Rubber Duck Thursdays! Let's talk about Microsoft Build!
Chapters8
Cassidy introduces Rubber Duck Thursdays, mentions Build announcements focusing on Copilot app updates, Copilot CLI, VS Code, and related tools.
Cassidy Williams dives into Microsoft Build highlights and demos the GitHub Copilot app, Copilot CLI, work trees, and canvases with live examples and playful demos.
Summary
Cassidy Williams from GitHub recaps the whirlwind of announcements from Microsoft Build, spotlighting the GitHub Copilot app, Copilot CLI updates, and new VS Code and Microsoft Scout developments. She showcases the Copilot app’s UI, its work-tree-based session management, and the playful Mona game, emphasizing how parallel work in work trees can accelerate multi-branch workstreams. Cassidy contrasts the Copilot app’s capabilities with the CLI, noting the Rust-based foundation and the ability to run slash commands and rubber duck model vetting across model families. She explains work trees in practical terms, illustrating how multiple isolated environments prevent cross-session conflicts when running concurrent AI-driven sessions. The stream also touches automations, canvases, and the potential for cloud versus local sessions, plus practical notes about setup, costs, and the current availability for free users. Throughout, Cassidy shares personal impressions, tradeoffs between work trees and traditional branching, and invites audience questions. The stream wraps with pointers to GitHub’s changelog and upcoming Rubber Duck Thursdays episodes, all while inviting feedback and future topics.
Key Takeaways
- GitHub Copilot app combines UI features (repos, PRs, issues) with Copilot CLI capabilities, powered by the Rust SDK, enabling parity with CLI workflows in a graphical interface.
- Work trees enable parallel, isolated sessions within the same repository; each session can run in cloud, local, or a dedicated work-tree directory to avoid cross-session conflicts.
- Rubber Duck is a cross-model vetting feature that compares planning across model families (e.g., Claude, GPT) to identify holes in the plan, improving code quality.
- Canvas sessions extend beyond chat by embedding interactive graphs or boards that can control or reflect agent sessions and even drive workflows.
- Automation features let you automate standups, change logs, or health checks across repos or cloud/local contexts, with configurable schedules and targets.
- Cassidy highlights practical tradeoffs between work trees and traditional branches—work trees excel for parallel work but may incur duplicated dependencies; branches are familiar and sometimes simpler for single-stream work.
Who Is This For?
Essential viewing for developers already exploring GitHub Copilot in GUI and CLI forms, teams experimenting with agent-driven automation, or anyone curious about parallel work with work trees and canvases in real-world projects.
Notable Quotes
"The GitHub Copilot app is a very serious development tool and you should never ever make fun of it."
—Cassidy sets the playful tone while underscoring the app’s capabilities.
"Anything you can do in the CLI, you can do in the app."
—Highlighting parity between Copilot CLI and Copilot app features.
"Work trees are the default and allow multiple sessions to run in parallel without stepping on each other."
—Explains why parallel sessions don’t conflict.
"Rubber duck on this is actually a slash command that dispatches a rubber duck sub-agent across models."
—Describes how Rubber Duck vetting compares model families.
"Canvas sessions interact with your sessions and can control the canvas back and forth."
—Showcases the bidirectional interaction between canvases and agents.
Questions This Video Answers
- How do GitHub Copilot app and Copilot CLI differ in daily workflows?
- What are git work trees and why would you use them for parallel development?
- What is Rubber Duck in GitHub Copilot and how does it vet models across families?
- How can canvases enhance interaction with GitHub Copilot sessions?
- What are best practices for automations in GitHub Copilot to manage change logs and standups?
Microsoft BuildGitHub Copilot appGitHub Copilot CLIRust SDKwork treescanvasesautomationsRubber Duckmodel vettingClaude Fable 5
Full Transcript
Hey, hey, hey. Hello everybody. I'm going to lower this music. Welcome to Rubber Duck Thursdays, which is our uh weekly live stream where we talk about all things open source, building things, things you're working on, things we're working on, and more. My name is Cassidy Williams. I'm on the developer advocacy team here at GitHub and I want to talk about Microsoft Build and whatever you're up to. Um, I'm getting a sty in my eye. I feel like I am stressed. But post post Microsoft Build has been a whole whirlwind. If you didn't see some of the announcements, there were so many cool things.
