GitHub Universe 2026 闭幕演讲：开源协作、AI 规划与编程的创造力

简要描述Brief Description

GitHub Universe 2026 闭幕主题演讲：开源声音、演示与编程技艺GitHub Universe 2026 closing keynote: open source voices， demos， and the craft of coding

目录Table of Contents

• 欢迎参加 GitHub Universe 2026
  Welcome to GitHub Universe 2026
• 可破解徽章项目
  The hackable badges project
• 使用 GitHub Copilot 学习
  Learning with GitHub copilot
• AI 对开源的影响
  Ai's impact on open source
• 借助 AI 智能体扩展项目规模
  Scaling projects with AI agents
• 不断演变的开发者角色
  The evolving Developer role
• 介绍 Chopin：协作式 AI 规划
  Introducing chopin: collaborative AI planning
• Maggie Appleton 谈肖邦
  Maggie Appleton on chopin
• AI 对科技领域之外的影响
  Ai's impact beyond tech
• 与校园专家 Jocelyn Balanda 的访谈
  Interview with campus expert jocelyn balanda
• 编码的艺术：最终演示
  The craft of coding: final demo
• GitHub Universe 2026 闭幕主题演讲
  GitHub Universe 2026 closing keynote

欢迎来到 GitHub Universe 2026Welcome to GitHub Universe 2026

请欢迎马丁·伍德沃德。你怎么样？好。Please welcome Martin Woodward. How you doing? Right.

那么，感谢大家前来。这是本次会议的最后一场议程。感谢大家抽出时间参加。So, thanks for coming. This is the very last session of the conference. So, I appreciate you all taking the time.

我叫 Martin。嗯，我负责帮助照顾开源项目维护者、学生、教师，以及 GitHub 这里的开发者体验。嗯，这段经历很棒。My name is Martin. Um, I help look after open source maintainers, uh, students, teachers, and sort of developer experience here at GitHub. Now, it's been great.

过去两天我一直在四处走走。嗯，和很多老朋友叙了旧，也结识了一些新朋友。这一年真是太不寻常了。I've been got to walk around these past 2 days. Um, got to catch up with a lot of old friends and you've got to make some new ones as well. And what a year it's been.

嗯，如果我们看一下，你知道，现在 GitHub 上已经有 2.25 亿名开发者了。这里的数量还在不断增长。继续。Um, if we look, you know, we see we've got 225 million developers on GitHub now. It's growing in here. Go.

哇。好了。是的。Woo. There we go. Yeah.

那也不错。它正在地图上的每一个角落蓬勃发展，这真是太棒了。所以，我们有美国和印度。Might as well. It's growing in every single corner of the map, which is fantastic. So, we have the United States and India.

你知道，目前它们是我们两个最大的成熟社群，但今年真正让我印象深刻的，其实是巴西的发展。You know, they're our two biggest developed communities right now, but it's actually uh Brazil's growth that's really stood out for me this year.

我们……对，巴西。我们已经新增了超过 300 万名开发者。那么，你准备好了吗？We've Yeah, Brazil. We've added over 3 million developers. So, you ready for this?

告诉我我做得对不对。Benindo Casa。嘿，欢迎大家回家。Tell me if I've done it right. Benindo Casa. Hey, welcome home, everyone.

好了。谢谢。我问了他们该说什么。There we go. Thanks. I asked them what to say.

他没事。对。嗯，正如你昨天听到的，你知道，这种增长有时前所未有，也令人应接不暇，你也知道。He's all right. Right. Um as you heard yesterday uh you know this growth it's been unprecedented at times it's been overwhelming as you know.

嗯，过去一年里，人们推送代码达180亿次。也就是说，仅这一年的推送次数就超过了2024年和2025年的总和。拉取请求的数量也不遑多让，平均每天合并超过350万次。Um people have pushed code 18 billion times in the last year. That's if you look that's more pushes in 2024 and 2025 combined in one year. And pull requests have kept pace too with an average of over 3 and a half million merges a day.

过去 12 个月里，这个数字超过了 13 亿。所以事情进展得非常快。我不知道你是什么感觉，呃，我也觉得速度真的很快。That's over 1.3 billion in the past 12 months. So things have been moving really fast. I don't know about you, uh it feels really fast for me too.

和每隔一周似的，我们好像又有了一个新的编程模型，呃，我总觉得自己一直跟不上潮流。Like every other week, we seem to have a new coding model and um I'm kind of constantly feeling out of date.

所以对我来说，我的学习方式就是通过实践。嗯，我的团队……So for me, the way I learn is by doing. Um one of the things that my team

可编程徽章项目The hackable badges project

Universe 的……就是你们都挂在脖子上的那些可破解徽章。所以谢谢大家。好了。does for Universe is the the hackable badges that you're all wearing around your neck. So thank you. There we go.

呃，是的，这些设计出来——这些的设计初衷，是给你们一个可以尝试、可以玩一玩，并在日常工作之外做些事情的地方。这是我们有意为之的。我这周看到你们当中很多人都在玩这些东西。Uh, yeah, these were designed these are designed to give you a place to experiment, you know, to to play and to work on things outside of your usual day-to-day work. It's a very deliberate thing that we do. And I've seen loads of you playing with them this week.

我确实看到你们走进分组讨论区时——呃，就在最开始的第一场活动，有些人已经改造了自己的徽章，而我们甚至还没开始，这太棒了。这些徽章最初是我的一个个人兴趣项目，这些年来已经发展了很多。我们最初有这么一个……呃，基于电子墨水的设备，上面有一个小孔……I actually saw when you walking into the breakout uh the first very first session, some of you had already hacked your badges and we haven't even started yet, which is amazing. These badges started as a as a passion project of mine and have evolved a lot over the years. We originally had this uh e- in based device with a little hole in.

你们当中有几个人有那些东西。后来我们升级到了完整的会议设备，还用上了墨水屏，现在我们有了这个——戴在脖子上的小型超级计算机。There's a few of you who had those. And then we moved up to the full conference and had the ink and now we have this, you know, small super computer you're wearing around your neck.

太疯狂了。嗯，做这个徽章……它仍然是……算是团队一个很有趣的项目。我们可以打造一些极客又令人愉悦的东西，这正是我们当初来这里工作的原因。现在，在 ARM 朋友们的帮助下，它成了每年送给你们的一份小礼物。It's crazy. Um, working on this badge is it's still a it's kind of it's a fun project for the team. we get to build something that's just nerdy and delightful and that's why we all came to work here in the first place. Now, with the help of our friends at ARM, it's become this little annual gift to you.

然后，嗯，说来也挺有意思的，今年我参与这个补丁的程度其实是历年来最低的。可它却远远是最棒的一次。部分原因在于，嗯，你知道，我是个经理。And then, um, actually, funny enough, this year has been the least I've been involved in this patch. And yet, it's the best one by far. Part of that is the Well, you know, I'm a manager.

我主要是删除邮件和审批费用。所以，呃，这大多是因为我们塞进了这么多技术。你知道，它里面好像装了一个 RP2350 Cortex 处理器，还堆叠着各种传感器。I mostly delete email and approve expenses. So, uh, most of that is because of all the technology that we've packed in. You know, there's got like it's an RP2350 Cortex processor stacked with sensors.

而且这东西，嗯，你知道，想想就觉得不可思议，它比我当年学编程用的电脑强大得多——那台电脑还是我父母攒钱给我买的。And the thing, um, you know, it's just it's crazy when you think about this is way more powerful than the computers I learned to to code on that my parents saved up to buy me.

这真是太棒了。今年把徽章送到大家手上的过程也非常顺利。希望你们都注意到了。It's it's fantastic. The badge has been getting getting the badge to you has been fantastically smooth this year as well. Hopefully, you all saw that.

嗯，去年就没那么顺利。呃，我们当时……现在里面有几个 GitHub 的人可能正暗自发笑，因为去年全球供应链出现了短缺，接着又好像遭遇了台风，导致屏幕制造商停产了几周。所以最后……货物延迟送到了办公室。Um, last year, not so much. Uh we actually had there's a few GitHub people kind of smirking and laughing in there now because we we had these global supply chain shortages last year and then kind of a typhoon hit and took out the uh screen manufacturer for a few weeks. So we ended up um the shipments were delayed getting into the office.

所以我们把所有零件都送到了旧金山办公室，包括主板和屏幕。活动前一晚，大家才赶到，所有 Hubber 都来了，还有不少社区成员也来帮忙。我们当时正手工组装徽章，这样你早上 8 点到场时……So we had all the parts delivered to the San Francisco office, the boards and then the screens. And then the night before the event is when they arrived, all the hubbers came in and with loads of some of the community folks as well, and we were hand assembling badges for you so that when you rocked up at 8:00 a.m.

