at some point all developers will have to understand how we coexist with agents. Coding agents from what I see are here to stay. They're already here. They're here to stay. So it's it's up to us to understand how do we want to to define boundaries between what I need to do and what an agent is allowed to do. I've previously seen comments or you know folks on X and other platforms complain that hey this agent went off and deleted my entire database and the first question that comes to mind is why was it able to do that in the first place so towards and what you're describing here is 2030 tech industry I do feel there will be a huge need for professionals to have a very clear understanding of this is the role that we play. For now, it may not be clear, but this is the role that we'll play as humans in the team and this is the role that our agents will have so that we can all coexist to drive the results that we're looking for. So that's that's the first thing that I feel will be really needed and then of course um to some extent soft skills will also be something that you'll have to invest some time in. I mean, if I can work successfully with a bunch of AI agents, then give me a reason to want to work with you as a human collaborator. So, again, that's also a personal um response, but I'm very curious to hear what chat thinks about this. So, yeah, let's all contribute to that conversation. Okay. Um, Amelio says, "I write everything in my visual studio and sometimes I write in txt in offline mode and save everything in memory which is separated from PC and um, full save." All right. So, what you mean is you're working with your agents in VS Code and then you basically duplicate that in an offline mode and save everything. What's the what's the risk there? What are you afraid of without having that offline um that offline memory? Could you maybe just let us know a bit more about that? It's an interesting workflow. Just curious about uh you know what's that risk that you've identified and you're like for me to avoid this risk let me save let me save my work offline in a txt file. Harry Peacock, great question. When you opt out, you're opting out of GitHub sharing your data, not Copilot, Chad, GPT, or Claude. They all share your data unless you're on enterprise plan, right? Okay. So, Har and let me know if I answer your question, if I or if I get it right. Um, number one is when you opt out, you're opting out to GitHub copilot and GitHub using your interaction data to retrain to train the models. Okay. And your comment about that does not apply to chat GPT or claude. Number one, chat GPT itself is not as part is not a GitHub offering. So what I believe you mean is codeex because codex is the other agent that you can access through GitHub but uh from from your question and to help clarify that is as part of your GitHub subscription you get access to codeex and you get access to cloud. So when we say that GitHub will not use your interaction data to train the models, we basically mean those two agents included as part of that GitHub offering. So there's it's we don't mean that there's a certain cloud agent that will have its own terms of service, but it's part of the GitHub copilot subscription that you have. So I hope that makes sense. uh if you're the in the business or enterprise plan, yes, that does not apply to you. Your data is not used for this particular purpose. But again, if you're on free pro or pro plus, you can opt out and this way GitHub will not use your interaction data to train models. Okay. Um good questions, good conversations. Is VIP coder does a VIP coder have a chance of landing a job? Not sure. Maybe, maybe not. Uh depends on what the organization is looking for. But yes, that's that's a very that's a very hot conversation and debate going on right now. Okay. Uh great. entering on age. All right. So, I'm just trying to catch up on the chat. There's so much going on, but this is all good. Good conversations and thank you to everyone who's contributing. So, um basically we have a few minutes remaining. So I'll just summarize the intent that I had and which is to basically separate the first workflow we talked about which is web coding where you just share your prompt have some air generated code and then edit your prompt and you stuck in that loop for some time until you get your desired output. But what we'll be looking at is co-pilots and how it fits into each of these stages. So of course this is not something we'll do in one hour. We'll start today of course and see how far we go. But uh on every live stream I do intend on basically picking up from where we leave off. So with that we can get into the first stage which is um I'll introduce us to our scenario and if you've attended a Microsoft AI tour of late or if you've watched any of the AI tour um content you must have come across the Zava organization. So, Zava, if I can just put it very clearly as uh the new Contoso. So, Zava is just a fictitious organization that we here at Microsoft are using to build demos around. So, you might have come across this uh in some of their talks, some of the conferences and events. And so basically what I want us to do is we're going to see how as an employee observer we can get to interact with copilot and GitHub at different stages. Again this will be grounded in this particular workflow. So we'll start from the very first which is the planning stage. Okay. And so for that I'm going to shift