The main thing that I thought was cool in particular is uh the GitHub Copilot app, which I am biased about. We had sessions about the GitHub Copilot app, the Copilot CLI. There's new updates to VS Code. There's new updates to Microsoft Scout, which is a whole thing that I haven't actually played with myself personally. So many new things, so many exciting things. I don't actually know if there was a list of releases directly from Build. And so I'm going to try to pull it up. But um while I do that, how are y'all doing? How how was your week?
Um, I have not been on this stream in a while because of all of the build prep and all of the things that I was working on uh for the event. And it it was it was a really really cool time. If you want to check out the sessions or uh any of the videos that are on demand, you can. And I'm going to share my screen over here. There we go. Um, and so if you want to check it out, you could go to build.microsoft.com and there's a blog on some of the sessions. So here's all the news and announcements of the things that happened in uh in the event in general.
But yeah, in my portion of the keynote that we had on the stage, it was all about the GitHub co-pilot app, which was very cool and a lot and I'm a square now. There we go. Okay. Um, but in the app, which I am going to pull up so that way you can check out in a second. Um, I'm also not going to run anything just yet because I don't want my computer to chug too much. Um, in the app, um, it looks like an agent window, huh, there's a whole lot of gray lines, and that is what the trendy thing is to be when you have an agent running app.
But what I like about it, in addition to, you know, being able to have quick chats that aren't attached to repos, your repos on the side, the ability to add repos, having an overview of all of the things that you're working on, PRs and issues and stuff, automations where you can have things run hourly, weekly, daily, monthly, or just on demand in the cloud locally. All of these things that you've seen before. This one has a game in it. Look at that. I can drag Mona around and she jumps around. But then she can jump around and play where uh this is something where if you have many agent sessions running uh this is the new like dino run game.
I'm I'm convinced. Uh you can jump around. And I'm going to do this. If you get to one of those branch icons, look, you skip some levels. How fun is that? Anyway, the GitHub Copilot app is a very serious development tool and you should never ever make fun of it. Uh because what other development tool lets you do this? Oh my gosh, please. Yeah. Yeah. I think this is the highest score I've ever had and I'm dying. Well, you know, that was good. What was my final score? Let's find out. 93. Hey. Um, but anyway, the that's a fun thing that's built in to the app.
But, um, if you like using the GitHub Copilot CLI personally at work, whatever, what have you, um, this is basically the powers of the CLI with some extra UI juicy goodness, because uh, it's built with the Rust SDK, the new Rust SDK built on top of the CLI. So, anything you can do in the CLI, you can do in the app. you if you for example want to have slash commands, they have those. You can do all of your usual slash commands that you do in the CLI in the app. And so if that's your workflow that you like, great, you can continue it.
Um, but I like pretty buttons. I I personally have been using the app way more than the CLI the the more it's been developed. It's it's still early days, full disclosure. But what has been particularly nice in addition to the games, in addition to seeing everything, you can, for example, pick models and choose your reasoning efforts. You can choose your different repos in here. Something that I think is very interesting and something that I'm probably going to make a whole video about is the fact that it works with work trees. Um, how do you access the game, Stephanie?
Great question. You grab Mona. See how you click her right here? You grab her, do a little wiggle, and then suddenly it's a game. That is that is literally how you access it. It's built into the app. Um kind of a fun thing. Um very silly but very professional and we love it. But who here has heard of work trees? Have you seen work trees? Work trees are very similar and yet different to branches. And so if you if you need some kind Oops. What was that? Sorry, that was my mouse doing something. Um, if you need an explanation on uh, git work trees, I would love to give it to you.
Long story short, the app defaults to work trees where whenever you, let's just say I wanted to run a session in this repo, I can choose if I want to do a worksheet or the local repository or in the cloud. And so cloud is some cloud sandbox where you could say, okay, I want to run this session and have it just run on github.com. I want to control it from my phone, whatever. um in the local repository, it'll work just wherever your folder is for that repo on your machine, which is typical. That's what you would kind of expect.
The work tree is the default. And git work trees are something where Let me actually I think I pulled up the docs right before this. And let me make sure that I did that correctly. Oh, that was my hotel expenses. One second as I pull that up. Pay no attention to uh my expense report behind the curtain. Lol. Okay. Anyway, I just found my work trees uh docs that I had pulled up. One second. I'm pulling it up and we'll reshare the screen. Haha. Okay. So, um get work trees are actually a thing that we did not make up for this app.
