← 返回任务列表

How to Get the Most from AI This Summer

189 段 · 1 位说话人 · 原片 20:19
M1
M10:00

今天这期 AI Daily Brief,我们来聊聊你的 AI 夏日冒险。

Today on the AI Daily Brief, we're talking about your AI summer adventure.

AI Daily Brief 是一档每天更新的播客和视频,聊的是 AI 领域最重要的新闻和讨论。

The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI.

好啦朋友们,在我们开始之前,先快速说几个通知。

Alright friends, quick announcements before we dive in.
M1
M10:34

好,今天是周末,而且还是个夏天的周末。

Alright, it is a weekend, a summer weekend, no less.

所以我们今天要来点行动导向的内容。

And we are going to go action-oriented today.

这周这期“大事件加长文”节目,我们其实要做两件事。

For this week's big things slash long reads episode, we're actually going to do two things.

而且这两件事都非常偏重实用。

Both of which are focused on the very practical.

第一件事是,Ethan Mollick 教授发布了他那篇博客文章的最新版,这类文章他之前已经写过很多次,核心就是针对不同任务推荐他觉得合适的工具。

The first is that Professor Ethan Mollick has published the latest version of a blog post which he's done a number of times that's all about which tools he recommends for different tasks.

他把这篇文章叫做:一份带有明确个人观点的“做事该用哪种 AI”的指南。

He called it an opinionated guide to which AI to use to do stuff.

所以我们会把他的所有推荐都过一遍,看看这些建议会不会影响你想怎么去做实验、怎么调整自己的工作流。

And so we're going to look at all his recommendations in case they might influence the way that you want to experiment and change your workflows.
M1
M11:24

不过我们先回到 Ethan 这边。

But let's start over with Ethan.

他在这份新的主观指南里提到的第一件事是,和他上次写这个相比,很多东西都已经变了。

The first thing that he notes in this new opinionated guide is that a lot has changed since the last time he did this.

事实上,他写道:现在所谓“用 AI 来做事”,涵盖的事情比以前多得多。直到最近,用 AI 还基本意味着通过 chatbot 跟一个模型来回对话。

In fact, he writes, What it means to use AI to do stuff encompasses so much more stuff than it used to until recently using AI meant talking to a model through a chatbot in a constant back and forth conversation.

而现在,它意味着使用 agentic system,也就是 AI 能把 AI 模型的“大脑”和一套能替你规划和执行的工具结合起来,一次性完成相当于真人好几个小时工作的系统。

Now it means using an agentic system where the AI is capable of doing the equivalent of many hours of real human work in one go by combining the brains of an AI model with a set of tools that let it plan and act for you.

基本上,agentic system 就是给 AI 一台可以用的电脑。

Basically, an agentic system gives an AI a computer to use.

所以一上来,他就把自己的建议分成了这两大类。

So right at the top even divides his advice into these two categories.

他实际上是在说,如果你只是想要一个低风险的答案,比如让 chatbot 给你个菜谱,或者帮你写封信之类的,那你用的几乎任何模型——包括免费的模型——都已经够用了。

And effectively, he says, If you are just looking for some low stakes answer such as asking a chatbot for a recipe or even asking it to help you write a letter or something pretty much any model that you might use including any of the free models is going to do well enough.

但反过来,只要是那种你真的很在意答案质量的事情,你就会想用 Claude 或 GPT-4o 这样的顶级模型,并把 thinking level 设到一个相对比较高的水平。

On the other hand, for anything where you really care about the answer, you're going to want to be using one of the premier models like Claude or GPT-4o set to a reasonably high thinking level.

至于真正高强度的工作,Ethan 认为现在实际上只有两个选择,就是 ChatGPT 或 Claude。

And when it comes to doing real intensive work, Ethan argues that there are effectively only two choices right now, which is ChatGPT or Claude.

这里值得一提的是,这基本意味着 Gemini 至少在现阶段已经正式掉出排行榜了。我觉得这对大多数人来说可能不算意外,但放在当下这个格局里,还是很能说明问题。

Now it is worth noting that that means that Gemini is officially out of the rankings at least at this point, which I don't think will come as a surprise to anyone, but still is pretty notable about the state of things.

接下来 Ethan 的下一部分叫做“给你的 AI 一台电脑”。

Now Ethan's next section is called giving your AI a computer.

