Rubber Duck Thursday!

Hey, hey, hey. Heat. Heat. N. Hey y'all. Welcome to Rob Thursday. How are you doing? How are you doing? Welcome to the stream. Welcome to the stream. It's growing. Yeah, I know. I love we love using that little ducky thing. That just gross. But hello everyone. Welcome to Robuck Thursday. If this is your first time joining, I'm Cadesa and I work here at GitHub as a developer advocate. And on Robot Thursdays, we come together and we kind of just learn and vibe and answer questions and build and just hang out for a little bit. just hang out for a little bit. Um, so super happy to be here with you. I'm going to switch off switch the music as usual over. Let me know how you're doing. How are you doing? How is it going? Um, how h how has your week been? I know for me it's been a long week. It's been a long week and you know, we're almost we're almost there to the finish line, but let me know how you're doing in the comments. Let me know what you've been working on and let me know what you want us to do today. Right. So last week, if you missed last week's stream, it was quite interesting because we tested out the auto model selector feature in Copilot CLI. And it was it was quite interesting. There were a lot of learnings and a lot of lessons. So I hope that we can kind of discover the same thing today. Yeah, I see Amy says, "Hit that like button, people. Hit that like button, people. Make sure you're following GitHub on all platforms. Um, we do post quite a bit of content every single day and we do these streams weekly. And tomorrow we have an open source Friday stream right here on the channel and I'll be chatting with the maintainer of Remotion. If you don't know Remotion, Remotion is the open- source project that kind of exploded in popularity this year wherein because of the ability of agents to do things, you know, on their own, people and and uh Remotion launched an MCP server. So, people connect to the remote MCP server with their favorite coding agent and they're just they've just been like building um videos programmatically. I tried it out for myself and it's it's actually pretty good and it's only going to get better. So, I'm super excited to talk to the maintainer of promotion tomorrow. So, if you want to see that conversation, be sure to join in. I'll be scheduling everything after this stream. But yeah, the week has definitely flown by. Amanda Amy, I I completely agree. Um, but we're here again and so let's hang out. So, I usually like to start these by with going through by going through the GitHub change log. So, let's just let's do that. Let's do that. Um, let me just pull up my I'm going to pop you guys over here. Whoop. That way I can still see the conversation that's going. All righty. and then I'm going to share my screen here. Let's actually go into full screen mode. But yeah, let me know how you doing, what you've been working on, how has the week been. I feel like things are just moving so fast and it's hard to to kind of keep up here. Okay, so Satie says, "Is this space for non-coders too? I'm only doing code through clot code." Sorry everyone. All love, all love. And I said, "Sady, I meant Soldi." Um, all love soldi welcome to the stream. Thank you for joining. Non-coders, welcome. Okay, if you're working in cloud code, that means you're doing all like not all but some of the amazing things that we're able to do in Copilot CLI. So, you are welcome. Let me know what you're building. What are you building? What are you working on? What have you been learning? How has it been going? Tell us stuff in the comments. Let's chat. Um, what was I doing again? Oh, yeah. I was I was about to share my screen. See, my brain is just like mishmash mish mash mish mosh. Let's see. Let's see what other what else we have in here. It's been a long week. So, true. some scared too with GitHub actions. You know, the team is working on it. We're working very very hard. Okay, like I said, it's been a long week. We're working very hard on all the things. So, never fear. What are we building today? That is a great question. What would you like us to build? you know init I was I'm just going to go through the change log and see if we see anything and try it out otherwise there is this experimental feature in copilot CLI called um rubber duck uh where you can uh plan with one model or like implement something with one model and then have another model review it and it actually is pretty interesting like it's it's such an interesting concept and I really like it so if anything we may be able to explore that let me know that's of interest to you. Um, I thought that's pretty pretty cool. So, welcome vibe coder. Vibe coders are always welcomed here. So, I've been trying to build a Roblox game with my six-year-old son. Oh, incredible. What are you using to build? Let me let me know. Okay, so for real now, for real, for real, let me share my screen. That's gonna pop you down here. And then let's open up another. Where' my tab go? Oh, right. Okay. Boop. Boop. Boop. Boop. Boop. All righty. And then I always like to switch over to this little one here so you can see the ducks. All right. That's cool or not. We can we can look at it and see. Hopefully it's not the you know like what happened last week. You remember last week with the auto model selection. So hopefully it will be a little better than that. But I mean that was