It is a thing that you can do where you have working trees within hence the name within your repository where typically you might for example have a main trunk for your repo and then you make branches then you merge things into your branch and they're all in these things. a work tree. It's like it clones your entire repo into a separate tree into a separate folder everything and works from that. And so the benefit of using work trees, which I'm actually going to paste the docs over here in the chat so that way you can you can see this if your heart desires.
If you want to use work trees in a normal repo, you can. What it does is it creates a whole separate folder of your repo and works completely in that. And so when you are using something like a session manager app or the GitHub copilot app, let's just say I were to run five sessions in the same time in one repo, the sessions won't step on each other. You don't have to think about the merge conflicts or anything because they're all happening in these isolated environments. And then from there, you can then push things. You don't have to stash anything when you want to make a change.
It'll just work from that work tree. Does that make sense? Let me know if you have questions about work trees. I think they are particularly cool. I'm going to pull that back up over here. I once again like when you want to add a work tree, it's very similar to adding a branch, but you could also create a branch within a work tree. It it lets you check out more than one branch at a time if you wanted to do multiple features at once. It's it's a really good way for working in parallel with yourself without having to just constantly branch off of the one thing.
There's pros and cons. I admit I hadn't used I'd worked tree I'd used work trees maybe once before the GitHub copilot app if I'm being honest just because I knew they existed but I personally was just like why wouldn't I just use a branch but uh that's because me as an individual human writing code at once and not necessarily in parallel it didn't really make sense to me but doing work trees with something where you might be running multiple agents at once it makes things work. But once again, there are pros and cons. Let's just say, for example, you have a Python repo where you have to pip install your requirements or or a node driven repo where you have to npm install your dependencies.
That means you have to npm install for every single work tree you have because it creates an actual separate folder for each work tree that you have. where if I were to pull it up right now, and I'm not going to right now just because I don't want to share that on my computer just yet because I don't know what I got in there. Um, when you have that separate folder for each of the features that you're working on, and let's just say you want to test each of those individual features, you have to npm install for example, every single time per folder.
Once again, you don't have any of the conflicts that you might be dealing with, but it is fully parallel work, which means you have fully parallel dependencies and stuff too. So, blessing and a curse. What happens when you have a large project? It might take a while to copy. Yeah, it git git is fast. Git can handle it. It still uses your same.Git folder and configuration underneath the hood. So, that part is cool. It's less like it's less cloning and like copying and more just like directing the work in a different folder. I guess that makes sense.
Um I'm trying to see if there's a nice explanation in the docs, but not really. Um I hope that makes sense. It's a lot. It's a lot. Don't get me wrong. Once again, I like it for parallel work. Typically in my workflows, I'm not doing as much parallel work at any given time in a single repo at a time. So I kind of like using the local copy on my computer rather than a work tree. But it's nice to have the option where when I was building things, for example, for uh for the demo at uh Microsoft Build, I have this repo called Signal Box.
I wonder if I still have it on this computer. One second. I'm going to pull it up. You can see how fast I clone it. Um, where is Signalbox? Right here with Signalbox. I added it to the repo or added it to the app. Here it is. I could say I Let me choose a nice uh Let's try the new MAI model, shall we? List issues an order by priority. How about that? And so I'm going to say we're going to list my issues that exist for this repo and order them by priority. Something that's fun is once again we're doing this with work trees.
When I click on this, it shows that it created this work tree from main and it comes up with a cute name. I like that this one is called special meme. That's great. And so it shows the tokens that's being used. It shows the context window being used. It shows the session spend. It shows everything including the changes which right now there's no uh there's no pushing happening right here. So I'm going to let it pull in the issues for this and uh we'll see what comes up and then we'll work with that. Um Stephanie, I love that you described it this way.
Yeah, work trees are like tabs in a browser. Exactly. You could open the same web page in a different browser, but that means things are cached on the front end per tab. You know that that sort of thing. Yeah. So npm installs for each work tree. Exactly. But right. Exactly. When things are working in parallel completely, it's a really nice way of working without having to deal with the stepping on each other and stashing amongst all of your different changes. Okay, it is. Oh, it's sorting my issues. Okay, we'll see what happens. I'm I'm excited to see.