第二天早上，大家都能领到自己的门禁卡，然后离开了。现在，没人怪我。没人抱怨，你知道，也没人需要被说服，趁着有一点空闲时间就来办公室。the following morning, people were able to get their badges and walk out. Now, no one was blaming me. No one was complaining, you know, no one had to be convinced to come into the office with every any bit of free time that they had.

那么，我为什么要提这件事呢？嗯，我一直说，文化，文化就是在无人注视时发生的事情。看到那种以社区为中心的创业精神依然在 GitHub 蓬勃发展，我真的很高兴。So, why am I even mentioning this? Well, I've always said that culture, culture is what happens when no one is looking. And it's been great for me to see that that that startup, that community focused spirit, it's still thriving here at GitHub.

我希望你这周在这里时能感受到这一点，也希望你在与所有 Hubber 的互动中都能感受到这一点。总之，那个……And I hope you felt that as you've been here this week, and I hope you feel that in all the interactions with all the hubbers that you've had. Anyway, the

使用 GitHub Copilot 学习Learning with GitHub copilot

我之所以提到这个徽章，是因为我们在实践中学得最好。嗯，我看到你们都动手改造了徽章。社区成员也拿到了徽章，大家做出来的东西真是太棒了。reason I mention this badge exists is because we learn best by doing. Um, and I see you've all hacked badges. We've got badges in the hand of community folks and it's been amazing what everybody's built.

老板在哪儿？你在这里吗？不在。Where's boss? Are you in here? No.

好吧。去你的。我们，我们有……在哪儿？我记得那个。Okay. Screw you. We We got Where's I remember that.

他做了一个收据打印机。我们有，呃，Steve Francia。好。He did a receipt printer. We've got uh Steve Francia. Okay.

你也一样。呃，他给自己的徽章做了个 NES 模拟器，还在上面玩《塞尔达传说》。哦，你在这儿啊。Same to you. We've got uh he built an NES emulator for his uh badge and he like runs Zelda on it. Oh, there you are.

嘿，抱歉。前排。不好意思，哥们。Hey, sorry. Front row. Sorry, dude.

抱歉，你又回到我的圣诞贺卡名单上了。总之，呃，为了让还没机会做这个的人，以及在线加入的朋友也能了解一下，我想我们不妨快速为这个徽章做个应用，顺便带大家看看。所以，我们切到我的笔记本电脑上，呃，大家可以看一下，我们这里有个网站，里面列出了这个徽章的所有网站。I apologize. you're back on my Christmas card list. Anyway, in an effort uh for folks that haven't had chance to do this or folks joining online, I thought we would actually quickly build an app for the badge and sort of show you around. So, if we jump on to my laptop, um if you want to take a look, we've actually got the uh a website here that's got all the websites for the badge.

背面的二维码里就有它，而且这是一个完整的模拟器。好得简直离谱。但如果我把它调成全屏，你就可以进来看看这个。It's on the QR code on the back and it's a full simulator. It's ridiculously good. But if I just make it full screen, you can come in here and this is look at this.

这简直太疯狂了。你可以进入这里，稍微缩小一点视图，还可以和朋友一起玩《Flappy Mono》之类的游戏，做各种各样的事。整个模拟器、整个体验，其实都是一位极具才华的暑期实习生查理·尼科尔森打造的。This is mad. You can come in here, you can zoom out a bit, you can start playing flappy mono with your friends, you know, and and do all this sort of stuff. That whole emulator, this whole experience was actually built by a incredibly talented summer intern called Charlie Nicholson.

而且，凭着他那令人难以置信的天赋，再加上大概 95 美元的 Copilot 额度之类的，他能做到的事情真是惊人。不过，好了，来这里。我们实际上要做一个 Hello World。And it's amazing what he could do with, you know, his ridiculous levels of talent and like $95 of co-pilot credits or something like that. But yeah, come in here. We're actually going to do a hello world.

我把这个切换成低分辨率，这样你就能看到了。然后，大家好。不错。Let me switch this to low res for you just so you can see it. And then hello universe. Cool.

我的天，太棒了。你一路奔波来到这里。锵锵！I mean, amazing. You came all this way. Tada.

谢谢大家。Thanks everybody.

晚安。不。是啊。Good night. No. Yeah.

哇。呃，这挺酷的。不过，这有点像我们以前编写代码的方式，你知道吧。Wow. Uh, that's cool. But that's kind of like how we used to code, you know.

那么，让我给你看看用那个……副驾驶应用来操作是什么样的。我这就给你提前预览一下。好了。So, let me show you what it's like to do it with the uh the co-pilot app. So, I'm going to come in bit of a sneak preview. There we go.

然后我进入 co-pilot 应用。实际上，如果你还没做过，可以把 home 仓库克隆下来，然后只需输入一个提示词就行。这个游戏对我来说真的是一次生成完成的，它创建了我这里这个打砖块游戏。我可以按下“选择”来推送，然后就能玩打砖块游戏，所有操作都可以在这里完成。And I come into the co-pilot app. And what you can actually do is if you haven't done this already, clone down, bring down the home repository and then you can just, you know, type a single prompt. This was a literally a oneshot for me and it created this breakout game that I've got here and I can press select to push and I can play breakout and I can do it all here.

我可以用键盘操作，也可以在上面玩，因为它就是画布。它是在 Copilot 应用里使用画布。它可以访问所有 MicroPython 调试功能，实际上还能进入其中查看和调试各种东西，真正实现验证闭环。关键就在于，你的流程中要有一个验证闭环，这样你的智能体就能做出令人惊叹的事情。比如，在我的组织里，我做了一个《超级 Mona》游戏，完全没有侵犯任何人的知识产权。我可以进来，然后……哦，我最好别这么做，不然我们会被下架。I can do the keyboard and I can play with that because it's the canvas. It's using um a canvas inside of the copilot app. it can access all the microython debug and it can actually get in and you know inspect things and debug things and really get that verification loop and that's the trick is to have a verification loop in your process and then your agent can do amazingly impressive things for example on my uh on my organization I've got the game of super Mona totally not infringing of anybody's IP and I can come in here and I can oh I better not do that or we'll get taken down.

好了，开始了。我居然已经死了。嗯，这算是一个很好的开头，展示了你能做些什么。我想通过实践来学习，而我对机器学习了解不多——对于一个总是在这里谈论 AI 的人来说，这确实有点尴尬。不过，我之前确实没怎么接触过任何基础知识。There we go. and I've died already amazingly. So, um that's just uh sort of a great start of what you can do. I wanted to learn by doing and I don't know much about machine learning, which is problematic for somebody who stands up here and talks about AI all the time, but I hadn't really played with any of the fundamentals.

我漏掉了那部分。我想自学一些机器学习，也想为我的徽章开发一个应用。所以我想做一个计步器。I'd missed that bit out. So, I wanted to teach myself some machine learning and I also wanted an app for my badge. So, I thought I'd build a pedometer.

嗯，我进来了，也连接上了我的徽章。好了。那么，首先，计步器应该是什么样子的？Um, I've come in here and I've hooked up to my badge. Here we go. So, first of all, what should a pedometer look like?

所以，我通过蓝牙让 Copilot 接上加速度计，实际通过蓝牙向我实时传输遥测数据。它会通过振动之类的方式，向我展示徽章的状态。所以，这还不错。So, over Bluetooth, I had co-pilot hook up the accelerometer and actually give me live telemetry over Bluetooth. Show me what's happening with the badge with vibrations and all this sort of thing. So, that's okay.

所以，接下来我需要训练一个机器学习模型。现在我要点击“walk”按钮，我们会完成一系列训练步骤，这样我们就可以……我构建这个应用是为了让它能够对我的数据进行分类。于是我们点击 walk，然后就有了音频。So, then I now need to train a machine learning model. So I'm now going to hit the walk button and we're going to go through a set of training steps so that we can I built the app so that it can categorize my data. And so we hit walk and then we have the audio.

好吧。还是All right. Still

站着别动。我们可以做到。这是我站着不动的样子，不过看起来并不怎么安静。stand still. We can do that. This is me standing still which isn't very still.

开始步行。Start walking.

好的。那么我现在要做的是收集行走数据，这样它就能开始识别我在做什么。Okay. So now what I'm doing is collecting walking data so that it can then start to see what I'm doing.