虽然他没有用 harness 这个词,但他实际上说的就是你要搭建哪些系统,包括控制、权限等等,

And although he isn't using the word harness, he's effectively talking about the systems that you're going to put in place, including controls, permissions, etc.

这些东西会决定、也会塑造模型到底能做什么。

that are going to shape what the models can do.

比如他说,我把这些系统接到了我的 email、Google Drive 里一部分非私密的内容,还有很多其他应用上,但你得自己决定你对什么程度会觉得舒服。

He says, for example, I have the systems connected to my email, a non-private part of my Google Drive and lots of other applications, but you have to decide what you're comfortable with.

而且很明显,Ethan 想给你的那个推动,就是让你真的去用这类 connector。

And it's clear that the nudge that Ethan is giving is to actually use these sort of connectors.

他写道,一旦你把这些都设置好了,你就能做出相当强大的事情。

He writes, once you're set up, you can do pretty powerful things.

比如,我对这两个系统都说:连接到我的 Gmail,帮我为二十一号星期一要讲的 MBA 研讨课做准备,包括做一些 presentation 和 demo 供我参考找灵感。

For example, I told both systems, connect to my Gmail and help me prep for the MBA seminar I'm giving on Monday the 21st, including building some presentation and demos as inspiration.

把这个主题下所有还没回复的消息也都处理掉。

Answer any outstanding messages on the topic.

两个系统都开始干活了。

Both systems got to work.

它们连接到了我的 email,也弄清楚了任务内容,甚至还正确判断出下一个“二十一号星期一”是在九月,不是在八月。

They connected to my email and figured out the task, including correctly figuring out that the next Monday the 21st was in September, not August.

在那之后,它们就直接开始干活了,这就是 agent 会做的事。

And after that, they just started working, which is what agents do.

它们会上网做研究,决定要做一个什么样的演示 demo,还会琢磨如果是我,该怎么回复那个给我发邮件的同事,等等。

They did research on the web, decided on a presentation demo, thought about how I might want to respond to the colleague who emailed me, and more.

大概十分钟之后,它们两个都交回了答案,做出了一整套教学材料,还给那位同事写好了一封邮件。

About 10 minutes later, both returned answers, having created a range of teaching materials and written an email to the colleague.

这真的很厉害,这些事本来可能要人类花上几个小时才能做完。

This is impressive stuff that would have taken a couple hours of human work.

但他说,Claude 只是准备了一个草稿,而 ChatGPT 实际上真的把邮件发给了我的同事。

But he says, while Claude only prepared a draft, ChatGPT actually sent an email to my colleagues.

这是怎么回事呢?嗯,其实是我的错。

What happened? Well, it was my fault.

因为我之前已经给过 ChatGPT 权限,让它可以代表我发邮件,而我对 Claude 的设置是要先问过我。

I had previously given ChatGPT permission to send email on my behalf, and Claude was told to ask me first.

所以他说,这是他学到的第一批重要经验之一:当你把这些系统用在真实工作里时,权限这件事非常重要。

So he says in one of his first big lessons, when you use these systems for real work, the permissions matter a lot.

这两家公司都让你自己决定,AI 在采取行动之前是不是必须先跟你确认,比如发邮件、买东西,或者修改文件。在你真正信任这个系统、也了解它会犯什么错之前,最好把所有设置都保留在先请求批准,这也是默认选项。

Both companies let you decide whether the AI must check with you before acting, such as sending an email, buying something, or changing a file. Until you trust the system and understand its mistakes, leave everything set to ask for approval first, which is the default.

随着你越来越熟悉,Ethan 建议,你可以逐步扩大这些 AI 对你电脑的访问范围,这样它们也就能帮你完成更复杂的工作。

As you get more comfortable, Ethan suggests that you can expand how much of your computer the AIs have access to and get more complicated work as a result.

Ethan 的原话是,这些应用可能最有意思的一个技巧,就是它们真的可以像你一样直接使用你的电脑。

Indeed, Ethan writes, Probably the most interesting trick of these apps is that they can just use your computer the way you would.

如果你在 Claude 或 Codex 里打开 computer use 这个选项,AI 真的就可以直接接管你的鼠标、浏览器和电脑。

If you turn on the computer use option in Claude or Codex, the AI can literally take over your mouse, browser, and computer.

对,这确实有安全方面的担忧,所以你得谨慎操作,不过它带来的结果也可能非常惊人。

Yes, this is a security concern, so you should proceed carefully, yet the results can be amazing.