fun, right? That was a lot of a lot of learnings. Okay. So I'm on the GitHub blog and I think I'm super zoomed in here. Let me just zoom out a little bit. If you go on the GitHub blog, right, I want you guys to go on the GitHub blog. It's github.blog. All right. And we have so much good content here. CK outlit bloggy. We have so much great content here that you can check out. Right. So, we've been running a GitHub Copilot CLI for beginners. So, like if you're brand new to Copilot CLI, we have a full series for you, blog posts, you know, and videos that you can follow along and learn how to use Copilot CLI. You will see Akla's beautiful face here. And then you also see my beautiful face here on on the GitHub blog where we have GitHub for beginners. If you're brand new to GitHub and you don't know how to use it, we have a full video series for you to learn from videos and blogs, you know, and all that jazz that you can uh and some projects for you to do as well in your learning. So, definitely go to the GitHub blog. We always do updates here. You can see we have some company news, some security updates, um just like all the latest and greatest things. This is where you'll go to find the news and we have an engineering section, skills section, enterprise section, lots of goodies, AI and ML section of course, lots and lots of goodies on the GitHub blog. All right, so make sure you check out the GitHub blog, but from the blog, you can click the change log. And this is where I always like to start our stream because we ship every single day, multiple times a day, and it's hard to kind of digest what's going on. So, let's see what's new. So, this is today. So, GitHub Copilot and Visual Studio. Oh, so some updates have have been pushed to the Visual Studio platform for GitHub Copilot. So, we have cloud agent integration. Awesome. So you can now use the Copilot cloud agent on github.com in Visual Studio. You can build your own custom agents. You can use agent skills and so many more goodies. So if you're a Visual Studio user, this is really good news for you. Copilot 5.3 CODS has been removed from the model picker. We all knew this was coming. So this is officially retired. We have some improvements to the copilot cloud agents where it starts 20 20% faster with actions customs images. Interesting. Get a copilot code review will start consuming GitHub actions minutes on June 1st 2026. This is important to look at. Okay. Important stuff. Okay. Lots of updates happening. Notice about upcoming new format for installation tokens. Okay. is if you have a GitHub app I would encourage you to look at this recent release 5.5 is now in generally available for GitHub copilot that is interesting like in and um very relevant for us especially if my vibe coders are here if you're vibe coding or if you're build or for vibe vibe engineering you know like you're building with aentic AI tools have you tried GPT 5.5 I think this is one of the things I really love and adore for about Copilot CLI is that I have access to all the frontier models. Like I have access to the latest and greatest models in one coding agent. So like I don't have to, you know, go to different tools to try them out. I can just try it out here. So I think today we're going to give this a go. And I mean this was shipped maybe like six days ago, so it's still pretty recent. It went into GA six days ago. So let's try out GPT 5.5. What should I build with GPT 5.5? I have a project that I've been using on these streams. Uh, so we could probably just still use that project for simplicity sake, but um, let me know. And note that this model is launching with a 7.5x premium request multiplier, you know, because models are not free, folks. you know, like we have to make it sustainable because right now it's such a core part of our engineering environment. It's it's important to make sure that it's sustainable. But 5.5 is available in Visual Studio Code, Visual Studio, Copil CLI, all these places. It's it will be surface for you right there. So, I think we can test out 5.5. I think that sounds really good to me. What should we do with it? Uh, let me see what the comments are saying here. All righty. So, mixing my Okay, so this person, the Stoke Bean, is telling us what they're building. So, they're mixing their systems engineering experience to map out what they what they want, then making clot code for me. Lao. Then I go through and test the game and make changes to the code as I need. It's rough. So my question for you is I love that you're using your system engineering experience to map out what you want. I think um in the mapping before you even start coding, right? It's important to make sure that you plan what your experience is and what your user's experience is and what the user's journey is going to be in your I think you said you're building a Roblox Roblox game. You want to make sure that you you do as much planning as possible because that's going to make it less rough for you because even during the planning you can say okay what skills am I going to need to give my coding agent to make sure that it's able to build this Roblox scheme for me in the most efficient and the most accurate way right so I would you know think about what does your environment look like what what type of skills are you using are using an MCP server are you using any custom agents? Like there's so much