to a different tab. And what I'll show you is right now for Zava again, I'm an employee. I've not figured out my role yet, but all I have is access to our company profile. So just documents describing who we are as server, what we do, our product catalog, etc. And actually how I got to this, as I said, this is a fictitious company that we're using just purely for demo purposes. So that means you can get a bit more creative in terms of how you want to use this scenario. So how I actually got to create these company profile documents, I had Copilot build them for me. Now this is not a typical conventional um use case that I expect most of you to have, but because it's possible, I wanted to just show how I went on to create these documents for the sake of this demo, right? And in the process we'll look at some of the features. So again we're in this first step that is planning. So what I did was as I said you can interact with copilot from different um platforms. You can interact with it directly on github.com. You can interact with it on VS code on visual studio and other idees through this chat interface. But you can also interact with it through the terminal. Right. And I'm just going to open my terminal here. and through the terminal if you have GitHub Copilot CLI installed you can easily just um invoke Copilot on your terminal and then you'll have an agent running purely on your terminal that you can interact with and you can have different engagements as you're used to having with these other different mediums. So, for example, you can select your models right here from the terminal. And I'm just going to expand this to have more room. So, you can pick the model that you want to work with. I'll just stick with Opas 4.6. You can work with different agents. So, you can create agents from the terminal. You can work with different agents. You can switch the mode that you want to interact with copilot. So, for example, I can easily switch to plan mode. I can switch to autopilot what I was talking about before or the generic agent mode. Okay. So in this case what I wanted was I knew that as Microsoft we have so many public facing documents and demos that already reference the Java organization but I didn't have like a profile. So for example I didn't have these documents. I didn't have like a portfolio PowerPoint that summarizes the the products that we have as Zava what we build. So I didn't have all this. So the first step that I did which is part of the planning step that you're talking about is research. In most cases if you're planning a project you're planning to start something you do have some research involved. And for copilot to facilitate that we have a research we have a research option. So right here on copilot cla I can easily use slash research and this is going to run a deep research investigation. It's going to use github search as well as web sources. So I already did this before. So I'm going to switch to resume. The session is called research Zava developer resources and I'm just going to scroll to the very top to show you the exact prompt that I gave it. So as you can see here and again let me just zoom it a bit zoom in a bit more. Yeah. So the very initial um prompt that I gave it is I used this research feature and then I told it to conduct thorough research on Zava which is a fictitious company from Microsoft that we use in developer content. Look for everything you'll find online but I'm giving you some specific resources that I want you to go and look at. So the coding agent went off. It did this thorough research and then so I'm going to scroll down. So it's it's now pulling in context about Zava online. It's looking at this material that I've pointed it to and then it created a research document. And then as you can see now it's giving me some research findings based on the search that it has just finished. And then I'm telling it that hey I want to now generate a company profile packet for Zava. And then I'm going to and and I want you to focus on the research that you've just finished. Okay. So for this profile um packet I want you to create an orchart. I want you to have our service portfolio. I want you to build all these and save them in my M365 tenant. Now I'm also telling it uh because by default agents may not have all the different skills to complete all the different tasks. So for example an agent may not be able to out of the box just split open a PDF and pass the information inside. But you can easily just give it a skill that will teach that agent how it can do that. Okay. So, if you're familiar with the GitHub skills, this is where I'm telling the agent, just figure out what skills you need to complete this task and then grab them from this particular website. And if I can just show you real quick, I know our time is almost up, but I'm glad we had today's conversations. So many insights shared, but if I can just quickly show you, we have this page, the skills, it's skills.sh. SH and I'm going I'm just going to drop this on the chat. So this is a page that has uh let me just zoom out a bit. So this is a page that has a consolidation of different skills. So you can have different agents come on this page and then find skills to do different things. So, if you want an agent to be extremely good at front-end design, there's a skill for that. If you want it to be really good at Azure AI, um, Azure messaging, you name it. Whatever skill you want your agent to