检测到行走。Walking detected.

现在这个房间里有这么多 2.4、4 GHz 无线电设备，挺有意思的。数据传输效果不太好。所以我想往这边走一点。好了。好了。This is interesting with all the 2.4 4 GHz radio in this room right now. It's not getting great data. So, I want to walk down here a little bit. Here we go. There we go.

停止。Stop.

摆弄你的徽章。Fidget with your badge.

好吧。摆弄我的徽章。Okay. Fidget with my badge.

好痒。That tickles.

哦，抱歉。Oh, sorry.

停下。Stop.

好的。Okay.

再走一次。Walk again.

再走一遍。现在，它足够智能，能在我移动到超出范围时把遥测数据发回来。Walk again. Now, what is smart enough to do is actually send the telemetry back when I move to move out of range.

分析。Analysis.

这就对了。开始认出你的步伐了。There we go. Starting to recognize your walk.

哦，哇。好吧，Oh, wow. Okay,

停下来。摆动你的手臂。stop. Swing your arms.

挥挥我的手臂。就这样。Swing my arms. There you go.

大型主题演讲的气场。Big keynote energy.

好的，Okay,

停下。再走一遍。stop. Walk again.

再走一遍。我就绕着讲台走走。Walk again. I'm just going to walk around the podium.

这才像是一个周五刚合并到缅因州的人走路的样子。Now that's the walk of someone who just merged to Maine on a Friday.

是的。Yes, it is.

停下。站着别动。好。结束了。结束了。Stop. Stand still. Okay. Done. Done.

不错。Nice.

不错。对。那么，我现在要做的是……对。Nice. Right. So, what I'm going to do now Yeah.

嗯，是的。太棒了。那么，接下来我要做的就是在浏览器里训练一个机器学习模型，毕竟现在是 2026 年了。Well, yeah. Amazing. So, now what I'm going to do is train a machine learning model in my browser cuz it's 2026.

我们现在可以做这类事情了。我要把那个机器学习模型部署到我的徽章上。它是一个简单的决策树。We can do that sort of thing now. And I'm going to deploy that machine learning model. It's a simple decision tree to my badge.

现在，当我开始走动时，你已经可以看到它确实检测到我在走路。如果我停下来，它就应该用绿色标示“没有，我没有在走路”。这样，它就能区分我的步伐和其他动作。And now when I start to walk, you can already see it's actually detecting that I'm walking. And if I stop because it's calling it in green, it should No, I'm not walking. So now it can disamiguate my steps from my actions.

来点大型主题演讲的气势。就是这样。太棒了。Let's do big keynote energy. There we go. Amazing.

好了，就是这样。耶！现在你可以鼓掌了。So there we go. Yay. Now you can clap.

谢谢。Thank you.

AI 对开源的影响Ai's impact on open source

幸运的是，我是 2021 年最早一批有机会体验 Copilot 的人之一。你知道，即使是我，也不断被 AI 能做到的事情，以及我们构建事物的速度所震撼。现在，如果我退一步来看，我们其实可以意识到，这种速度正在对全世界的代码库产生怎样的影响。Now, I was one of the first people to uh get to play with Copilot, luckily enough, back in 2021. And you know even I'm constantly blown away by what AI can do and and how fast we can build things. Now if I take a step back we can actually appreciate that the you know the impact that this speed is having on on the entire world's codebase.

所以我觉得，我们不妨看看几件事：它带来的一些好处、一些意想不到的后果，以及我们能做些什么。我一直在反复问自己一个问题：在这个 AI 时代，我们怎样才能确保创造时依然用心？因为这就是我来到这里的原因。So I thought it'd be good to sort of look at a few things you know the good things that that gives us but also some of the unintended consequences and what we can do. The question I've been asking myself a lot is um how can we make sure that we're still building with heart in this age of AI? Because that's that's why I'm here.

以 Octiverse 为例，今天 GitHub 上 40% 的拉取请求都始于代理的初稿。现在是 40%，而三月份时这个比例还是 17%。Looking at Octiverse, 40% of the pull requests on GitHub today start with an agent's first draft. 40% and that was 17% in March.

这正在释放出巨大的创造力。我想，我们大家之所以都在这里，是因为我们都体验过：以前需要几个月才能实现的想法，现在几分钟就能推出。但这也关乎另一种同样宝贵的时间：你开始尝试一个想法，也许只做到10%，就能判断它是否值得继续推进。This is unleashing huge amounts of creativity. I think all of us here because we're here and we've had that feeling of shipping an idea in minutes when previously it would have taken us months to build. But it was also about time that's just as useful where you start an idea maybe get 10% of the way there and then you're able to figure out is it you know is this worth pursuing or not?

就像你知道的，你可以做原型。虽然创新步伐很快，但说实话，并不是所有的 PR 都是精品。好吧。Like you know you can prototype things. While this pace of innovation is huge let's be honest not all those PRs are diamonds. All right.

创作如今比以往任何时候都容易，但这并不意味着你手头的东西已经成熟到可以投入开发。有些只是糟糕的点子，有些甚至根本算不上点子。Just because it's easier to create than ever doesn't mean what you have is ready to commit. Some are just bad ideas. Some are barely ideas at all.

你知道，插槽。而且，嗯，我们看到低质量贡献有所增加。很多时候，贡献者是出于好意，但开源库的维护者要筛选这些内容，可能会非常非常耗时。You know, slot. And um we've seen a rise in lowquality contributions. Often often times well-meaning, but it can be very very time consuming for maintainers of open source libraries to wade through that work.

传统上，开源之所以能够运作，是因为它与这种贡献漏斗相契合。大多数人只是使用，只有一小部分人真正作出贡献；而这种平衡之所以能够维持，是因为编写代码本身存在各种门槛。你得先学会编程。Traditionally, open source has worked because it's been aligned with this contribution funnel. Most people consume and then only a small percentage actually contribute and and then that balance was in place because of all the friction that kind of came with coding. You had to learn how to code.

你必须了解社区的运作方式。而 GitHub 的一大优势在于，我们帮助人们更轻松地参与贡献。这也推动了开源社区的发展。You had to learn how the community worked. And the great thing about GitHub was that we helped make contributing easier. And that helped open-source communities grow.

但正是这种便利，让我能在周末创建项目；与此同时，也意味着有些人现在会在没有花时间了解项目文化、理解哪些东西受到重视、哪些不受重视的情况下，就开始为项目做贡献。因此，这些贡献，尤其是那些看起来像安全漏洞之类的东西，可能会占用维护者一大块空闲时间，也会让人很难找到那些真正有意义的贡献。But that same ease that helps me create projects on the weekend, it also means that some people are now contributing to projects without taking time to learn the culture, taking time to understand what's valued and what's not. So those contributions, especially if they're like, you know, things that look like security vulnerabilities or whatever, they can take a chunk of your free time as a maintainer, they can make it really hard to find those meaningful contributions.

在 GitHub，我们有一种“先交付、再学习”的文化。嗯，我们也学到了很多。我们和许多维护者交流过，并了解到需要更多控制措施，以确保开源仍然是一件有趣的事，也是我们愿意利用业余时间去做的事。At GitHub, we have a culture of ship to learn. Um, and we've learned a lot. We talked to a lot of maintainers and we learned that more controls are needed to help make sure open source remains fun and the thing that we want to do with our spare time.

比如说，我们为代码仓库设置了进行中工作（WIP）数量上限。这意味着，贡献者必须决定，在他们提交的拉取请求中，哪些一两个（具体数量由你设置）希望你优先查看。在你合并其中一个，或者他们撤下自己的请求之前，他们不能再提交其他请求。我们还让维护者更容易设置互动限制。So, for example, uh, we added um, work in progress limits for a repo. So, that means that uh, your contributors, they have to decide which of their pool requests is the one or two, however you set it, that they would like you to look at first. and they can't contribute another one until you've, you know, until you've either merged it or they take theirs down. We've also made it easier for maintainers to set interaction limits.

所以，也许他们想把项目设置成：你必须先有一个已合并的 PR，才能为这个项目做贡献。So maybe they want to set their project up so that um you have to have had a PR merge before before you can contribute to a project.

最后，还有一点，这件事在我们实施之前多少让我停下来想了想，因为我一直以来都在做开源。不过我们增加了一个功能，让维护者可以直接说：“你知道吗，我目前不想接收 PR，谢谢。”我觉得这样挺好的。And then lastly, and this one kind of, you know, made me pause for thought before we did it because I'm I've been open source forever, but we added the ability for maintainers to just say, you know what, I don't want PRs at this time. Thanks. It's I'm cool.