configuration you can do to your environment before you start building something. Um, and so I would take a look at that. But that sounds that sounds cool though because you learn the most when it's rough. Honestly, when there's a lot of tears, that's when you're going to learn learn a lot. Okay, so Sullie says, "Going to watch your video ASAP." That's exactly what I was looking for. Awesome, Sie. Welcome to the GitHub fam. We love to have it, you know, we love to hear it. So yes, definitely check out Get Up for Beginners and Copilot for Beginners on our YouTube channel. It is free education for you and the content is excellent. Okay, it's excellent. It's very good. Okay, so I like this question. Companies ask for GitHub, but most real work is proprietary. Yep. How do we showcase skills when they can't share employer code? So, Joseph, I would encourage you to build personal projects. So, you can create your own portfolio site to host your personal projects on and that is one way to showcase the skills that you can do. So, whatever you're doing at work, build a project that kind of reflects what you're doing, the the skills that you have at the job, right? Like for example, something that I recently built that's been so great is I built an actual AI receptionist for my brother and it was overengineered. But why was it overengineered? Because I was getting into vector databases. I was learning about vector databases. I was learning about you know going deeper on rag and you know building these systems out. And so I I overengineered it to kind of showcase my skills and also to improve my learning. So personal projects is how you're going to be able to show an employer the skills that you have without showing proprietary code because you never want to do that. So Ryan said GPT in my opinion is the weaker model. Oh Ryan them fighting words. Why do you say that? Let me know because I I tested 5.5 a little bit. Not like a lot, but just like in a I was having the biggest bug the other day. Um what was it again? I forgot what it was. Um but I think like my environments were I was Oh, I was using the same um database for Okay, don't judge me too hard. Okay, don't judge me too hard. But I was using the same database for staging and local. And because of that whenever I you know went through the authentication flow um I had like so many different type of type of users um with the same name the same email. So my permissions were getting mixed up because one user didn't have the permissions that it needed. The other user had the permissions that it needed but then somehow some users were deleted. Anyway, it was a whole mess. Okay. And I actually went to 5.5. I went to GPT 5.5 and I connected um my database through the through MCP and I you know I gave it all the context and I said hey here's what's going on. I have no idea why when I log in I don't have the permissions I'm supposed to have. And within like I maybe like going back and forth maybe like 15 minutes, GPT 5.5 helped me to realize that, you know, I was lazy in how I configured my environment, you know, because I didn't want to do like three databases. Um, but I mean that's literally what I need to do. I need to have one for local, one for staging that mirrors production and then you know um in order to do it properly. I was just kind of being lazy and it and I and I suffered because I I just couldn't figure it out. But but having a conversation with 5.5, it was the fastest resolution I've been able to have with a model because I find OP plus 4.6 to be quite verbose at times. Um, and even 4.7, I really haven't been using 4.7, but 5.5, it was able to get it like that. So, Ryan, I'm curious what your experience has been like. lots of great comments here. 5.5 is crazy. It's ability to make crazy UI. Oh, let's test it out. Who has time for personal projects? That's so real. No one. But you know, with with agents by your side, Joseph, with agents by your side, you can make it possible. Uh or you can like, you know, post on LinkedIn about what you're doing without showing proprietary code. LinkedIn is great. Okay, LinkedIn gives you a lot of exposure and also like the GitHub community, you know, just like getting involved. All right, so let's test out 5.5. And if there's time, I want to show you that rubber duck that rubber duck feature that I was talking about because it's already like 24 minutes into the stream, which is crazy. Okay, so I've been thinking about ways to fit it into my unit testing, but I'm still unsure. current agents still hallucinate sometimes. Yes. Um yes, hallucinations will still happen sometimes, but I think that's why like um MCP servers uh are really good for giving your agents the context that it needs. Even it's not perfect, right? Like not all of this is so new. Like none of it is going to be perfect, but I find that, you know, giving it giving the agent the context that it needs on what I'm asking it to do kind of helps with the elucination a bit. But yeah, it does like to make stuff up and then say I'm you are absolutely right. I did do that, you know, just don't do it. Just don't do it. And also, you know, like your agent skills on MD file, your instructions files, your, you know, like your global like you all those all those things are there to help to lessen hallucination. Okay. So, this is good high quality content. Thank you. I'm happy. I'm happy that you think