我通过 Codex 里的 Sora,让 ChatGPT 5 去下载一个 3D 建模程序,然后用它做一个非常具体的设计。

I asked ChatGPT 5 via Sora in Codex to download a 3D modeling program and use it to create a very particular design.

下载 Blender,然后做一只在飞机上用笔记本电脑的水獭。

Download Blender and make an otter using a laptop on an airplane.

接着他还分享了一段视频,里面 AI 就真的在做这件事。

He then shares a video of the AI doing exactly that.

他说,如果把这些能力放在一起看,你会发现,AI 几乎能做任何一个能接触到你电脑的人能做的事,而且有时候做得还更好。也就是说,我自己根本不知道 Blender 是怎么用的,当然有时候它也会做得更差。

If you put this all together, he says, you will find the AI can do almost anything that a person with access to your computer can do, sometimes much better, i.e. I have no idea how Blender works, although sometimes worse.

当然啦,我还是更愿意自己做 slides、自己写邮件,谢谢,不过 AI 一直在变强,所以它的能力也在不断提升。

I'd rather make my own slides and write my own emails, thank you, but the AI keeps getting better so the capabilities keep improving.

现在,很明显是在回应一些他到现在还经常听到的抱怨。在他举的一个例子里,他指出,有些人可能是基于 AI 之前几代版本,形成了错误的预设。

Now, clearly responding to some of the complaints that he still hears, in one of the examples he gives, he points out where people maybe have mistaken assumptions based on previous iterations of AI.

Ethan 说,我有一本新书会在十月出版。

Ethan says, I have a new book coming out in October.

这本书已经经过了好几轮专业编辑和校对,不过我还是把完整的 PDF 交给了通过 Codex 使用的 GPT-5,让它从头到尾再检查一遍。

It has been through rounds of professional editing and proofreading, but I gave GPT-5 via Codex the full PDF anyway and asked it to check it all over.

这个 AI 工作了三十分钟,核查了 一百九十五 条参考文献,还给了我好几页笔记,这些工作本来要一个研究团队花很多个小时才能完成。

The AI worked for 30 minutes, chased down 195 references, and gave me pages of notes that would have taken a team of researchers many hours.

AI 进步有多大,有一个迹象特别明显:它给我的每一条备注都准确无误,没有幻觉出来的页码,没有编造的文本,也没有任何我能看出来的错误。

One sign of how far AIs have come is that every one of the AI's notes was accurate and there were no hallucinated page numbers, no invented text, no errors I could spot at all.

事实上,我遇到的还是相反的问题。

In fact, I had the opposite issue.

这个 AI 实在是太吹毛求疵了。

The AI was incredibly nitpicky.

好在,我还是用了人类自己的判断力,把这类挑剔意见给否掉了。这也很符合一个主题:和这些系统一起工作,更像是在管理,而不是在聊天。

Fortunately, I used my human judgment to reject these sort of complaints, which fits the theme that working with these systems is more like managing than it is chatting.

你几乎可以把这些 AI agents 想成一个你可以委派工作的团队。

You can almost think of the AI agents as a team you delegate work to.

最后,Ethan 还是又讲回到了 Google。

Now lastly, Ethan does come back to Google.

他写道,Google 不久前在 benchmarks 上还处于领先,但现在在真正重要的地方已经落后了。

He writes, Google, which led on benchmarks not that long ago, has fallen behind where it now counts.

它既没有领先的 frontier model,也没有任何接近 Codex 和 Claude Code 的东西。

It has no leading frontier model and has nothing close to Codex and Claude Code.

所以这就是为什么我现在不建议你把 Gemini 当成主要系统,不过这个情况也可能很快就会变。

That is why I don't suggest Gemini as your primary system right now, though this could change quickly.

不过他说,这也不代表 Google 就没有东西可补充。

However, he says that doesn't mean that Google has nothing to add.

他提到 Gemini Notebook,也就是以前叫 NotebookLM 的那个,还有他们一些更偏丰富媒体的模型,比如能处理视频的 Gemini Omni。

He points to Gemini Notebook, formerly known as NotebookLM, also some of their more rich media models like Gemini Omni, which works with video.

整体来说,这套说法其实挺简单的。

Now overall, this is fairly simplistic.

说真的,完全不是在阴阳 Ethan,但你坐在这儿可能会想,等等,就这些吗?

In fact, with absolutely no shade to Ethan, you might be sitting here thinking, wait, is that it?