这些只是我们一直在添加的一些功能。我们添加了大量功能，努力减轻维护者的工作负担，因为要记住，维护者往往是在利用业余时间做志愿工作，贡献自己的劳动，好让你们能够依赖这些代码，并由他们来维护这些代码。现在，随着人们陆续加入社区，他们也需要学习这一点。And these are just some of the features we've been adding. We've been adding a ton of features to try and make it easier to reduce the workload for maintainers because you got to remember maintainers are volunteering often in their spare time volunteering their work to you to depend all that code that we depend upon to maintain it. Now, as folks come new into the communities, it's a lesson that people have to learn.

它是开源的，你可以，嗯，分叉项目，也可以添加自己的功能，但维护者是否愿意接受你的更改，就看他们了。Open source, it lets you um fork, it let you add your own features, but it's up to the maintainer if they want to accept your changes or not.

对吧？开源意味着可以自由查看、改编和分享，而无需征求许可。但这并不意味着可以把自己的意愿强加给他人。不过，事情也并非一片黯淡。我们的Right? Open source is the freedom to inspect, to remix, and to share without asking for permission. But it's not the freedom to impose your will on others. But it's not all doom and gloom. Our

借助 AI 智能体扩展项目规模Scaling projects with AI agents

数据显示，维护者正越来越多地使用 AI（包括 Agentic Workflows）来管理贡献，并帮助项目扩大规模。我看到 Frank 就坐在前排。Home Assistant 项目这次实际合并了 5,000 个拉取请求。data shows us that maintainers are increasingly using AI, including Aentic Workflows, to actually manage contributions and help scale their projects. I see Frank on the front row. The Home Assistant project has actually merged 5,000 pool requests this time.

就在过去一年里，开源软件的创建和分享量超过了我们历史上的任何时期。CNCF 一直是我们的长期合作伙伴，他们实际上一直在使用一大批 GitHub Copilot 智能体——这些智能体由我们协助赞助——来扩展和管理他们的项目网络，而众所周知，这个网络规模庞大。嗯，CNCF 一直在通过一个名为 OpenPrinting 的项目使用 AI。Now, more open source has been created and shared than at any point in our history in the last year. The CNCF has been a longtime partner of ours and they've been actually using a swarm of GitHub copilot agents that we we help sponsor and they use them to scale and manage their network of projects which as we all know is vast. Um the CNCF has been using AI with a project called open printing.

现在，我不知道你怎么想，但这听起来是个绝妙的应用场景。如果逢年过节回家时，能让我的智能体帮我叔叔修打印机，那我就全力支持 AI。Now don't know about you sounds like an amazing use case. If I could go home at the holidays and is my agent fixing my uncle's printer, I'm all in for AI.

好吧。我当时在和我的朋友乔治·卡斯特罗聊天。乔治在吗？我知道。Okay. when I was chatting with my friend George Castro. Is George in? I know.

好耶。好耶。就是这样。Yay. Yay. There you go.

乔治来了。耶。好吧。There's George. Yay. Right.

乔治是 CNCF 的一名工作人员。呃，他跟我解释说，他实际上发现，维护者们知道如何维护打印机驱动程序，而且他们其实很喜欢修打印机。George, he's a staffer at the CNCF. And uh he was explaining to me that actually what he found um is that uh the maintainers, they knew how to maintain printer drivers. They actually enjoyed fixing printers.

这正是他们的核心专长。差距其实在基础设施方面。他们需要帮助，把自己的专业知识转化成别人真正能运行的东西。Like that's their core expertise. The gap was actually around the infrastructure. They needed help getting their expertise into something people could actually run.

CNCF 有一个名为 Hive 的框架，George 和团队已经将 Hive 部署到四个打印机代码仓库中，帮助你管理各种容器更新和维护工作，以及 CI/CD 工作。这实际上也是我们看到的一个行业趋势，正在不断发展。DevOps 做得好的团队，最有条件充分利用 AI。So the CNCF have this framework called Hive and George and the team have deployed Hive into four of the printer repositories to help you know manage all that maintain kind of container refresh work, CI/CD work. And that's actually a trend that we've seen growing across the industry. Teams that do DevOps well, they're the best equipped to take advantage of AI.

你们为让人类能够安全、高速地工作而设立的同样的防护措施，也能帮助你们的智能体。不过我们都知道，在我们合作的公司中，DevOps 实践的成熟度差异很大。像 OpenPrinting 这样的团队，我们一次又一次地看到，AI 正在帮助团队偿还一部分技术债务，而这些债务原本阻碍着他们做好 DevOps。显然，其中很重要的一部分就是 GitHub Actions。The same safeguards that you put in place to allow humans to operate safely at high velocity, they also help your agents. But we all know that there's a there's a vast spread in the maturity of DevOps practices in the companies that we work with, you know, and like open printing, we see example after example where AI is is helping teams pay down some of that technical debt that was that was kind of stopping them from doing good DevOps in the first place. Now, a huge part of this is obviously um GitHub actions.

嗯，Actions 在过去一年中实现了大幅增长。去年，Actions 每秒运行作业近 1,500 次，也就是每天 1.24 亿次。Uh actions has seen massive growth in the last year. Actions ran a job almost 1 and a half,000 times a second last year. That's 124 million times a day.

这相当于我们在 2025 年看到的数量的两倍。在过去 12 个月里，我们捐赠了超过 115 万亿个 Action Core 分钟，帮助开源维护者维护和验证他们的项目。That's double what we were seeing in 2025. We've donated over 115 trillion Action Core minutes in the past 12 months to help open source maintainers maintain their projects and verify them.

这么多零啊。我一开始念出来的时候读错了。光是更快地构建和交付产品也没用，除非你构建的是正确的产品。That's a lot of zeros. I got this wrong when I was reading out originally. And it's no good building and shipping things faster unless you're building the right things.

明白吗？这就是我们需要人类的原因：判断力。Okay? And that's what we need humans for. Judgment.

我们需要准确理解自己要解决的究竟是什么问题，才能引导人工智能系统帮助我们解决问题。然后，我们还需要反思并弄清楚，我们所做的事情究竟是在让问题变得更好还是更糟。恐怕这种商业判断力不是人工智能能赋予你的。We need to understand exactly what problems we're trying to solve so that we can direct our AIS to help us solve them. And then we need to reflect and understand if what we're doing, are we making the problem better or worse. That business judgment, I'm afraid that's nothing AI can give you.

这就是为什么它叫工作。你知道，你的职责是定义什么才算好，之后 AI 才能帮你实现目标。It's why they call it work. You know, it's your job to define what good is, and only then can AI help you get there.

不断演变的开发者角色The evolving Developer role

这也是我坚信我们始终都需要开发者的众多原因之一。不过，如今开发者的定义正在发生巨大变化。我们与 AI 模型的交互方式也在变化。That's one of the many reasons why I firmly believe we're always going to need developers. But what a developer is now that's changing massively for us. So is how we're interacting with our AI models.

嗯，如果我们看一下最新数据，就会发现，随着这些 AI 模型快速发展，开发者实际上正在为具体任务选择合适的模型。现在，你们大约 25% 的任务会主动选择快速、轻量级的模型；其余任务则会在深度推理模型或通用模型之间切换。我们的数据清楚地表明，没有哪一个模型，甚至没有哪一个模型系列能够包揽所有任务；而且一旦有新模型推出，大家也愿意切换过去。Um if we look at the latest data, we see with the rapid uh pace of development of these AI models, developers are actually picking the right model for the job. Now you're actively choosing fast and lightweight models for roughly 25% of your tasks. and then you kind of switch between deep reasoning or general purpose models for the rest. It's clear from our data there isn't one model or even one model family to rule them all and the willingness to switch between to a new model when a new one's available.

这种变化是前所未有的。通过权衡手头任务的速度、成本和复杂度，你就能使用最适合自己的模型。That that change is unprecedented. By balancing the speed, the cost, and the complexity complexity of a task at hand, you're using the models that work best for you.

接下来，我们再来看看可以使用 Copilot 的各种各样的平台。每个人都有自己的看法。好了，我们都是开发者。And then let's layer on the many, many surfaces where you can use Copilot. Everyone has an opinion. All right, we're all developers.

每个开发者都有自己的看法。我们都坚信自己的看法才是最好的，对吧？因为事实就是如此。Every developer has an opinion. We're convinced our opinion is the best, right? Cuz it is.