so. Sharing the GitHub copilot CLA with my team. We're in the process of deploying GitHub Copilot orwad and I'm super excited. I am super excited for you. Be sure to check out uh Copilot CLI for beginners and go on our blog. We have some really really good content on there. And if you ever need anything, um, always come to Robuck Thursday. You always get, you know, have like full access to us right here and then we can always send whatever feedback to the team, send us DMs. We're here to help you. Okay. Um, so I hope that process goes well. Love the fact that you built a virtual receptionist. I know, Puja. Also, hey girl. Um, it was such a fun build. It's currently in production. It's actually in production um, at a mechanic shop. So that was that's pretty cool. But I learned so much during the build, right? And so I think um I I hear you Joseph as in who has time for personal projects, but honestly it's like the the current uh tech ecosystem and this current engineering ecosystem is so fun. Um I try to just make time for personal projects, but also I do I learn for a living, right? So like I I do this for a living. So like I feel like I I I I can find time for personal projects probably more than a straight sweet can. Okay. Wow, you guys are really chatty today. Okay, let's go. So, okay, Soldi, let's talk about this. Let's talk about this. So, I just tried GPD 5.5 yesterday for making a website. It was great, but I still don't understand what GitHub is used for. In the middle of building through plot or codeex or copilot CLI solely, you know, get into get into it. Give it a try. Give it a try. Give it a try. Um, I've been mainly building agents in CLA, but just trying to understand how and when I would use GitHub. I was sharing skills through Google Drive with a friend, but obviously I would like to help businesses as well. Okay, so Si's question folks, and please help her in the comments. What is the purpose of GitHub in all of this agentic stuff? And I think in your in your instance right here so where like you're building a website and you're like what is the purpose of GitHub? So the purpose of GitHub is to make sure that you don't lose your work. Okay. So imagine you build out this website or you build out these this beautiful things for your client and you know like you tell your agent, hey make this one change and then it changed too many things and now you don't have any historical reference of your work to go back to. That is why we're here at GitHub. Like that's one of the reasons right in in your instance that's one of the reasons why we're here. So GitHub allows you to save your work. Think of it like um the the Microsoft Word doc for code. So because of GitHub, you'll be you'll be able to have a historical reference of your work and you can go back to the the next best version of your code that worked and that functioned um very seamlessly. Definitely check out GitHub for beginners, watch all the videos, read all the blog posts and I promise you like you will you will get into it. you will understand so much better um about what the purpose of GitHub is in all of this. So we just don't want to we you you don't want to lose your work. So learn how to use it. Some really good questions here. Okay, so I really like GitHub agentic workflows and I saw that you can now bring bring your own key, but I can't find anything in the docs and I'm now having some issues configuring the secrets. Do you maybe have more information? You know, Mara, I actually just made a video about bring your own keys in um Copilot CLI. So, let's take a look at it. I'm just going to go based on what you guys need, right? So, let's take a look at how to bring your O keys in. And are you talking about bringing your own keys in copilot CLI or bringing your own keys in the GitHub agentic workflows? Just to clarify, if it's the bring your own keys in GitHub agentic workflows, I would have to take a pause and and look into that because I haven't tried it yet. But if it's bring your own keys in copilot CLI, I actually just did something about that. Okay, so we have a list. Somebody keep a running list. So we're going to look at GPT 5.5 and we're going to look at bringing your own keys to Copilot CLI. That's a lot to look at. So let's get into it right now. Okay. And so I want to show you Copilot CLI. Okay. Okay. Let's go. Okay, I would love for you to give it a go. Primarily because you have access to so many models in Copilot CLI and as you just heard, you can also bring your own um local models to Copilot CLI. So, it's just like it's worth the squeeze, you know, like it's it's worth it. All right, so you should be able to see my uh my terminal here. Let me zoom in a little bit more. So, I have this project here that I've been, you know, I I have not been working on it. I So, I can't even say that I've been working on that I've been meaning to work on, right? It's called Chapter Smith and essentially it allows people who make long form videos to upload it and get chapters like those YouTube chapters automatically. Nothing crazy happening, right? And so I've been using this project to kind of show you the ropes around Copilot CLI. And so we have some features here that we can implement. So I kind of want to go in and see if we can implement this chapter editing UI. Remember last week we attempted to do it with the auto model selector. I nuked all that