但你得记住,Ethan 面向的是更广泛的受众,而且确实很可能是比每天听 AI 播客的这些人,AI 使用成熟度低得多的一群人。

But you have to remember that Ethan is writing to a much wider audience and indeed probably a much less mature AI-using audience than the folks who are listening to a daily AI podcast.

而我觉得这篇东西有意思的地方,尤其是跟这个思路之前的版本相比,不是他再按一个个 use case、一个个要点去讲,也不是去指出他喜欢哪个模型;Ethan 在这里真正做的,其实是把两种完全不同的交互模式,清清楚楚地划出了一条界线。

And what's interesting to me about this piece, especially as opposed to previous editions of this same idea, is that rather than going use case by use case and point by point and pointing to which model he likes, really what Ethan is doing here is drawing a bright line between two entirely different interaction patterns.

一边是老式的、基于聊天的交互方式,这种基本上用什么都行;另一边则是这种全新的管理 agents 的模式。

On the one side, there is the old chat-based interaction for which pretty much anything will suffice and on the other is this totally new managing-agents pattern.

在我看来很明显,Ethan 是想把这种“管理 agents”的交互模式,讲得让那些一看到 harness 这种词就会直接听懵的人,也觉得容易接近、容易上手。

It is very clear to me that Ethan is trying to make the agents-management interaction pattern feel approachable and accessible for people who if they see words like harness are just gonna have their eyes rolled to the back of their heads.

但这确实是工作方式上的一个大转变。

But it really is a big shift in working.
M1
M19:22

你可以按等级或者按主题来筛选它们,主题包括像 tools、agents 和 automation,building and creating,或者搭建 knowledge 和 context 这类。

You can filter them by level or by topic, with the topics being things like tools, agents and automation, building and creating or setting up knowledge and context.

你还可以给自己的 passport 盖章。

You can also stamp your passport.

这个 passport 可以只自己用,或者你也可以把它设成公开,分享你做过的所有很酷的事情。

The passport can be just for you, or you can set it to public and share all the cool things that you've done.

你拿到的章越多,你的 street cred 也就越高。

The more stamps you get, the more street cred you get.

一个章,我们叫你 day tripper;三个章,你就会变成 certified AI explorer。

One stamp and we call you a day tripper, three stamps and you become a certified AI explorer.

拿到六个章,你就是 trailblazer;拿到十个章,你就是 globetrotter。

At six stamps you're a trailblazer and at ten stamps you're a globetrotter.

这是不是把 millennial 式尴尬拉满了?

Is that the height of millennial cringe?

是。那我在乎吗?一点都不。

Yes. Do I care? Not even a little.

我在 enterprise AI 里一直反复看到的一件事,就是公司在每个 cloud、每个 model、每个 framework 之间来回对冲,或者花钱找 GSI 做一个永远做不完的 pilot。

One thing I keep seeing in enterprise AI, companies hedging across every cloud, every model, every framework, or paying a GSI for a pilot that never ends.

而那些真正把东西 ship 出去的团队,他们已经选定了一条路,然后动作非常快。

The teams actually shipping, they've picked a lane and they move fast.
M1
M113:04

我对 AI Adventure 很兴奋,其中一个原因是,就像你在 Claude campaign agent OS 那边看到的,我们当时做的很多东西都非常进阶。

Part of what I'm excited about for the AI Adventure is that, as you saw with both Claude campaign agent OS, a lot of the stuff that we were doing there was very advanced.

但到了 AIDB New Year 这边,很多内容其实就非常容易上手。

Whereas when it came to AIDB New Year, a lot of it was really accessible.

它本来就是给那些在 AI 这件事上还刚刚站稳脚跟的人准备的。

It was meant to be for people who were still just getting their feet under them when it came to AI.
M1
M113:37

比如有一个 quick trip 项目,我们叫它 Pack Your ID。

One example of a quick trip project we're calling Pack Your ID.

这当然是一个 context 项目,它会帮你建立一份关于你自己的资料,你可以把这份资料交给任何一个你正在互动的 AI,这样它就能更了解你的背景,也能给你更好、更加贴合语境的结果。

This is, of course, a context project that helps you build a profile of yourself that you can give any AI you're interacting with so that it has better context about you and can provide you better, more contextual results.

这些项目或者 labs 的运作方式是,每一个都会附带一套说明,以及一组核心任务,其中也包括你可以直接复制的 prompts。

The way that these projects or labs work is that they're each going to come with a set of instructions and some set of core tasks, including prompts that you can copy.