I wonder if there's any complications with git lfs and work trees. Oh, that's a good question. I had not considered that. That's a great point. Okay, so it has priority high model the product intelligence entities and raven do all of these create insights dashboard all of that. I can sort these by ready for implementation or by last updated. Sure. But once again because you can say I want to work on all of these in parallel in different work trees and just go what I can do instead of just saying okay let's do number 10 then let's do number 11 I'm going to say I want to work on let's see number 20 number 17 number 16 and number 37.
Make a new session for each and keep me updated. So now this session right here that I have has turned into kind of my dashboard for the work that's about to happen. And you can see on the side over here now it's starting to list the issues and it's turning this into like a little folder of things that it's working on. And so slowly but surely it's creating work trees and creating sessions for everything in this particular repo. And so now it's going to it's creating these sessions. It's doing all of these individually. And right now we're just in the starting part, but let's see here.
Okay, they're started with planning mode enabled. Great. And it gives me all these session IDs. And so I can start looking at each of these individually. And let's see, we're seeing starting issue session. That one says work tree created. I'm gathering the issue details from the project and it's it's working. And so it's nice because once again I can say these issues are the ones that I want taken care of. Great. Start go in parallel. I didn't have to do any stashing. I didn't have to think about the state of my project. It just took what it had, made the work tree, and ran with it, which is really nice.
Let me know if you need any more explanations on that. I think this feature is particularly cool. I admit once again, I'm still early in my like agentic development journey where I do a lot of things one by one. I feel like I do a whole lot more on the planning and specification angle and then delegate more of the actual individual coding work to uh agents depending on the models that I'm using and the things that I'm doing. But uh yeah, it saves you from having to wait for the AI to do each individual thing.
Exactly. It because it can do things in parallel, you can let it go. And once again, because these are all happening right now and I'm just kind of watching them go. While it's doing that, I could then go back to my home, look at my work, start work in another repo, grab some tea, or once again, play this game built into the app. Hooray. I love that we have this. It's so fun. Um, it has been it has been a cool thing to be able to work on and I'm excited to see where people take uh take take the app and take and take their projects and being able to just really accelerate how they build because I don't know about you, I'm building way faster than I ever have before and losing games way faster, too.
I think of the work tree like an actual tree and each branch I create is a separate part of that tree. Yeah. Right. Exactly. And so it kind of lets you have a bit more of a green field on your project where let's just say I wanted to completely rewrite one aspect of the front end of a product that I'm working on. Let's just say it would probably be nice to do that in a work tree rather than on a branch so that I can see all that's happening. And let's just say, okay, I don't like issue 17.
It's taking up too much effort or something. I don't know. Oh, look at that. It did a rubber duck. Really quick, I want to explain what a rubber duck is. Um, I'm actually just going to stop one of these because my computer's really loud and chugging and we are indeed live streaming. But rubber ducking is a cool feature that's a part of GitHub copilot in general. And what I like about rubber duck is it consults other models for you where let's just say once again I'm using the new MAI model to check that one out.
But you could do that with GPT. You could do that with with uh any of the cloud models. Whatever you'd like. I'm stopping a bunch of these because my computer is breathing hard. um because you have access to multiple model families with a single GitHub copilot subscription if you'll forgive me sounding like a shill. What's nice about that is you can compare across model families what your plans are like where for example I did a really long planning session for one of my side projects recently where I was truly just saying okay I want to do this I want to do this and I was mostly using clawed model families to do that but then I said could you you call a rubber duck on this that's actually a slash command that you can run where you can do slash and then uh rubber duck slash rubberduck right there where it dispatches a rubber duck sub aent.
And what that does is it looks at the model that you're using and consults a different AI model and says, "Hey, great. I can vet this model with any other model family and let you know like what holes are in the plan. What which ones do you think?" Because if you for example do everything with Claude Haiku, just picking a random one off the top of my head. If you are using that a ton for everything and you switch to like a Claude sonnet, Claude sonnet might be just like, "Yeah, no, that looks pretty good." And leave it at that because they come from the same biases.
But if we have GPT who comes from a completely different model training world, they can then say, "Actually, that's not so great. Let's let's fix that." And so it yes, it always uses a different vendor. It it always uses a different model family. And we've found I wish I had the numbers on me. we found that there's an actual statistically significant improvement when you rubber duck um in in terms of the code that stays in your codebase as a result. Um and so it's it's very very cool. Also, I'm not sharing my audio, but it's really funny.