code. I I got I got rid of it because it's, you know, it didn't it didn't turn out the way that we wanted it to. So, let's try GPT 5.5 to see how it would implement this feature, right? So, I'm just going to copy this here. I'm not going to do any crazy like prompting or configuring all that jazz. I'm on the security review branch, but I'm just going to stay there for now. I like to go into the co-pilot d-allow all mode or the yolo mode or the vibe coding mode. Uh because sometimes I just want don't want to say accept accept except good good you know I just wanted to go go forth and do my voice isn't coming. Oh no. Are you here? If I put my mic closer you can hear me chef. If I put my mic closer can you hear me? Let me know. Hey Jeremiah, you're my Rubik sunshine. My only sunshine. Um, love that. That's a great song. Let me know if you can hear me. I think my mic seems to be on. Um, I can't hold it like this throughout the stream, so I just have it right here. I'll try not to go all the way back here. Okay. So, let's try GPT 5.5. And as you see, I already have it enabled here. Oh, let me let me take that down. Welcome, Blessing. Welcome to the stream. We're going to be building something. We're going to test out GPC 5.5. So, I hope you stay and see what it has to offer. And we're using Copilot CLI because Copilot CLI is awesome because it has all the models, all the best models, and you can even bring your own model, you know. So that's cool. Okay. So I have 5.5 here, right? But if I go to slashmodel, you can see that I have so much more than just one model. We have auto mode which automatically selects the model for you based on thanks. We have Sonic 4.6, 4.5, Haiku 4.5, Opus 4.7. We have these hidden models. Don't don't worry about them, okay? Just pretend you didn't see those. We have Opus 4.6. So, we have GPT 5.5, 5.4, Codex 5.3. Uh, so we have a lot of models here. Oh. Oh, and I just noticed that there are some models that I'm not seeing, but that's besides the question. But you can see that we have GPD 5.5 and we have claw models and boom. Right. So, I'm going to stick to 5.5 here. Medium. Let's Let's do that again. I I kind of want to do high thinking just even though high thinking is going to take some time. So, let's just stick with medium. Uh, and I'm going to say I'm going to go into plan mode. So, how I do that is I do shift tab on a Mac here and I'm going to say help me create a plan to and I'm going to paste. Oh, wait. I lost the thing I copied here. Help me create to create a plan to and I'm going to say like let users edit blah blah blah, right? To let users edit timestamp, titles, and descriptions before exporting. So, let's see what this experience is like with 5.5. And what I can do, let's not do that. Never mind. Scratch that. Pretend you didn't hear me thinking out loud. Okay, so the usual process is happening. It's going to search through the project to see what the current flow is to determine what the plan should look like, right? It's going to it's reviewing the improvement MD file, which is the file I have here that has some improvements that I can make to the application. And let's actually run it. Uh, I'll just run it from the terminal. She didn't carry over. Oh, she did carry over. Okay, let's do an install. And I know I pro I need to, you know, switch from npm probably. I I'm so used to using it, you know. Okay, so the current code keeps chapters in appearance state but renders each chapter as read only text. So now it's asking me, okay, what do you want to do? So, what do you want the editing UI style to be? Inline fields, always visible. Click edit per chapter, then save, cancel, single edit mode for I kind of want it to be like inline time stamp validation behavior. Automatize impossible. What should happen when a user enters an invalid or out of order timestamp? Ah, let's just do that. Should edited chapters persist beyond the current generated results page? Just the current session. Always show an editable option optional description field. These are such great questions. I I really like plan mode for that because it it it thinks about things that I'm not thinking about too deeply, especially when I'm doing these types of um demo type streams. And so I I really like that. So if you're not using plan mode to plan, get into it. It's very good. Okay, it's planning getting something there. Let's run the project. I probably need to do some updates to stuff. Um, I need to update my dependencies. Are you guys still here? Oh, you're probably just like looking at what I'm doing. I'm like, why did all the talking stop? I'm happy you're paying attention. Oh, I love this. Someone says, "I always look forward to rubber duck Thursday." So glad, Sherifa, thank you for joining. We love doing it. Oh, well, that's good to know. Thank you for letting letting me know this. We're going to I'm going to see if I can look into that and bring it up to the the social team. Thank you. We are listening attentively. I see. I see. Okay. So, let me go back. Let me go back and pay attention then since you are paying attention. So, what happened here? Right. So, let's let's take a look at it. So, I know it asked me all those questions. Then it went in and it created the file plan. Um, and then it tell me that the plan is ready for review. So that's essentially it was it was just like you know creating the plan. So the plan is