但实际上，我们需要的是选择权。我们需要这些系统能够按照我们自己的方式运行，从终端到 Copilot 应用，再到 g.com 或移动端。我们很多人也希望能在自己选择的编辑器中编写代码。But in reality, what we need is choice. We need the systems to work our way from the terminal to the co-pilot app to g.com or mobile. And many of us want to code in our editor of choices.

你知道，VS Code、Visual Studio，还有 Eclipse。IntelliJ 系列编辑器搭配 Copilot 非常受欢迎。实际上，自从今年夏天的 WWDC 之后，Xcode 开发者现在也能原生使用 Copilot 了。You know, VS Code, Visual Studio, uh, Eclipse. The IntelligJ family of editors is incredibly popular with Copilot. And actually since Dubdubdc um in the summer Xcode developers, they've now got native access to Copilot too.

不过，如今我们与编码智能体交流的大多数方式还是通过聊天。但到了 2026 年，我们开始看到智能体编程远远超越了聊天。我们已经见到了大量开源和商业设备，比如罗技的这款设备，以及可改装的数码相框。Most of the ways we talk to our coding agents today though is is through chat. But in 2026, we're starting to see Aentic coding go way beyond chat. We've seen an abundance of devices open source and commercial like this one from Logitech and hackable picture frames.

你知道，人们正在构建各种工具，以便随时了解自己的智能体在做什么，并帮助控制它们。可在这里，我们仍然关注的是一位开发者与自己的智能体团队交流。软件开发是团队协作，对吧？you know, all these things that people are building to kind of keep on top ambiently of what their agents are doing and help control them. And yet here we're still focused on an individual developer talking to their team of agents. Software development is a team sport, right?

GitHub非常关注这样的问题：社交化编程的下一步是什么？它会是什么样子？我们如何与团队合作，如何与我们的智能体团队协作？GitHub is heavily focused on what like what's the next of social coding? What does it look like? How can we work with teams in with our teams of agents?

所以，我打算……So, what I'm going to

介绍 chopin：协作式 AI 规划Introducing chopin: collaborative AI planning

接下来要做的是介绍我们研究团队的 Maggie。接下来，我们会聊聊如何尝试与 AI 协作规划。那么，Maggie，你好。谢谢你，Martin。嗯，是的，do is introduce Maggie uh from our research team. go up next and we're going to talk see about experimenting with how we plan collaboratively with AI. So Maggie Hi, thank you Martin. Uh so yes,

Maggie Appleton 谈肖邦Maggie Appleton on chopin

我是 Maggie，是 GitHub Next 的设计师和工程师。我们是一个研究与原型开发团队，致力于探索智能体式开发前沿的各种可能性。I'm Maggie. I'm a designer and engineer at GitHub Next. So we are a research and prototyping team figuring out what's possible at the frontier of agentic development.

你们中的一些人可能了解我们过去的项目，比如最初的 GitHub Copilot、SpecKit 和 Agentic Workflows。最近，我一直在思考如何与智能体一起制定计划，以及这种做法有多糟糕。如今，我们通常会在命令行界面或桌面应用中开始规划，回答一连串没完没了的选择题。Some of you might know our past projects like the OG GitHub Copilot, SpecKit, and Agentic Workflows. And lately, I've been thinking a lot about planning with agents and how bad it is. When we plan today, we usually start in a CLI or desktop app answering an endless series of multiple choice questions.

大概做到第九题左右，我个人就已经觉得很累了，开始想：“行吧，选 A，推荐选项听起来不错。”嗯，这种环境让我觉得不太适合梳理问题、做出艰难的决定。最后主要变成了凭感觉做决定，因为一个小小的选择题框根本不是理解复杂选项及其取舍的合适形式。Somewhere around question nine, I personally get very tired and I just start being like, "Sure, option A, recommended option sounds good to me. uh this does not feel like the best environment for me to problem solve and to make hard decisions. This ends up being vibes based decision-making primarily because a very small multiplechoice question box is the wrong format to understand complex choices and trade-offs.

我这里没有足够的信息，也无法妥善地分析这个问题。等我终于回答完所有这些问题后，代理又会丢给我一份超长的 Markdown 计划让我阅读。这些计划实在太冗长了。I don't have enough information here and I can't properly explore the problem. Once I'm finally done answering all these questions, the agent then dumps out a huge markdown plan for me to read. These plans are so long and verbose.

一页接一页地阅读大段文字，我经常会感到疲惫。不过我觉得更大的问题是，这些事全都只有我一个人在做。队友看不到我正在阅读和审批的计划，我也看不到他们的。I often get tired reading page after page of walls of text. I think the bigger problem though is I'm doing all of this alone. My teammates can't see the plans that I'm reading and approving and I can't see theirs.

我没法很快地征求他们对我的界面设计或后端架构选择的意见。所以，我们都在各自私下与代理进行交流，做出会彼此产生影响、却没有彼此参与的决策。对我来说，直接把想法或功能做出来，然后提交一个 PR 来征求反馈，往往更容易。I can't quickly ask them about their opinion on my interface design or my backend schema choices. So, we're all having these private conversations with agents, making decisions that affect one another without involving one another. It's often easier for me to just build the idea or feature and put up a PR in order to get feedback on it.

但等到提交 PR 时才问“我该不该构建这个？”，已经太晚了。But opening a PR is way too late in the process to ask, should I build this?

或者说，这才是正确的架构方案？在我花费 token 搭建这个东西之前，我们本该早些讨论这件事——因为我们现在可能得重写它，甚至彻底弃用它，才能更好地探索人与智能体之间的实时协作规划。我们做了一个名为 Shopan 的原型，这个名字取自那位著名的波兰作曲家。Or is this the right architectural approach? We should have had that conversation much earlier before I spent the tokens building something we might need to now rewrite or throw out altogether. to explore better real-time collaborative planning with people and agents. We built a prototype called Shopan named for the famous Polish composer.

呃，我来给你们看看它能做什么。现在，我们在 Shopan 的一份新规划文档里，Martin 也和我一起在这里。Uh, and it let me show you what it does. So, here we are in Shopan in a new planning document. And Martin's in here with me.

我们可以同时在文档中输入内容，也可以在聊天中互相交流。这个文档与 Martin 的计步器代码位于同一个代码仓库中。Shopan 代理也在这里协助我们，一起梳理想法、讨论选项，并在确定方案之前探索各种可能性。And we can both type into the document or message each other in the chat. And this document is connected to the same repository as Martin's pedometer code. And we also have the Shopan agent in here with us to help us think through our ideas, discuss options, and explore possibilities before we commit to an approach.

所以，我们已经教会这些酷炫的小徽章如何计步了。So, we've taught these cool little badges how to count steps.

现在，我们想把它扩展成一场多人协作的步行冒险。这样，我们就可以各自戴着徽章四处走动、收集步数，然后把步数汇总起来，看看能否达到目标数量。我已经在项目聊天中描述了我的想法，Martin 补充说，他认为我们走动时必须能在没有 Wi-Fi 的情况下使用，因为大家都知道，会议场所的 Wi-Fi 从来都不可靠。Now, we want to extend that into a multiplayer collaborative walking adventure. So, we'll both walk around with our badges and collect steps, and then we'll pull them together to try to reach a goal amount. So, I've described my idea here in the project chat, and Martin has added that he thinks it will need to work without Wi-Fi while we're walking around because we all know you can never trust conference Wi-Fi.

然后我补充说，我觉得这个界面应该像一片小小的景观，让我们可以一起漫步其中。但 Martin 说，他更希望屏幕上显示一个大而清晰的数字。所以，这个决定我们之后还得再商量。And then I've added that I think the interface should be like a little landscape that we can walk across together. But Martin has said he would prefer a big readable number on the screen. So, we're going to have to figure out that decision later.

在我们聊天的同时，Shapan 正在为我们的计划补充备注。然后，它会根据我们的谈话，开始找出我们需要做出的重要决定。As we chat, Shapan is filling in our plan with notes. And then based on what we said, it begins to identify important decisions we'll need to make.

我们在这里的后端使用了决策模型。它还会开始为我们进行调研和构建原型。我们可以看到，Shopan 为我们制作了许多不同的界面模型供我们挑选。We're using decision models on the back end here. It also starts doing research and building prototypes for us. So we can see here Shopan has made a bunch of different interface mockups for us to pick from.