I have I set custom notification sounds in the app and I'm like hearing things happen. Um my custom notification sound for like, oh, you need to answer this prompt. See this icon right here? It wants me to say like, do I want to approve and implement the plan or not? Um, and let me stop this one. Hey, I don't want you to work on this anymore. I get it. You've served your purpose as a demo. Um, anyway, my custom notification sound is from Zelda where it goes, "Hey, listen." Um, so as as I was talking, I just got a little, "Hey, listen there." But anyway, yeah, it uses different model families to vet plans and find holes in it and it has been a huge improvement to a lot of developers workflows that we've seen on the platform.
I have confidence in the new Microsoft model uh M A as long as the GitHub and VS Codes teams are steering it. It's been interesting to play with so far. My pros for it, I it's so fast. It's great. That being said, I feel like when I've used it, whe to write code specifically, I do have to be really explicit with my prompting. I think it's it is a smaller model. Like for for now, I think I think they actually did a benchmark and it's closer to haiku than it is to an opus. But, uh, for a small model, it was zippy.
I was able to say like, "Oh, could you just change the padding here and adjust this on one of my projects?" And it did it in like two seconds. So, I was very very happy with that. Um, so though once again I work here. I I've got to be biased towards the MAI models. I don't think they're quite ready for like big showtime yet. If you were to ask me about this exact model two years ago, it would blow my mind, but we've moved past that very rapidly in uh our society. Let's see. This is a good question.
What would you prefer? either using a work trees for agents to work on multiple features in parallel or set up a local folder to set up a new branch and work on that instead in parallel because in practice they both set up a different folder. I personally have my comfort zone in branches. Pros and cons of a lot of different things. I think that like it's nice to not have the artifacts of changes in work trees. So, it really just depends on what you're working on. In branches though, like once again, let's just say I have npm installed on main and then make a few branches.
The fact that it keeps those dependencies, yay, great. I have to do like a new reinstall and then might have conflicts when I merge back into main. If I'm on a a branch work trees, it turns it into a whole separate entity. Ultimately, you're working on things and organizing yourself depending on how you feel. I've found that work trees are once again particularly good when I'm doing like more than three things at once. Um, but because I personally am rarely doing more than three things at once, I tend to use branches myself. So th that's that's generally when I prefer one over the other.
If you all have different opinions, I would love to hear them. Again, work trees was were new to me until I understood their purposes in uh Agentic apps. And it is a really interesting way of thinking and working. And I think I think it works well, but it is different. It's it's a it's a different mindset where I could totally see people saying like, why wouldn't you just use a branch? and it's a perfectly valid way of suggesting how to work. I have been using the app a bunch though. I I recently cleared out my automations because I want to start fresh and see which things work well for my workflow.
My manager does like 11 automations a day which is amazing. Um because you can hook up so many different things to your automations. Um, this is just doing like a change log draft as a test that I was running. But you could say, for example, based on the amount of features that I shipped in this particular week, every Monday, I want you to say like write me a change log for all of the issues that I closed or give me like a standup list that I can give to my team saying, "Hey, this is what I worked on this week.
This is what I want to work on next week." Something like that. You could say you want to run it in the cloud or you want to run it locally. You could give it access to different tools. You could say you wanted to work on specific repos. There's a lot of options in here and it's just automations that you can do. And so, uh, also, hey, yes, I am Cassidy. Welcome. Um, the automations are very neat. I have mostly used them for testing and not as much for my daily workflows. And yet, I see some of my teammates thriving on them.
And so what automations would you do? I know that uh for example I I mentioned my manager before. She hooked up the work IQ MCP server and and a bunch of other things where it now like summarizes her email inbox every morning and it directs traffic for different things that happen in Slack and on Teams across messaging and and lets her know like hey you should reply to certain things which I think is a really cool practical workflow. I think for myself that like both would work, but I'd also get nosy. I have I've I've personally found for myself I don't do well with just summaries.
I'd rather just read the whole thing, but uh that's probably not the most efficient thing to do. You could also, for example, hook up the Playright MCP server. I'm making something up. Playright MCP server and say like, "Go to this website every day. Take a screenshot and show me like what does it look like in a given day?" I don't know. Make something up. Um you the workflows don't have to be restricted to just a specific repository. It could be just a job that you want to run that's AIdriven like a cron job but for your own personal tasks which is particularly neat I think.