to add current session in line keep the chapter smmith app as the canonical chapter state owner excuse me normalize valid timestamp dates preserve existing copy export behavior update change log and validate with existing lint build commands. Awesome. So we can accept the plan and build. We can exit and do it ourselves. I'm just going to accept the plan and build. Okay, we're going to we're going to just do that. Let's see. And I remember last week when we were using uh I believe it was one of the high cool models in auto model mode, like auto mode. It took forever. It took forever to build something that didn't work. So, let's see what the output of 5.5 will give us here. So it's first checking existing validation baseline then it will make the smallest set of state type UI changes. Okay. So it's scoping its work to what we're asking it to do. request fail. So, right now it's implementing edits. are there any files? I don't think so. good job. Okay, so so far I'm liking how focused the model is so far. So like I remember last week when we tried to implement this feature, Haiku went ahead and did things that I didn't ask it to do. So, I asked it to implement the edit um the edit feature, but then it saw some error and it was just like, "Oh, let me just fix this real quick." And I was like, "What are you doing? I didn't ask you to do that." And here you see GPC 515 says, um, the baseline lens, you know, that there's an existing issue, but I'm just going to keep my fixes scoped to the edited chapter files and not touch anything else. Hello. That's what I want. That's what I want to be done. Please focus. Focus. Focus on the work. So, I already like that it's doing that, right? It's keep it's it's being very focused on what the um the goal is. Okay. Yeah. See y'all are really locked in. Let me lock in too. Okay. okay. Um, while it's doing that, let me see if the app is running. H. I keep doing that. It should be. And I should have a Yeah, I should have a thingy here. You can just open. Ah, you open on this screen even though I clicked over there. Let's just do that. So, this is Tractor Smith. You know, it's I like it cuz I use it. Um, so for me it's useful and that's what I love about even like a coding agentically. It's like I'm able to just build my own personal tools and I love it. So, we're just waiting. What do you do while you're waiting for your agent to build? Um, I have a bad habit of I won't tell you what I do. What do you do? What do you do when you're waiting for your agent to build? Let me know. let's see. So, this is the chapter list timestamp time stamp error second. So, this just look like standard stuff, right? trim testing parts numeric parts. Okay, we deleted the export button from here. We removed the copy to clipboard button. Why? I'm just looking at what's what's been done. So, you change or maybe like improve the error messaging to say fail to process video specifically. All right, you removed you guys. Why is the copy functionality being removed? Is it like refactoring? Yeah. Yeah. Alex, like these are this is not what I expected at all. And you know, this is why it's important to review the work that your agent is doing. Um I am a huge fan of vibe coding. Okay, the biggest the biggest fan. I love it. But you have to review the work. And in order to review the work, you kind of have to know a little bit of a little bit a little bit of a little bit of something. Okay. Why? Why? Well, let me see. Let me see the reasoning. Let me see if it's change log is updated. I'll run the existent build. So, I'm going to come in here. I'm just going to say, tell me why you remove the copy copy button and also the export button. I don't know if it removed it or if it like put it somewhere else. It has a change log entry that I can review. So whenever it makes a change, I tell it to um explicitly up update a change log file with all the work that it did. It's also going to like spit out everything it did in in our UI here in compile CLI. I'm just going to send that here. And while it's building, you can actually send it messages. You'll see a message that pops up that says pending. And when it finishes its current, I guess like work or whatever it's doing, the process is doing here, it's going to like surface that message that you send. Yeah, I think so too, Alex. Okay, so it triggered the TypeScript reviewer skill. um which is awesome. It's going to review the chapter. So, it's doing its own review. So, there's some TypeScript failure that's happening. So, it's trying to get that resolved and so it's invoked the skill to do that. I think that's one of the reasons I love skills because it's literally empowering your agent to do the work that it needs to do without you constantly having to say, "Oh, here's some context. Oh, here's how you do this. Oh, here's some information. This is still pending cuz it's still working on resolving here. Okay, so that's 5.5. Um, I didn't really look at all the changes it did. So, let me go back to the editor. Okay, so it moved the copy button from like here and it put it in its own copy button. Okay. So, it just created its own thing for it. I mean, I kind of agree because now I can import it everywhere else. Sure. Sure. I think that was a good choice. I think that was a good choice. It was a good call. What do you think? I think it was a good call. So, I didn't finish like going through all