我们有了横向布局的构想、大地图、径向拨盘，以及 Martin 想要的那个醒目的大数字。打开这个原型后，你会看到这是一个可交互、可点击的界面，展示了各种状态和选项，比如为每个徽章添加步数，或者我们碰面并同步时会发生什么。We've got my landscape idea. We've got a big map, a radial dial, and that big readable number that Martin wanted. If I open up this prototype, you can see this is an interactive clickable interface that's showing us all the different states and options like adding steps to each badge or what happens when we meet up and sync together.

它已经弄清楚了我们需要查看的所有不同事件。通过查看实际的模型图，而不只是文字描述，我们能更直观地了解应用可能是什么样子，这真的很有帮助。因此，试用之后，我可以回到我的决策卡片，并决定我们想把横向布局和大数字结合起来。It's figured out all the different events we need to see. This is really helpful to get a sense of how our app would look by looking at actual mockups, not just text descriptions. So, after trying this out, I can come back to my decision card and decide that we want a mix of the landscape and the big number.

Shopan 会根据这个选择更新计划。呃，并在这里添加一个小的行内标记。打开它后，你可以看到我的名字，以及我做出这个决定的时间和日期。Shopan is going to update the plan with that choice. Uh, and adds a little inline marker here. If I open it, you can see my name and the time and date that I made this decision.

这样我们就有了一份记录，记下是谁做了什么决定。所以，几个月后如果有人想知道我们为什么选择这种设计方案，他们就可以找到这份记录，然后来问我。现在我们已经确定了界面，我正在问 Shopan 能否使用蓝牙让我们的总数保持同步。This gives us a record of who decided what. So, if in a couple of months someone wants to know why we chose this design approach, they can find the record and come ask me. Now that we've decided on our interface, I'm asking Shopan whether we can use Bluetooth to keep our totals in sync.

马丁，他说，马丁说他想知道我们走出信号范围后会发生什么，以及这些徽章的实际有效范围大概有多远。看完我们的聊天后，肖潘去网上做了一番调查，并整理了一份关于蓝牙支持和有效范围的报告。主动式智能体是我们正在探索的一个重要方向。Martin, he said, Martin says he wants to know what happens when we walk out of range and what range we could realistically expect from these badges. After reading our chat, Shopan has gone off and done some web research and put together a report on Bluetooth support and range. Proactive agents are one of the big ideas we're exploring here.

智能体如何预判我们的需求，并在我们开口之前就着手处理？这样一来，我和 Martin 可以继续聊天，或在别处做决定，同时由 Japan 为我们下一轮的选择准备材料。你还可以看到，文中有很多有用的示意图，探讨了将 IR 配对作为一种替代方案，并展示了如何在我们的两个徽章之间进行同步。How can agents anticipate what we'll need and go do that work before we ask for it? This means that Martin and I can keep chatting or making decisions elsewhere while Japan preps material for our next set of choices. You can see there are also lots of helpful diagrams in line exploring IR pairing as an alternative and showing how syncing would work between our two badges.

我个人觉得，特别大段的文字很难理清思路，而像这样的可视化说明对于理解数据流或架构决策非常有帮助。现在我们有了更多背景信息来帮助做出决策，接下来就需要做出一些决定了。首先，要考虑的是这些徽章应该在什么时候共享它们的步骤。I personally find very large walls of text difficult to reason about and visual explanations like this are extremely helpful in understanding data flow or architectural decisions. So now that we've got more context to help make our decisions, we need to make some. So first it's asking when should the badges share their steps.

我正在选择一种有意安排的会面与同步机制。接下来我们需要决定，成功的同步应该是什么感觉。I'm choosing a deliberate meet and sync system. We then need to decide how should a successful sync feel.

我最初选择了一条安静的内联消息，但 Martin 把它改成了一个小小的庆祝效果。你可以看到，他做出更改并保存时，他的头像会显示出来。这样我们双方都可以做出决定并保存。I initially select a quiet inline message, but Martin changes that selection to be a little celebration. So you can see that his face appears as he makes that change and saves it. So we can both make and save decisions.

最终方案的决策记录中会注明这些最终选择分别由我们谁做出。随后，Shopan 会根据这些更改，为我们的方案补充更多图表和说明。好了，既然我们喜欢这个方案的呈现效果，就可以告诉 Shopan，是时候开始构建了。And the final choices are attributed to each of us in the final plan's decision records. Shopan then uses those changes to expand our plan with more diagrams and explanations. Okay, now that we like the look of this plan, we can tell Shopan that it's time to build.

好了，Shopan 已经把计划拆分成一个个独立的任务，并为每项任务都发起了拉取请求。既然这个游戏已经开发完成，我们现在就来试试吧。我已经拿到徽章了，呃，我要开始走步数了。So, Shopan has broken the plan down into individual tasks and opened a pull request for each one. So, let's try this game out now that it's finished building it. I've got my badge uh and I'm going to start getting my steps in.

我们必须齐心协力，争取达到 10,000。We have to try and get to 10,000 together.

嘿，马丁，你能帮我个忙吗？Hey, Martin, can you help me out?

天哪，拜托。Jeez. Come on.

好吧，那我觉得你没问题。All right, then. I think you good.

我们开始吧。来吧。你准备好了吗？Let's do it. Here. You ready?

哦，好大。Oh, big.

好。再近一点。再靠近那个……Okay. Closer. Closer to the

嗯，Well,

最近怎么样？how's it going?

呃，这不太好。但屏幕上显示的是：“好了。你已经……哦，”Uh, it's not good. But what it said on the screen is There we go. You've Oh,

我们赢了。哦，好吧。是啊。好，我们试试看。we won. Oh, well. Yeah. Okay, we're going to try.

Y 演示成功。Y demo success.

我们就是这样证明这场现场演示的。太棒了。呃，谢谢你，Maggie。这充分展现了团队合作和创造力的力量。来，让我们为 Maggie 热烈鼓掌。谢谢你，Maggie。我们要把它弄掉。弄掉。That's how we prove this live demo. That's amazing. Uh, thanks Maggie. It goes to show you the power of, you know, teamwork and creativity. So, let's give Maggie a big round of applause. Thanks, Maggie. We're going to get rid of it. Get rid.

AI 对科技领域之外的影响Ai's impact beyond tech

现在，我喜欢和 Showben 合作的地方，其实是这种合作带来的感受。呃，感觉不再那么像是在做背景中的待办事项梳理，而更像是大家一起创造。这就是我们都在这里的原因。我们此刻之所以聚在这里，正是因为彼此协作所带来的力量。Now, what I love about working with Showben is actually how it feels. uh it feels a lot less, you know, like background uh backlog grooming and a lot more like creating together. And that's why we're we're all here. That's why we're here right now because of the power that comes with working collaboratively with each other.

现在，我们确实开始看到这种趋势也出现在计算机科学以外的领域。嗯，越来越多的人开始接触编程，并看到了开放共享的好处。科学界实际上一直是推动这一巨大增长的关键力量。专业技术服务和科学领域增长了五倍，如今已成为科技行业之外参与开源工作的领先群体。Now, we're actually seeing this start to show up outside of computer science, too. Um more people are getting into coding and seeing the benefits of sharing in the open. The scientific community, they've actually been a key part of of driving this huge growth. the professional technical services and scientific sector that's actually grown fivefold and they're the leading group now working on open source outside of tech.

开源代码库已悄然成为已发表研究背后可追溯记录的一部分。The open source repository it's quietly become the part of the the paper trail behind published research.

现在，为了帮助科学界，我们增加了引用支持，也添加了 ORCID ID。这有助于你识别研究人员。这样你就能弄清楚哪些作品出自杰出的计算机科学家 Dr.Now to help the scientific community we've added citation support and we also added orchid ID. This helps you identify researchers. So you know you can figure out which work is from the distinguished computer scientist Dr.

Martin Woodward，以及哪些是我说的废话。好吧，你可以在所有论文中正确引用它。嗯，我们已经看到世界各地的研究机构和大学越来越多地使用 GitHub。Martin Woodward and which is the nonsense from me. All right and you can reference it correctly in all your papers. Um and we've seen research bodies and universities from all across the world pushing more and more on GitHub.

来自 60 多个国家的近 2,500 所学术机构加入了我们的校园计划。我们拥有 100 多位校园专家。为我们喝彩吧！Nearly 2 and a half thousand academic institutions in over 60 countries are part of our campus program. We've got over a 100 campus experts. Give us a cheer.

嘿，他们来了。还有 500 多位校园专家和校友也参与其中，嗯，他们实际上会帮助各自大学里的学生学习 Git、GitHub、C-Pilot，并在这个不断变化的 AI 世界中摸索前行。Hey there they are. and over 500 campus experts, alumni, they're are getting involved um and they actually help students inside their universities learn Git, GitHub, C-Pilot, and just navigate this whole, you know, world of AI that's changing.