比如说,这里就有一个复杂 prompt,它要求 AI——引号——“帮我为这个 AI 产品写好并安装一个轻量级的全局身份设定,这样以后新的对话一开始就知道我是谁。”

So, for example, this is a complex prompt that asks the AI to, quote, help me write and install a lightweight global identity for this AI product so future chats start knowing who I am.

这个 prompt 里还包括,要根据当前系统来判断,持续生效的个人 context 应该放在哪里。

The prompt involves figuring out where standing personal context should live based on the current system.

它会指示 AI 通过采访用户的方式,先起草一份简短说明。

It instructs the AI to interview the user to draft a short brief.

最后,它会要求 AI 产出这样几样东西:第一,一个可以直接粘贴使用的全局 ID 模块,大概一百五十到三百词,写成给未来 AI assistant 的指令;第二,针对这个产品当前 UI 的准确安装步骤;第三,一个我应该在新对话里提出的测试问题,用来验证它是不是已经加载成功了——也就是说,一个只有我的 ID 才能回答得好的问题。

Then, finally, it asks the AI to produce one, a paste-ready global ID block of roughly 150 to 300 words written as instructions to a future AI assistant, two, exact install steps for this product's current UI, and three, one test question I should ask in a fresh chat to verify it loaded, i.e. something only my ID would answer well.

接下来,对每个 lab 或 project,我们还会再提供一些额外的层级,也就是 project extensions。

Now, for each lab or project, we're also going to provide some additional layers, i.e. project extensions.

这里的话,就是第二个模块:怎么对我提出反对,什么时候该质疑,什么时候该问澄清问题,什么时候该拒绝一个模糊的请求。

In this case, it's a second block, how to push back on me, when to challenge, when to ask a clarifying question, when to refuse a vague request.

最后,我们还会给你一个 stretch goal。

And finally, we're also going to give you a stretch goal.

说到 Pack Your ID,其实这个 context project 还有一个更高级的版本,叫 Personal Brain,也就是一小组文件,用来教任何 AI 你是谁、你在做什么,以及你喜欢别人怎么做事。

Now, when it comes to the Pack Your ID, there's actually an even more advanced version of this context project called Personal Brain, i.e. a small set of files that teach any AI who you are, what you're working on, and how you like things done.

这基本上就是一种 context pack,类型上和我在 contextportfolio.ai 那个 portfolio builder 里分享过的一样,只不过这里是把它整理成一个个人学习实验。

This is basically a context pack of the type that I shared at contextportfolio.ai in that portfolio builder, but organized as a personal learning experiment instead.

如果说这些 quick trips 是让你从你已经在用的 AI 里轻松获得更多价值的方法,那么 excursions 就是一类 project,你会去构建一些真正能在当前 session 结束后还继续存在的东西。

Now, if these quick trips are easy ways to get more from the AI you already have, excursions are a type of project where you are going to build something that actually outlasts the current session.

我刚刚说的 Personal Brain 就是一个例子,不过我们也有一些 excursions,是关于去执导一个完整成品的创意作品,也就是说,不只是一次性 prompt 一下,而是真正做出一个完整的、由 AI 生成的创意 project。

The Personal Brain that I was just talking about is an example of that, but we also have excursions for directing a complete and finished creative piece, i.e. not just one-shot prompted, but actually a full, complete, AI-generated creative project.

还有一个 excursion,是关于用 vibe coding 做点东西出来。

There is an excursion for vibe coding something.

我知道,虽然我们经常聊这些工具,但还是有很多人根本没时间,或者没有那个精力带宽,真的去做出自己的第一个 application。而我真心觉得,这其实是你现在使用 AI 时能获得的最大突破之一,而且它比你几乎能做的任何别的事,都更能把你从“AI 只是一个助手,帮你把现有工作做得更快、更便宜或者更好”这种想法里带出来,

I know that as much as we talk about these tools, there are still plenty of people that just haven't had the time or bandwidth to actually go build their first application, which I believe is genuinely one of the biggest unlocks that you can have using AI right now, and more than just about anything else you can do will shift you from thinking of AI as an assistant that can help you do your current work faster, cheaper, or better,

带到一种能解锁全新能力的东西上。

to something that unlocks totally new capabilities.

然后,对那些想更有野心一点的人,我们还有我们称之为 expeditions 的内容,它们的目标是彻底改变你和 AI 协作的方式。

And then for those who want to be really ambitious, we have what we call expeditions, which are meant to totally change how you work with AI.

