探索增强版Codex的功能

欢迎收听本播客。

Welcome to the podcast.

我是你的主持人，贾登·谢弗。

I'm your host, Jaden Shafer.

今天在节目中，我想谈谈OpenAI如何应对Anthropic大规模功能推送所带来的负面舆论，同时也要介绍一款刚刚发布的新工具——Claude Design。

Today on the show, I wanna talk about OpenAI fighting back about against the giant onslaught of features Anthropic has been pushing some negative PR Anthropic has been getting, but at the same time, an incredible new tool called Claude Design that just came out.

我还想聊聊风险投资在AI领域的去向，那里有一些出人意料的有趣动向，以及一个新术语——令牌最大化。

I also wanna talk about where some VC dollars are going in the AI space, some surprisingly interesting things there, and a new term called token maxing.

此外，OpenAI刚刚大幅增强了Codex的桌面控制、内存功能和应用内浏览器，并支持超过100个插件集成。

In addition, OpenAI just massively beefed up Codex for desktop control, memory, and in app browser, and over a 100 plug in integrations.

这本质上是他们直接针对Anthropic的Claude Code和Claude Cowork发起的反击，我认为这对未来编辑型智能体的发展方向至关重要，让我们深入探讨一下。

Basically, this is them swinging directly at Anthropics Claude Code and Claude Cowork, and I think it matters a lot for where editing agents is gonna be going in the future, so let's get into it.

在开始之前，我想提一下AI Box。

Before we do, wanted to mention AI Box.

我经常听到人们说，他们正在为ChatGPT、Claude、Gemini、Perplexity甚至Midjourney付费，而当把这些费用加总起来时，每月的开销很可能达到70到80美元，还要管理多个账户。

The thing I keep hearing from people is that they're paying for ChatGPT, Claude, Gemini Perplexity, even Midjourney, and by the time you get all of that added up, you're likely 70 or $80 a month across a bunch of different logins.

AI Box 让你在一个界面中访问超过80个不同的AI模型。

AI box gives you access to over 80 different AI models in one interface.

所有这些服务每月只需8.99美元。

All of this is just $8.99 a month.

如果你想使用，可以在描述中找到 aibox.ai 的链接。

If you wanna get access to it, there's a link description to aibox.ai.

此外，我们还有一个叫 AI Box 构建器的功能，你可以将多个AI模型串联起来。

And in addition, we have something called the AI box builder where you can essentially link together multiple AI models.

我们为你搭建完整的流程，让你无需任何编程知识就能轻松创建工具。

We build up the entire workflow for you, and you vibe vibe build tools without needing to know any code at all.

我不是开发者，我做这个是为了帮助其他非开发者的人。

I'm not a developer, and I built this for other people that are not developers.

所以如果你想试试，描述里有 aibox.ai 的链接。

So if you wanna check it out, there's a link in the description to aibox.ai.

好的。

Okay.

我想先谈一家叫Factory的公司。

The first thing I wanna talk about is a company called Factory.

这是一家AI编程初创公司。

So this is an AI coding startup.

他们专门专注于企业工程团队。

They're focused specifically on enterprise engineering teams.

他们刚刚完成了1.5亿美元的A轮融资，估值为15亿美元。

They just closed a $150,000,000 series a at a 1,500,000,000 valuation.

Kochsla Ventures领投了这一轮。

Kochsla Ventures led the round.

红杉资本、Insight Partners和黑石都参与了本轮投资。

Sequoia Insight Partners and Blackstone all were participating in this.

创始人名叫Matten Grinberg。

The founder is named Matten Grinberg.

他曾是加州大学伯克利分校的物理博士生。

He was a physics PhD student at Berkeley.

他于2023年直接给红杉资本的合伙人肖恩·马奎尔发了冷邮件，据说他们关系很好。

He basically cold emailed Sequoia partner Sean Maguire in 2023, and they apparently were good friends.

他们因为物理研究而建立了联系。

They bonded over physics research.

马奎尔劝他退学，红杉资本随后对这家公司进行了种子轮投资。

Maguire convinced him to drop out, and Sequoia seeded the company.