Um, you could have a change log where it just says checks, hey, or a change log, an automation where it says, are these services up or down today? And it goes to certain websites and just checks, is the service up or down? Something that you need to check often in this agentic era, am I right? Um, so anyway, the the automations are particularly cool. I need to figure out a better way to do them. But it's it's fun to experiment with for sure. I guess you could build yourself a single inbox that links everything you get messages in.
Yes. No, I actually I one of my teammates did a really good job of that because so many of their messaging platforms have an API or an MCP server of sorts. They're able to just pull in all of the data into one spot. And so this can really become like a place for a hub for you to work. Um, a GitHub. But yeah, you can you can do that. There's also a very cool concept called a canvas. That's really neat. And actually, I'm going to I'm going to pull it up in signal box really quick.
It should work. It should work. I am I am hoping this works because I haven't actually done this since uh open the gesture canvas. Okay. I haven't done this since my builds demo, my Microsoft build demo. So, we'll see. Canvases in the app are glorified web pages. Let's be real. They like it is technically under the hood an iframe, but it has really specific APIs for interacting with agents. So, I'm actually going to I'm going to turn off my camera. No, wait. I don't think you can still hear me. Okay. Um, I'm going to allow my camera and we'll see if it works.
And I hope it doesn't break anything. Okay, it's using this webcam. Nice. My computer's getting so hot. What I can do, approve PRs. And it'll work. Confetti. Okay, I'm going to close this because my computer's chugging. Um, let's see. Stop. Stop. Okay. Um, you're able to interact with your session managers via however you want. And so, for example, that was just a gesture canvas where it pulls in it pulls that in. But you can make a canvas that does anything and it can interact and and integrate with sessions. I'm going to stop that work there because we don't need that to happen right now.
But let's just say for example, you made a conbon board canvas. I could do that in here where let's just say I'll make a new session and I'll do slashcreate canvas. Make a canvas that is a conbon board of my existing GitHub issues. And as I drag cards across the swim lanes, it kicks off sessions and actively updates as sessions change. Let's just try it. A canvas can interact with your sessions, can interact with your agents, and control what's going on. It's it's it's like a birectional communication which is really cool. It it was awesome to be able to experiment this with this where let's just say you pull in different MCP servers or something or you have architecture diagrams or something you could say hey create a canvas based on this architecture diagram showing my app and then let me start sessions based on the sections of the app that I want to talk through and you can then control the canvas and the canvas can control the session back back and forth.
It's really interesting. Honestly, was kind of hard for me to wrap my head around it when it was first out because it's so new. But it is really cool to be able to have that as an option where once again, I didn't have to build this. Let me say repo project scope or current user. Let's just do user scope for now. Um, it did that. Hey, listen at me. But, uh, these canvases, you don't have to touch the code for them. They're just made in the app. based on your natural language. And it allows you to interact with your agents beyond just text and chat.
Because in 2026, working with agents should be more than just chat, you know. And so being able to do that and and being able to interact however you want, whether it's via camera, via your microphone, via some kind of graphical interface, it's nice to have that as an option in the app. Good evening, Nadine. Thank you so much for coming by. What's up, One Pablo? Welcome. Completely lost it when you couldn't make it to the finals on March CSS. H I know. I even have this water bottle for the March CSS thing. If if y'all didn't see that, it's okay.
It's It's a tragic story. But um and also, while this is running, I'm going to hide my screen for a second just to save some bandwidth. But anyway, um I was in a video series called Marchmad CSS run by syntax.fm where it was a bracket tournament for live coding in CSS. If you want to look it up, Marchmad CSS, check it right there. I know that's backwards, but you you get it. Marchmad CSS um on syntax.fm. they they had a bracket where people would live code CSS uh based on a prompt and they would like have percentage matches and I lost by such a small percentage it was painful and then the team kept like releasing graphics being like look how close this was and I was like I know literally I think it was less than.1% difference between myself and the person that beat me uh Josh Kumo and then he ended up winning the whole tournament so I'm a little less salty now but It was uh it was heartbreaking.
Um let me see if I can pull it up. I I think I have that. Uh madcss.com. Let me let me pull that up while we're uh organizing everything. Um let's see. Okay, the graphics are not up. So, it is uh madcss.com, but I've got to pull it up on the syntax. Channel. So, please hold while I do that and then I'll share my screen again. But it was a very fun tournament to be a part of. I have already asked to come back and do it again because it was really, really fun. Um, and really just good content in general.