现在已经上线的 expeditions 有两个。

There are two expeditions that are live right now.

第一个叫 Lemonade Stand,它的目标是真正做出一个由 AI 配置团队的 micro-business,并且带有一个真实、可以马上运行的需求测试。

The first is called Lemonade Stand, and the goal of it is to actually create an AI-staffed micro-business with a real, ready-to-run demand test.

这个不只是给 entrepreneurs 或 solopreneurs 准备的。

Now, this one is not just for entrepreneurs or solopreneurs.

就算是那些对自己的日常工作很满意、也确定自己永远不会离职的人,完整走一遍设计一个 micro-business 的过程,也会对你在日常场景里怎么使用 AI 产生非常惊人的学习效果。

Even for folks who are completely happy in their day job and sure they'll never leave, going through the process of designing a complete micro-business is going to have incredible learning effects for how you use AI in your normal context as well.

这个 expedition 具体分成三个 sprint。

Now, this particular expedition is organized into three sprints.

在 Discovery Sprint 里,你会从你过往经历、技能、可调用的资源,以及外部的 idea labs 里提炼出一些不那么平庸的想法,最后得到一个五到八个点子的短名单,以及一个排好序的前三名,并附上 founder fit notes,也就是它们里面哪些看起来最适合你。

In the Discovery Sprint, you extract non-trivial ideas from your history, skills, access, and external idea labs, leading to a short list of five to eight ideas, and a ranked top three with founder fit notes, i.e. which of them seem most likely to be a good fit for you.

第二个 sprint 是做 business plan:从那前三个概念里挑一个,把它规划成一个真实的 micro-business,配上一个 AI org chart,其中包括第一周的任务清单、一页纸计划,以及 AI staff map。

Sprint two is about creating a business plan, taking one of those top three concepts, and mapping it as a real micro-business with an AI org chart, including a week one task list, a one-pager plan, and an AI staff map.

第三个 sprint 是 Validation Sprint,在这里你会学会在过度建设之前先测试需求,包括给软件做一个轻量 demo、设计实验,以及从真人那里拿到真实信号。

The third sprint is the Validation Sprint, where you learn to test demand before you overbuild, including a thin demo for software, an experiment design, and getting real signal from a human.

再说一次,这些 sprint 里的每一个,都会配一个核心任务或者一组任务,再加上一个可复制的 prompt,你可以直接丢进任何你正在使用的 AI 里。

Now, again, each of these sprints is going to come with a core task or a set of tasks, and a copyable prompt that you can drop into whatever AI you're using.

不过很重要的一点是,尤其是像这种更大型的 excursions,这些可复制的 prompts 可能更多只是帮你起步;实际上,很多 prompt 真正展开后,会变成你和 AI 之间一整套对话和互动。

Now, importantly, especially when it comes to the more extensive excursions like this, the copyable prompts may be more about getting you started, and indeed, many of the prompts actually open up an entire set of conversations and interactions between you and the AI.

比如在 idea mining sprint 里,那个可复制 prompt 实际上会把整个过程组织成一系列阶段,其中包括像 AI 对你做一次深度访谈这样的环节,来帮助它判断哪些 business ideas 可能更适合你。

For example, in the idea mining sprint, the copyable prompt actually organizes things into a set of phases that include things like a deep interview by the AI of you to help it figure out what business ideas might be a good fit.

另外,尤其是针对其中一些更复杂的想法,AI adventure platform 也会附带一套资源,告诉你可以去哪里继续深入。

Now, especially for some of these more complex ideas, the AI adventure platform is also going to come with a set of resources for where you can go deeper.

还有一个 expedition,我也特别想推荐给你,我们叫它 The Loop。

One more expedition that I want to sell you on we're calling The Loop.

你已经听我在这个节目里没完没了地讲 loops 了:你拿一个定义清楚的任务,配上测试构建和 agent 可以重新检查的 diffs。但把它用到非技术工作里,在很多方面都更难,或者至少说,没有把它用在 software engineering 上那么直观。

Now, you have heard me talk endlessly about loops on this show, where you take a well-defined task with test builds and diffs the agent can recheck, but applying it in non-technical work is in many ways more difficult or at least less intuitive than it is in applying it to software engineering.

The Loop 这个 expedition 的目标,就是带你一步步做出一个真正的 agentic loop,而且是在一个对你有用的工作领域里。

The goal of The Loop expedition is to help you walk through building an actual agentic loop in an area of work that is useful for you.