所以，嗯，你会看到他们穿着红色连帽衫，看到时就跟我打个招呼。So, yeah, you see them in the red hoodies, so say hi to me when you see

与校园专家 Jocelyn Balanda 的访谈Interview with campus expert jocelyn balanda

他们。为了让你了解这些校园专家有多么出色，我来向你介绍校园专家 Jocelyn Valande。Joseline 是加州大学戴维斯分校神经假体实验室的访问本科生，也是 MLH 教练和 Fellowship 项目成员。欢迎你，Joselyn。them. To give you an idea how amazing these campus experts are, let me introduce you to campus expert Jocelyn Valande. Joseline is a visiting undergraduate at UC Davis Neuroposthetics Lab and an MLH coach and fellowship. So, welcome Joselyn.

主持人：你还好吗？Host: You all right?

访客：你好。Guest: Hi there.

主持人：你最近怎么样？所以，你是在小学开始接触乐高头脑风暴和这类东西的。是什么让你对机器人产生了兴趣？你是怎么意识到它不只是个爱好的？后来又是怎么把它发展成职业的？Host: How you doing? So, now you started getting into Lego Mindstorms and things at school, elementary school. What drew you into robotics? How did you realize that was more than just a hobby? So, you turned into a career?

嘉宾：是的。所以，我会不断追问自己，事物的用途是什么，以及它们内部是如何运作的。在我的青少年学校里，乐高机器人听起来是一项有趣又新颖的活动。Guest: Yeah. So, I will question myself about the purpose of things and how they worked on the inside. Lego robots sounded like a fun and new activity at my youth school.

我记得小学时第一次参加比赛，他们问了我一个特别大的问题：“你长大后想做什么？”我回答：“我想制造机器人来帮助人们。”那时我还很小，完全不知道未来会有什么在等着我，但我想，小时候的我已经知道自己将走向何方。I remember going to my first competition during elementary school and they asked me like this huge question, what do you want to do when I grow up? And I said, I want to build robots to help people. I was pretty young and I had no clue of what the future awaited for me, but I guess little me knew where I was headed.

主持人：是啊，我的意思是，你肯定做过吧，因为你高中时就做了一个脑机接口来完成各种事情。呃，这东西到底是怎么运作的？你是怎么把大脑里那些杂乱的信号转化成实际动作，让假肢动起来的？Host: Yeah. I mean, well, you must have done because this is you in high school and you built a a brain computer interface to do stuff. Uh, how how does how does this work? Like, how do you take all that messy signals from our brain and actually turn them into action through a prothetic?

嘉宾：是的。所以，高中时，我必须完成一个研究项目才能毕业。我想重点研究的主要课题之一就是脑机接口。Guest: Yeah. So, in high school, I had to do a research project in order to graduate. So, one of the main topics I wanted to focus on was brain computer interfaces.

嗯，这对我来说是个全新的领域。所以，找到一位导师并不容易。碰巧，我可以给博士发邮件。Uh, this is a new area for me. So, finding a mentor wasn't easy. By chance, I could email Dr.

Tech Demter 的生物医学教授 Premier Malanise。呃，关于这个项目，我在这里制作了一个 BCI，基本上就是一种可以通过大脑信号控制的手部假肢。所以，当这个人——呃，这位朋友——想到张开手时，假手就会张开；当他想到合上手时，假手就会合上。Premier Malanise, a biomedical professor at Tech Demter. Uh for the project, I built this BCI over here, which is basically uh hand prosthetics that you can control using your brain signals. So when the person um this friend uh taught about opening her hand, the prosthetics hand will open and when he thought on closing the hand, it will close.

我使用了一个机器学习模型，呃，叫作线性判别分析，利用她自己的信号对她进行训练。我认为效果很好，因为我只有两个训练类别。I used an ML model uh called linear discriminant analysis in order to train her on her own signals. And I believe this worked great because I only had two training classes.

主持人：太疯狂了。所以，嗯，那你现在上大学了，现在在学校，而且你目前就在加利福尼亚。对。Host: That's crazy. And so um when you then now you're at college, now you're at school, you're currently here in in California. Yeah.

你现在具体在学什么？你在做什么项目？What are you actually uh studying right now? What are you working on?

嘉宾：是的。我最近开始在加州大学戴维斯分校 Neuropthetics 实验室担任访问本科生，实验室的主要目标是利用植入式阵列，通过 BCI 开发恢复言语能力的技术。我参与的项目叫“brain to voice”，旨在帮助 ALS 等疾病患者。Guest: Yeah. So I recently started a visiting undergrad position at the UC Davis Neuropthetics Lab where the main goal is to develop technology uh to restore speech with BCI using implanted arrays. I'm working on a project um called brain to voice to help people with disabilities such as ALS.

嗯，人工智能和机器学习算法会在说话者试图说话时解码电信号，而我的具体工作是在多个计算平台上开发底层硬件系统，以降低音频流传输延迟。Um AI and ML algorithms decode electrical signals uh when the person is trying to speak and my task there is specifically to work on low-level hardware systems on multiple computing platforms to reduce audio streaming latency.

主持人：太棒了。显然，你一直积极参与 MLH 的校园专家项目。社区在你的成长历程中发挥了很大作用。你有什么建议吗？我想，现在有很多学生正在家里观看。对于刚刚起步的人，你有什么建议？Host: Amazing. Now you've obviously been involved in campus experts in MLH. Community has been a huge part of your journey. What advice would you give? There's lots of students watching at home, I think. What advice would you give to somebody starting out?

嘉宾：是的。所以，我的社区归属感始于 Robbo Regos，这是 techree 的一个机器人社区。在我的家乡，呃，我们制造了一台自主机器人，嗯，它是一款通信助手，可以帮助人们。Guest: Yeah. So, my sense of community started with Robbo Regos, a robotics community at techree. So, my hometown uh where we built an autonomous robot um which is a com assistant to help people.

因为我想学习机器人技术，嗯，帮助他人，所以我回到家乡后加入了这个社群，寻找志同道合的人。对于所有学生，我的直接建议是：规划好自己的短期、中期和长期目标，并在一路前行的过程中找到能支持你的社群。对我来说，这些社群和支持包括 Robo Regos、校园专家计划、MLH、我在加州大学戴维斯分校的实验室社群，以及个人导师。Given my desire to learn robotics uh to help people, I got involved in this community back home to find people who shared my own passions. My direct advice to all students out there will be to map out your short, medium, and long-term goals and find communities along the way to support you. For me, it has been Robo Regos, the campus expert program, MLH, my lab community at UC Davis, or individual mentors.

当然，社区有各种各样的形式。所以，不要害怕寻求帮助，发封邮件吧。我当时就是这么做的。Communities of course come in all shapes and forms. So, do not be scared to ask for help, send an email. That's what I did.

嗯，只要你足够坚持并表现出兴趣，就会有人愿意帮助你。Um, if you are persistent enough and show interest, there will be people out there to help you.

主持人：你会帮忙的，对吧？嗯。Host: You will help, won't you? Yeah.

嘉宾：是的。好了，太棒了。好，那么谢谢你，Jos。请大家给他热烈鼓掌。Guest: Yeah. There we go. Amazing. Well, so thanks Jos. Give him a big round of applause.

主持人：现在我们构建软件的方式，它……Host: Now the way we build software, it's

编码的技艺：最终演示The craft of coding: final demo

完全变了。但、但我们构建软件的初衷并没有变。作为开发者，我们在实践中学习。completely changed. But but the why we build software, the why hasn't. As developers, we learn by doing.

我们通过解决问题、构建事物、与他人合作，以及不断突破自己的能力边界来学习。这正是 GitHub 存在的意义：为这些事提供支持。只用了少量 AI 积分，我就学会了如何构建一个机器学习模型，并将它部署到我的徽章上。We learn by solving problems, by building things, by working with others, by pushing the limits of what we know how to do. That that's what GitHub exists to support. With only a few AI credits, I learn how to build a machine learning model and deploy it to my badge.

这简直不可思议。不过我在想，我们或许可以把它推向更夸张的极端，做点不太一样的东西。That's just incredible. But I was thinking we could maybe, you know, push it to some more like ludicrous extremes. Do something a bit different.

那么，我们一路回到这边的讲台吧。幸好我带了吸汗带。嗯，好了。So, we'll go back to the podium all the way over here. Glad I brought my sweatb band with me. Um, there we are.