But, uh, when you see the stats, h devastating. I'm pulling it up. Look at that. In case you can't see, that's me at number four. Listen, I was and I was against Josh who was number three right there. Literally less than 0.1% difference in our live coding CSS prowess. H devastating. I just had to be up against them. But anyway, it was a very fun tournament to be a part of. Ah, and also a devastating one. Anyway, do you all have questions about things? Microsoft builds, GitHub, anything like that. Nearly a rounding error. I know, right?
uh painful painful but uh once again we've been uh we've been doing so much at GitHub shipping so many things and I we're only going to be shipping faster and faster and faster. I'm excited about some of the things coming out soon. Um some some of the things that we can see in general. If you Thank you for welcoming me back. I appreciate it. I didn't stream here for like three weeks and it feels it feels so long. I missed you all and I almost didn't stream today because I'm getting a sty in my eye and it's driving me crazy.
But anyway, if you ever want to see what GitHub is releasing at any given point, you could go to github.blogchangelog. and I'm going to drop that in the chat there. On the change log, you can see the latest things that have come out. And even today, we have 1, two, three, four releases where aentic workflows no longer need a personal access token. Ooh, these are now in public preview. That's amazing. Where you can kick off GitHub actions with Markdown. This one is really cool because it has like guard rails built in to how uh like security accesses actions and everything.
It's it's particularly cool. Um, oh, the set slash settings in the GitHub copilot CLI is really nice where you're able to configure all of your preferences in one spot by doing slash settings. Once again, all of this that you see is also available in the GitHub copilot app. It's just different surface depending on what you want to do. You can use VS Code, you can use the agents window there, you could use the GitHub Copilot app, you can use the terminal. there's there's so many different options and surfaces for you which can be overwhelming but and and I do acknowledge that it can be quite overwhelming but at the same time it's nice to have a choice where depending on the given day or what I'm working on sometimes I'm just like I'm just going to open the CLI run a thing close it and move on sometimes I want to have a more like I want to talk to some agents and have it all in the GitHub copilot app and then sometimes I'm just in my IDE and in VS Code I'm just doing plain old autocomplete as I'm typing.
You have choices and that is particularly nice. Just Oh, I feel bad for not having signed up for Copilot, but because they've gone into pausing. I do understand that. Sorry about that. That pause will be lifted very soon if it hasn't already. I thought it I thought it was lifted. I can look into that. But anyway, you should be able to if you don't sign up if you can't sign up now, you can very soon. Um, but anyway, so many things. Oh, do managing discussions in the CLI I thought was nice. Really, just if you want to live in the CLI and be your Lee tacker self in the CLI, it's pretty nice to be able to do that.
Um, let's see, what other releases have we had this week? There's been so many. Um, CodeQL has been getting tons of improvements lately for just security in general. Um, let's see, there was new dependabot version updates. Oh yeah, Claude Fable 5 is now in GitHub Copilot. I actually haven't used that one yet. I'm a little nervous. Um, but I've heard good things, so I got to I got to check that one out. Um, lots of enterprise stuff. Some old models have been deprecated. It it has been a busy time. This is this is all just in the past week of updates and there's even more to come.
I because especially like business-wise, it's the end of the fiscal year coming up. So, all of the companies you know and love will have plenty of things releasing and stuff to catch up on in general. Okay, I am noting the time. What questions do you all have, if any, in the next few minutes before we sign off for the day. If you have any questions about what happened with Microsoft Build or anything, let me know. Oh, does the co-pilot app support hooks for creation and tear down of work trees? Yes. Um, yes. the end. You can have it do auto clean up.
Let me actually pull it up really quick. One second. let Oh, I forgot to share my screen again. So, see how we have like all all of these different work trees that we created earlier. First of all, you could just delete it. When you delete the session, it deletes the work tree. When the work is done, it deletes that. So, you can do it manually. Oh, actually, let me show you that popup. This deletes it deletes the session and removes its work tree. The committed work on its branch is safe. There's no uncommitted changes to use.
yada yada yada. It it has that warning for you. But if I didn't want to do that, um in settings you can have let's see where is it. Um archive delete delete confirmation. There's a way. Oh yeah, you can arc. This is it. Session cleanup. There's a session cleanup session which lets you archive merged sessions where it like hides them. And then for deleting archived sessions where it just deletes the entire work tree, deletes everything. And so it creates the work trees as needed typically per session unless you want to say make a separate work tree just for this.