在这个项目甚至还没进入 prompts 和具体步骤之前,就已经有一整套背景学习内容,来帮助你更熟悉这些概念。

Now, before this one even gets into the prompts and steps, there is a bunch of background learning that's going to get you more familiar with the concepts as well.

其中一些部分包括:什么是 loop、什么不算是 loop,成本、models、loops 会怎么失败,以及怎么在 Claude Code、Cursor 和 Codex 这些主要工具里启动一个 loop。

Some of the sections include what a loop is and isn't, costs, models, and how loops fail, and how to fire a loop in major tools like Claude Code, Cursor, and Codex.
M1
M119:11

我们还会在 AIDB Operator Circle 里专门为参加 summer adventure 的人建一个新的子社区,我也会把进入那个社区的链接放到 show notes 里。

We'll also set up a new sub-community for people who are doing the summer adventure in the AIDB Operator Circle, and I'll include links to sign into that in the show notes as well.

最后收尾的时候我想说,贯穿整个二零二四年、尤其特别重要的一个概念,就是这种 capability overhang——也就是 AI 能做到的事情,和我们实际上拿它去做的事情之间的差距。

As we wrap up here, one of the concepts that has been most important throughout 2024 especially is this idea of the capability overhang, the gap between what AI can do and what we're actually using it for.

现在我想说的是,这个世界上几乎没有谁,不是在面对某种自己能力跟不上的落差。

Now I would contend that there is almost no one on the planet who doesn't have some capability overhang that they are dealing with.

哪怕是实验室里的研究人员,哪怕是那些全部工作就是关注 AI 里在发生什么、然后把所有新东西都试一遍的内容创作者,也一样。

Even researchers inside the labs, even content creators whose entire job is to pay attention to what is happening in AI and try out everything new.

我们想做的事情实在太多了,时间根本不够,而 AI 的能力还在一路飞快往前冲。

There is simply not enough time to do everything that we would like, and AI capabilities keep racing ahead.

这个 summer adventure 的目标,就是帮你用一些有趣的方式,缩小你自己的能力差距——不管这其中哪一部分对你来说最重要——而且是真的能在这个过程中玩得开心。

The goal of the summer adventure is to find fun ways for you to close your own personal capability gap, whatever part of that gap matters most to you, and actually have a good time doing it.

跟这种项目一贯的情况一样,这个 summer adventure 是完全免费的。

As always with programs like this, the summer adventure is completely free.
M1
M120:13

好,那今天这期就先到这里。也一如既往地感谢你收听或者收看,我们下次见,peace!

For now that's going to do it for today's episode. I appreciate you listening or watching, as always, and until next time, peace!
已剔除 7 处广告(点击展开查看)
M1
M10:17广告 · 已剔除

First of all, thank you to today's sponsors, Robots & Pencils, Rackspace, Blitzy, and Airtable.

To get an ad-free version of the show, go to patreon.com/ai-dailybrief or you can subscribe on Apple Podcasts.

To learn more about sponsoring the show, send us a note at [email protected].

M1
M11:08广告 · 已剔除

Secondly though, we're going to be introducing the latest training program from AIDB.

This one is at summeradventure.ai and is a choose your own adventure to expanding your AI skills.

We'll talk about how it works and give you a preview of some of the projects.

M1
M17:30广告 · 已剔除

One that requires not only learning new skills but reprogramming your brain, which is where now we get to the AI Summer Adventure from The AI Daily Brief and Superintelligent.

I have had a ton of fun over here releasing these various educational-type experiences throughout the year.

The first was AI DB New Year, which was a 10-week challenge that was meant to give you 10 basic skills of using AI in a way that could level you up over the course of the first period of the year.

Tons of people engaged with that, including many in teams, and given the context of what we were just discussing, it's almost quaint looking back how many of the skills from that were really in that pre-agentic paradigm.

Now, a couple months later, as we all got familiar with agents and harnesses, and Claude took over our brains.

Next up was Claude Camp.

This was a crawl-through-glass experience designed to help you build agents in a way that really led to the learners understanding the guts of these systems and the type of opportunities that they represented.

However, Claude was only one platform, and pretty quickly people stopped thinking about can I build an agent, to how do I use Claude Code or Codex or whatever harness I'm working within to build an actual agentic operating system that can help me do lots of different types of work.