basically have, it can grab a skills configuration from this page and then just add that into its own agent configurations. Right? So, if you're yet to have a look at this, you can just browse through and see some of the available skills. Or better yet, you can just point your agent to this website and then ask it, hey, for this task that I'm giving you, figure out the skills that you need and then you already have access to the website. So, you'll figure out how to install them. And so, as a result, I ended up with this skills folder. So if you're using GitHub copilot, this is how you create your skills. Inside agitub folder, you add a skills folder and then for each skill, you basically create a folder for it. So one of the skills that I needed and again this is part of step one. We want to do some research. Okay, we are we are just an employee at Zava. We want to do some research about our organization and then start one of the projects which is building a website. And we've said that we're not taking the conventional vibe coding route, but we want to be as close as possible to the software development life cycle framework. So, one of the skills that I need for for research purposes and for planning purposes is work IQ. Okay. And I'm just going to open this uh skill configuration just so we can all know what work IQ is. If you've not yet tried Work IQ and you basically work with M365, this is something that you can try out because basically Work IQ allows your AI agents to connect to your M365 tenant, right? So for now it has only read only permission so it's a bit safe. Uh so the idea is that you can have let's say copilot CLI go off and check your emails give you a summary of the emails give you a summary of your team's messages read through documents in one drive and that's exactly the skill that we used to build out um these particular profiles that we have for the company. Okay. So I know that's uh that's a lot. We didn't get so much into the demo part of this, but I'm glad that we had a chance to talk through some of the uh the questions that you all had. So, I know we're out of time, but what we'll do in the next live stream, so again, this is rubber duck Thursdays. We do this every Thursday. So, in the next one, we'll basically just pick it up from here. And I'll just give you what I have in mind for this session is that with our company profile documents, we're going to get into a planning mode with copilot and then we're going to ask copilot, hey, we need to build a company website. Then let's build a plan out of it and then we'll go all the way and follow this process. So this is what we'll cover from next time. Um, I hope this was helpful. I hope you learned something. So, I'm just going to again scroll through the chats. Okay, so there's a comment here that programming has come a long way. It used to be done with punch cards, then came guies and with binary and with it binary was converted to programming languages. Now those programming languages have turned to plain English but critical thinking is still what and always and what has always been about that's so true. Um critical thinking as is at the core of it at the end of the day you still need to critically think about a problem that is being solved and how it's being solved. So that's a very good comment. Uh David, thank you so much for sharing and I hope that also goes into ask answering the question that was asked earlier. okay, really good questions and I'm sorry if I missed some of them. I'm just trying to skim through because again I'm being mindful of time. Um, so Perez is saying, "I'm a vibe coder as well, but I always spare time every day to learn about software. I believe you need to have a degree of understanding." And that's that's 100% true. If at this point writing code is not expensive, anyone can do that. So what's valuable is you taking that extra step to basically learn the skills, right? And an analogy that I had earlier was um let's say if I'm getting from point A to B, I might choose to drive to that place. But that doesn't mean that I can't walk and exercise. Okay? So I can easily just choose to walk. That will contribute to me exercising. It's a good choice. But me choosing to drive doesn't mean I'm lazy. Okay? But what I will need to do is if I drive to the office, then I need to carve out some time for exercise. So I need to go that extra mile and realize that exercise is still relevant is still useful get some time for exercise right so it's the same thing Perez right now you can easily get something done with AI so it will have a site ready for you it will have um a certain software working for you but at the end of the day you need to realize if my intention is to basically grow my skills then I do need to carve out time out of my day dedicated to learning So I do agree with the approach you're taking there and I believe that it's what most people are doing right now. You basically play around with AI to stay ahead of this um curve but again you do take some time back and really work on your skills. So that's a very good um feedback. All right. Well, thank you all so much for joining. This has been really exciting. So for next time probably we'll have let's say 15 minutes just dedicated to answering questions before we switch over officially to the demo. And so with that I'll again hand give everyone your day back. So thank you for joining and I hope to see you all on the next live stream. Thank you.