of it before sending my Why did you do this message? Okay. So, I think it made a good call in regards to like creating, you know, that new file there. So, here's what it did. Added in that editing timestamp edits are normalized. Copy and text export buttons are disabled until time stamp issues are fixed. What user prompt help me create a plan to let users da Okay, sir. This is still pending, but let's see what it says. It's still going. So Claude So Claude is not doing the code review. So what you're seeing here is everything Claude code is a I think it's a mark. Hold on, let me let me tell you. I think it's a marketplace. So everything Claude Code is this incredible repository of skills and agents and custom agents and plugins and oh my goodness all the things that I came across one day on on GitHub. And so because with copilot CLI you're able to use um you're able to use configurations from other agents. So like this everything cloud code was specifically made for cloud code but because copila is able to use those settings as well. Um it's able to also use those skills. This is the repository that I'm talking about the everything quadcode one. It has so much goodies and so that's what I installed in Copilot CLI. And so whenever it's trying to do something like reviewing TypeScript errors, it will like trigger one of the skills that's in there because a TypeScript reviewer is a skill that comes from the everything cla code. I guess it's a is it a marketplace? Um I'm still getting up, you know, with the terminology of all this stuff, but that's that's what was happening there. So it wasn't that Claude Cod was reviewing it. It's a skill that came from this repo that was reviewing it. So Copilot was reviewing it, you know. Okay. So it looks like it's finished it work. So it says I didn't intend to remove the copy or export UI. What I removed was the old parent level copy plumbing on copy chapter on copy all and the unused copy state variables in chapter smith but I didn't ask you to do that you know those are redundant because copy button already performs the clipboard copy itself the visible button should still exist one thing I did change if your intent is to add the full export chapters model then I should wire the da interesting. So, it answered my question. Let's see if it completed the build. So if I do boop boop. unexpected token. I internal s Do I have my I should have my keys in here. Um, hold on. I should have my API key set up here, but Cloud's going to look, not cloud, but compiler is going to look at that. Let's check my environment here. Yeah, both keys are here. That should be fine. Done. That was my exact thought, but I didn't say it. Bye, Sie. Thank you for joining. Internal server error, you guys. Now it's not loading at all. Do you ever miss like the old way of engineering? You know, like the old way. Um It could have been the URL. Exactly. And that's the beauty of GitHub branching and you know version control and all this all this stuff. But let's see if we can get somewhere. Let's see if we can get somewhere. It's crazy because like whenever I'm doing whenever I'm working on my own, um I don't run into these issues that I do live. so I always find it interesting. Maybe uh maybe copilot doesn't like to be put on camera, you know. So, what I'm seeing is um where did it first start? manifest.js JS. Okay. Yeah, but now we have a server error. Dude, hold on. Let me kill all sometimes that alone will help. Honestly, Brian, um, so like it's it was built a while ago. I really haven't been actively I'm I'm not actively working on it. I use it to do demos and, you know, like all this stuff. It's just a great um it's just a great project for me to pull for whenever I need to do something like this like the stream thing. let's see if it works. I got a new link. literally has you. So, I need to update the validation here. Let's see. Let's choose one of the ones I have here already. Uh, okay. So, copy. Where's edit? If I click in line, I'm supposed to edit. So again, the edit isn't it's not editing, you guys. It's not editing. You guys. You guys, you Nice. So, the lesson here is don't rush through things. um verify the plan that you get from the agent and you know make sure that you're reviewing the work that it's doing and if it does something you didn't ask it to do tell it to stop so it can remain focused. Usually if I'm building something real, you know, I have like I still create issues and I do one issue at a time on like different branches and all that stuff, but maybe I need to implement my actual maybe I need to do my actual workflow on streams instead of vibe coding because I find it's it just takes so much longer. read chapters in speculi. Yeah, this is still the same copy stuff. Uh, I mean, I literally did reload the app right before we tried that. Okay. So, here we go. So, now I'm able to just do stuff here. And if I do export all chapters, do I get all of them? I do like that I can copy one by one. But let's see. I'm just going to pop it at the bottom of the read me here. I do I do get all chapters. So, this is like, you know, those YouTube chapters you see. This is kind of what it looks like. So, yeah. And the edit works. So, how do we feel about 5.5? I don't I don't know. I I find it a little like like I did it, but I didn't finish it and then I touched something else that he didn't ask me to do. Um, but that