That's what led to Agent OS, which is still, I would argue, a totally valuable program to go check out if you haven't yet, that once again is a free self-directed program for building an agentic system that can help you do lots of types of work.

Agent OS was designed by Nufar Gaspar, who also now leads our training programs at Superintelligent, including our executive catch-up program, and our executive agent leadership program, which is effectively the enterprise-grade version of Agent OS.

Now, when it comes to the AI Summer Adventure, this one is purposely designed to be more dynamic and fun and, well, choose your own adventure.

Building on the adventure and travel theme, it's organized into different destinations.

Each of the 20-plus destinations has a lab or a project that allows you to learn some new skill.

Some of them are for beginners, some of them are for intermediate users, and some of them are even advanced.

M1
M110:09广告 · 已剔除

That's one of the reasons I like today's sponsor, Robots & Pencils.

They've gone all in on AWS.

They're an advanced-tier AWS partner and they ship production AI coworkers in 45 days.

That's led to them doing some of the more interesting work I've seen on AI coworkers.

And by that I'm not talking about chatbots, I'm talking about actual agentic systems that sit inside a business architecture and do real work.

That kind of focus matters if you're an enterprise leader trying to get something real into production or an AWS rep trying to move a customer from interested to deployed.

Request an AI briefing at robotsandpencils.com.

One conversation with Robots & Pencils and you'll know.

One of the more interesting shifts in enterprise AI right now is how quickly the conversation is moving towards infrastructure and operations.

As AI moves into core workflows, regulated data environments and agentic systems, enterprises need governed infrastructure and inference that can operate reliably day to day with clear operational accountability built in from the start.

As those systems scale, the operating model increasingly becomes part of the AI strategy itself.

Rackspace Technology is the operator of the full enterprise AI stack, from agents to infrastructure across private cloud, hybrid cloud and edge environments.

Rackspace builds and operates governed AI infrastructure, inference and production AI systems for organizations where sovereignty, compliance and uptime are non-negotiable.

Their deployed engineers stay embedded beyond deployment to help operationalize and run AI in live environments.

To learn more about where enterprise AI runs and outcomes scale, go to rackspace.com.

Weekends are for vibe coding.

It has never been easier to bring a passion project to life, so go ahead and fire up your favorite vibe-coding tool.

But Monday is coming and before you know it, you'll be staring down a maze of microservices, a legacy COBOL system from the 1970s and an engineering roadmap that will exist well past your retirement party.

That's why you need Blitzy, the first autonomous platform designed for enterprise-scale codebases.

Deploy at the beginning of every sprint and tackle your roadmap 500% faster.

Blitzy's agents ingest your entire code base, plan the work, and deliver over 80% autonomously.

Validated, end-to-end tested, premium quality code at the speed of compute, months of engineering compressed into days.

Vibe code your passion projects on the weekend, bring Blitzy to work on Monday.

See why Fortune 500s trust Blitzy for the code that matters at blitzy.com.

That's blitzy.com.

This episode of the AI Daily Brief is brought to you by HyperAgent, where you run fleets of agents your team can manage together.

New users get $1,000 in inference.

Forget local agents and chat workflows waiting on your laptop to be prompted.

HyperAgent deploys always-on agents in the cloud, doing real work across the tools your team already uses.

Marketing's agent turns competitor moves into landing pages.

Sales' agent enriches leads, drafts emails, and updates the CRM.

Ops' agent chases the paperwork and tracks the budget.

Every agent has access to shared context and follows your rules about scope and approvals.

It's time you add agents that feel like teammates.

Hire yours at HyperAgent, built by the team at Airtable.

Claim your $1,000 in inference at hyperagent.com/ai-dailybrief.

M1
M113:26广告 · 已剔除

We've brought more of that back with AI Adventure, and so if you have friends or colleagues or even yourself have just areas where you feel really behind, even some of the beginner quick trips might be something that you want to check out.

M1
M118:43广告 · 已剔除

Now, once you listen to this show, summeradventure.ai is going to be fully live.

You can sign up for free and start taking it in.

So far, about half of the destinations are unlocked, and each week here through the end of the summer in the US, i.e. basically the beginning of September, we'll be unlocking even more projects.

Some of those projects we already have planned, but some of them we're anticipating building on the fly based on what's happening in the industry.

Who knows, maybe we'll go build ourselves a model router, given that it seems like everyone else on the planet has at this point.

M1
M120:09广告 · 已剔除

You can sign up at summeradventure.ai, and I'm excited to see you there.