对。所以，显然呃，这是我的徽章。给你。Right. So, we've got obviously uh my badge. Here it is.

我们已经有了往下走的动作，不过我觉得要是还能处理一些其他动作就很酷，比如投掷和放下。We've got, you know, walking down, but I thought it'd be cool to some handle some other gestures like throwing and dropping.

所以，在我的朋友 Ship It Squirrel 的帮助下，我们就来做这件事。呃，我把徽章放到这里面，放进一艘船里。接下来我们要做一些训练，先练习投掷。所以我点到“投掷”。准备好。扔出去。哎呀。So, with the help of my friend Ship It Squirrel, we're going to do exactly that. So, uh, and I pop in my badge inside of here, inside a ship. And then what we're going to do is now some training uh for we'll do throw first of all. So if I come to throw. Get ready. Throw it. Oops.

等等。Hold it.

我应该拿着它，而不是把它掉了。算了。I should hold it, not dropped it. Never mind.

把它剪成波波头。Bob it.

把它弹出来。好。嗯。Pop it. Okay. Yeah.

等等。Hold it.

等一下。Hold it.

把它扔了。Throw it.

好的。All right.

等等。Hold it.

好的。Okay.

放下吧。好，挺好。完成了。不错。Set it down. Okay, cool. Done. Nice.

不错。好，现在我们来做 drop 训练。我这里只使用三个数据集，所以效果不会很完美。Nice. Right, let's do drop training now. I'm only doing three data sets here, so it's not going to be perfect.

准备好。Get ready.

我准备好了。I'm ready.

放下它。把它捡起来。Drop it. Pick it up.

好的。Okay.

放下它。Drop it.

别提了。Drop it.

把它捡起来。像烫手一样赶紧扔掉。Pick it up. Drop it like it's hot.

是的。Yeah.

把它捡起来。Pick it up.

明白了。Got it.

让它保持不动。好吧。算是静止吧。Leave it still. Okay. Stillish.

完成。Done.

完成。Done.

不错。Nice.

不错。好。那么，我们要训练机器学习模型了。Nice. Right. So, we're going to train our machine learning model.

我们要把它部署到 Squirrel 上，然后测试一下。训练。开工吧。We're going to deploy it to the squirrel and then we're going to give it a test. Training. Let's work.

继续。我要把它放在最右边。已部署。Go on. I'm going to leave it very right. Deployed.

已部署。86%。相当不错。对吧。Deployed. 86%. That's pretty good. Right.

好。那么我们要做的就是试着把它扔出去。准备好了吗？Right. So, what we're going to do is we're going to try uh throwing it. You ready?

不错。接下来我们试着把它放下。准备好了吗？Nice. And then we're going to try dropping it. You ready?

哦，我们再试一次。嘿，成功了。好了。Oh, let's try one more time. Hey, there we go. Right.

所以，现在我们已经把它弄好了。我觉得来场小比赛会挺酷的。那么，我们来找 Ship it squirrel 吧。So, now we've got that going. What I thought would be cool would be to do a bit of a race. So, let's get Ship it squirrel.

好吧。我相信你，对吧？你会表现好的。Okay. I'm trusting you here, right? You're going to be good.

好。那么我把你转接过去，我切换一下。Okay. So, I'm gonna pass you. Let me switch.

我来把这段音频打开。我们再试一次吧。Let me switch the audio on for this one. Let's just try again.

好。对。嗯。谢了，兄弟。我要发一些毛绒玩具。好，先拿着。请等我说开始再动。给你。哦，糟了。Okay. Right. Yeah. Thanks, bro. So, I'm gonna pass some plushies out. Okay. Hold on to them. Okay. Until I say go, please. So, there you go. Oh, nuts.

好了。谢谢。那是莫娜。There we go. Thanks. There's Mona.

我们有副驾驶了。希望如此。拜托。We've got co-pilot. Hopefully. Come on.

你能做到。好了。弗兰克，替我摇晃一下莫娜。You can do it. There we go. Give Mona a shake for me, Frank.

给你，Kasa。然后是 Ducky。接着还是 Ducky。There you go, Kasa. And then Ducky. And then Ducky.

对。抓住他们。按住他们。Right. You got them. Hold them.

把毛绒玩具拿出来。好。开始吧。Bring out the plushies. Okay. Here we go.

再来一些吧。好吧。不。Let's have some more. All right. No.

不。先别把它们扔来扔去。我们还想要更多。No. Don't throw them around yet. We want more.

我告诉过你，你只有一件事要做。拜托。好吧。I told you you had one job. Come on. Right.

准备好了吗？各就各位，预备，开始。我们要找出谁的腾空时间最长。Ready? On your marks, get set, go. What we're trying to do is find who has the biggest air time.

好。小心碰头。我们走。Okay. Watch your heads. Let's go.

确保我们这边也能弄到一些。不过你还没碰上一个难缠的呢。来了。Make sure we get some over here. But you haven't had a pushy yet. There we go.

小心头。好，准备好了吗？Watch your heads. Right. You ready?

倒计时，开始。10。莫娜领先。Countdown. Let's go. 10. Mona's in the lead.

来吧，莫娜。你能行的。莫娜要去……哦，卷轴。Come on, Mona. You can do it. Is Mona going to Oh, Scrolls.

准备好了吗？二。一。Ready? Two. One.

嘿。嘿。请抓紧你们的毛绒玩具。Hey. Hey. HOLD ON TO YOUR PLUSHIES, PLEASE.

等等。等等。嘿，拜托。看在老天的份上，你们替我紧紧抱住那些毛绒玩具。抱紧。答应我。抱紧。谢谢。我要静音了，不过一定要抱紧。好。好吧，刚才那个例子非常非常傻气，但也很有趣，展现了当我们突破已知可能性的边界时，会发生什么。我现在站在这里，呃，我们……Hold on. Hold on. Hey, come on. Hold on tight to those plushies for me if you're crying out loud. Hold on tight. Promise. Hold on tight. Thank you. I'm going to mute, but hold on tight. All right. Okay, that was uh very very silly but delightful example of what can happen when we we push the boundaries of what we know to be possible. As I as I stand here, uh we're

GitHub Universe 2026 闭幕主题演讲GitHub Universe 2026 closing keynote

这是我漫长职业生涯中见过的最大一次技术变革。毫无疑问，AI 让我们做事更快，也让我们更轻松地创建那些一直想做、却始终没有时间或技能去实现的应用程序。in the biggest technological shift that that I've seen in my long career. AI, it undoubtedly makes us faster. It makes it even easier to create applications that you've always wanted to create, but you've never had the time or the skills.

你知道，当你从 Universe 回家时，我想请你做几件事。首先，你如何扩大自己学习的圈子？你打算如何拓展自己所知的边界？You know, as you head home from universe, I want to ask you to do a few things. First of all, how can you how do you expand the circle of the people that you learn from? How are you going to push the boundaries of what you know?

那你又打算如何通过实践来学习呢？记住，在这个等式中，AI不会、也不可能取代人类的心。在这个新时代，你必须继续发挥创造力、判断力和品味。And how are you how are you going to learn by doing? Remember this, AI does not and cannot replace the human heart in this equation. You must continue to bring the creativity, the judgment, and the taste in this new era.

AI 不会写故事。故事由你来写。好了。AI doesn't write the story. You write the story. Okay.

好。各位，拿到毛绒玩具的人能不能站起来，把玩具举起来给我看看？Right. Who, if you got a plushy, can you stand up for me and hold your plushy in the air, folks?

站起来。好。Abby，把你的扔给别人。如果你戴着绿色挂绳，手里还有毛绒玩具，就把它送给别人。好，明白了。太好了。你们每个人都可以免费享用 12 个月的 Copilot Max。恭喜。Stand up. Okay. Abby, throw yours to somebody else. If you've got a green lanyard and you've got a plushy, give it to somebody else. Okay, got it. Great. You all you all get 12 months of co-pilot max on me. So, congratulations.

离开时，如果你能把这位 VIP X 留在这里，会有人给你一张优惠券，最多可用 12 12 个月。至于其他所有人，我们要做的就是从那些门口列队出去。祝大家平安到家，呃，也谢谢大家成为 GitHub 宇宙的一员。非常感谢大家。谢谢。When you leave, if you can leave out this this VIP X over here, somebody will hand you a coupon for your 12 12 months max. For everybody else, what we're going to do is we're going to march out of those doors. So safe home and uh thank you for being a part of the GitHub universe. Thank you very much everybody. Thank you.