And you can do that. You can do any git operations you want in the app. And then for the sessions and for the tearown, I have everything at a week. You could do it at 30, 60, 90 days. You could do it at um a one day. You could also say disabled. Don't delete them ever. Let them live forever. You you can do that. When is remote coming to the GitHub Copilot app? Is it not already there? Oh, that's where you just say work in the cloud. See how it says cloud right here? Sorry, I should have said I shared that earlier.
If you do cloud, that's remote. Done. Uh, you could also run something in the CLI, hit SL remote, and then come and edit it in the app or on the web or whatever you want to do. I do think we need to I I thought the slash command was in there, too, but I didn't see it, so I need to edit it. I suppose the work tree will appear in any guey tools you're using. Yes. Yes. And so and you can configure how you want your work trees to be set up where let's see uh show and explorer.
Let me this is really tiny. It's not letting me zoom in but see all of these folders. Those are my work trees. And so uh those are my work trees in my actual GitHub repo and stuff. And so you can yeah you can show them in your particular uh git tools of choice. Yes, you can also uh my defaults in particular is I have all the work trees live in my co-pilot folder and so I I literally have a co-pilot work trees folder and then it has all of my repos that I've added to the app in that folder and then the work trees live in there but you can move that um into your uh into your repos themselves.
Can I use the app with a free account with the limitations? I'm excited about the integration with VS Code. Not currently. I think currently it's only available for paid users. I think right now the team, we know that it's not the best experience for free users to be able to use the app because just of how token usage works. And so the team is trying to optimize the token usage before we release it to free users so that it doesn't feel like, okay, I used it once, guess I got to pay. like that's not a that's not a good experience for anybody to have to deal with.
And so what we're hoping to do is improve that experience, improve some optimization so that free users can enjoy it. Um but yeah, otherwise it's users if you have if your business uses it like copilot enterprise and copilot business users um they already have access and they can use those as well. Are there any other questions? just let me know. Otherwise, we will see you back again next week. And so, if you have never been here before, Rubber Duck Thursdays is our weekly stream. It is here on GitHub's YouTube, Twitch, and LinkedIn. We do it multiple time zones throughout the day on Thursdays.
So, we have one that's at like I want to say it's 10:00 a.m. UK time. earlier in the day. It's like 4:00 am my time and I'm in Chicago, so I think that's correct. Like around 10:00 a.m. UK time. Then there's one earlier in the day. I think it's 10:00 a.m. Eastern. I should have this on the top of my head. And that one is in Spanish. So, and you can hang out with us on the Spanish version of this stream. And then uh right now right here where it is now uh 2:00 p.m. Central, noon Pacific at 10:00 a.m.
Pacific, you can check out this stream as well. But uh it's been fun chatting with all of you. Let's see. Can you chat in the app link? I'm not sure what you mean there, but you can chat in the co-pilot app. Yes, if that's what you're asking. Anyway, we are here every week. We will see you again next week. Also on our streams um on Fridays is open source Fridays and so in open source Fridays which happens on Fridays um we interview open source maintainers around the world and it's really cool to see the projects that they're maintaining the things that they're doing and that's a really fun stream to be a part of to just learn more about really cool open source projects.
We have so many different videos you could see if you check out the GitHub YouTube. We have GitHub for beginners, which shows you not only how to use Git and everything for beginners, but also GitHub Copilot. We also have the GitHub Copilot CLI for beginners, which shows you slash commands, how terminals work, that sort of thing. And we also have the download, which is a bi-weekly show of just like the latest news and updates. The latest episode should come out, if it doesn't come out today, it's coming out tomorrow. My co-orker Christine is hosting that one.
and a bunch of other ad hoc videos that will hopefully help you over time. But anyway, it's been lovely hanging out with all of you. I can't wait to see you next week. Hopefully my eye sty goes away. I'm cursed. This is like my fourth one this year. Anyway, small children, I'll talk to you all next time. It's been lovely. I'll talk to you later. Bye.
More from GitHub
Get daily recaps from
GitHub
AI-powered summaries delivered to your inbox. Save hours every week while staying fully informed.