could be user error, right? That could be me not taking the time to do it properly, right? Like to do my usual flow of like here's all the things you need to do. It's sculpted out properly. Um, instead of saying like, you know, do this, plan, and then implement it, go forth and run, I think it's important to take our times and do things whenever we're doing things. But yeah, what do you guys think? I think there's a way for me to see that actually. That's a really good question. Um, Ryan, So, if I do slash usage, it's here. So, did I use that much? Honestly, you guys, I feel myself like shutting down a little bit. but yeah, what do we think of 5.5? Um, I didn't get to show you the bring your own key stuff, but we do have we do have some docs on it. Um, let's see. Bring your own key. Yeah, we do have some pretty good docs on it. Um, you can use OpenAI models, Azure, Azure Foundry stuff, Anthropic, O Lama, which goes into those local LLM models, and it has instructions on how to do it for each one. So, you'll need an API key. You need to know your provider. It's all here. Oh, what is this? Wait. Oh, I lost the ability to comment, folks. I planning mode may not work as well as just incrementally building and integrating features. So, yeah, I think like usually I'll use planning mode and then I say, "Okay, let's break down the plan into into sculpt issues." And then I'll like create issues and then I'll like tackle one issue at a time just so I do like one section at a time. It takes a little bit longer but I get a better output that way. So it's like yeah incrementally building and integrating. I don't see an increase in productivity compared to previous models, you know, but I I I don't know if it's like user like maybe I wasn't being the most efficient. You know what I mean? Yeah. Like some setup is needed to make sure it doesn't steer into a fence. Exactly. Okay. Found it for CLI but still don't find the docks for agentic flows. We'll surface that to the team. let's see. GitHub, aentic workflows. Um, by just going to do a quick Google search here. One second. Yeah, you are correct my dear. I cannot find anything on that but yeah the it is model agnostic. Let's go pass CLI. If I GitHub aentic workflows. Okay. This is a thingy from GitHub. Quick start contributing guard wheels. See how it did. You go to this repository. Have you seen this repo? Um, Marorrow, it is a very new feature. Yeah, but have you seen this one? Quick start. How it works? All right, let me see if I can say something. Documentation. It's not what I want. Oh, let's check the FAQs capabilities. Guard rails. Can I use MPCB servers? Can I can I use um Okay. Oh, what do I need? Why do I need a token or key? So you configure it as a GitHub access secret like in your settings where you know you can configure secrets for actions. Have you tried that one? Have you tried that? You are correct that we do need um clearer docs and probably like demos on how to do bring your own keys into the GitHub aentic workflows. got you. Yeah. Yeah. I I I I'll bring this back to the team so we can, you know, get into it and get on some docs. I would send you the repo link, but for some reason I'm not able to add comments anymore. So the repository is gh- aw for GitHub agentic workflows. So it's github gh.io/ gh-- aw copy link address. So, let me see if I can pop it into a um a banner. Can I still add a banner? I cannot add a banner. Um Oh, I can add banner. Okay, so that's the repository that you can go to for GitHub agentic workflows. And yeah, Maro, thank you. That was that's a really good flag. I will send that note back to the team so we can um get into it and add better documentation so that you can you know add your own key to GitHub agentic workflows. But I believe that's all I have for you today folks. what did we learn today? Today we learned that it's important to um take our time when we're building with agents and also to you know like properly set up our environments so that we can get the best results as possible from agents. That's that's my main takeaway, you know, like I I need to bring my actual workflow to the stream and not say, "Oh, yeah, let's let's vibe, but you know, actually build with you properly." Cuz whenever I'm working whenever I'm working on my own, I I do do things differently. I take a more rigid and more like methodical approach. But hopefully you had a good time. I have enjoyed hanging with you. Tomorrow I'll definitely be sending a note over to the docs team um in regards to bring your own keys in GitHub agentic flows. But this has been rubber duck Thursday This has been rubber duck Thursday. Hopefully you enjoyed. Thanks for coming and I will see you next week. Oh, any docs on setting up VS Code for vibe coding? Oh, setting up VS Code for vibe coding. What a good question. So, setting up VS Code for Vibe coding. What would that even look like? Another thing I will send back to the team. Let me actually write this down because my girl will not remember. Let me um find a text. Hold on. Pinner. Pinner. Pinner. Pinner. Okay. Docs for setting up VS Code for by coding and then docs for agentic workflows with um bring your own key. How to set that up? Great question. Any more docs requests? Just kidding. Thank you for hanging with me. I'm glad you enjoyed the session. and I really like hanging with you all. But I will see you next week. Bye. Heat. Heat.