他们的客户名单已经包括摩根士丹利、安永和帕洛阿尔托网络公司。

Their customer list already included Morgan Stanley, Ernst and Young, and Palo Alto Networks.

所以，很明显，这是一家非常专注于企业市场的公司。

So, obviously, this is a very, enterprise focused.

他们根本不是在针对个人开发者。

You know, they're not targeting individual developers in any way.

他们的目标是企业市场。

They're focused on the enterprise.

格林伯格的论点是，Factory之所以能脱颖而出，是因为其模型的灵活性。

Grinberg's pitch is why Factory, I think, kind of differentiated their model flexibility.

他们基本上允许你在Claude、DeepSeek之间切换，选择最适合的工具。

They basically can let you switch between Claude, DeepSeek, whatever makes sense.

不过说实话，Cursor也能做到这一点，我认为目前大多数主流玩家也都具备这个功能。

Although, honestly, Cursor does that too, as do, I think, most of the serious players at this point.

我认为这告诉我们，即使Anthropic、OpenAI和Cursor已经进入市场，企业级AI编程领域仍然有空间留给一些专注于特定领域的玩家。

What I think this does tell us is that even with Anthropic and OpenAI and Cursor already in the market, enterprise AI coding still has room for some category specific players.

摩根士丹利不会允许某个随机的开发者工具运行在他们的网络内部，除非这个工具是根据他们的合规性和安全要求构建的。

Morgan Stanley isn't gonna let some, you know, random developer tool run inside their network unless it's built with their compliance and security posture in mind.

我认为这正是Factory正在填补的空白。

And I think that's basically the gap that factory is filling.

我认为他们目前15亿美元的估值表明，风投们相信这里确实存在一个真正的市场缺口。

And I think $1,500,000,000 in their current valuation says that VCs are believing there is a real gap here.

当然，Claude Code也在试图进入这个领域。

Of course, Claude Code is trying to get in there as well.

你可以看看Cognizant这样的公司，刚刚让他们的全部35万名员工都使用Claude和Anthropic的所有工具。

And you can look at things like Cognizant just getting, you know, all of their employees, 350,000 of them onto Claude and and all the Anthropic tools.

所以我认为可能会有很多竞争者，但有趣的是他们正在开辟一个细分市场。

So I think there's probably competition from a lot of players, but it's interesting that they're carving out a niche there.

好的。

Okay.

接下来我想谈谈Anthropic推出的一个全新工具，叫做Claude Design。

The next thing I wanna talk about is a brand new tool from Anthropic called Claude Design.

目前它还处于研究预览阶段。

It's a research preview right now.

它面向Pro、Max、Teams和企业订阅用户开放，由刚刚在一两天前发布的Claude Opus 4.7模型驱动。

It's available to pro, max, teams, and enterprise subscribers, and it's powered by a Claude Opus 4.7, the model that just came out a day or two ago.

这就是Anthropic刚刚发布的内容。

So this is what Anthropic just shipped.

基本上，你可以描述你想要的内容，比如演示文稿、一页纸摘要或着陆页原型，Claude会生成一个初稿。

And, basically, you can describe what you want, a pitch deck, a one pager, a landing page prototype, and Claude generates a first draft.

你可能已经见过它之前生成网页的样子。

You've probably seen it kinda make web pages before.

这很有趣，因为你实际上可以提前用Claude Design来制作原型，然后通过直接编辑或与它对话来进一步完善。

So it's interesting because you actually can use Claude Design to kinda come up with the mock ups ahead of time, and then you can refine it by either directly editing it or just talking to it.

我其实很欣赏这两种方式，因为我用过很多工具，我的意思是，我不想过多批评Level，因为我知道它们确实有一些直接编辑功能。

I actually appreciate both of these options as I've used a lot of tools, like, I mean, I wanna throw too much shade at Level, because I know they do have some direct editing features.

但过去这些功能对我来说从来都没怎么好用过。

It actually never worked super good for me in the past.

也许现在它们变好了。

Maybe they're better now.

但过去，Level会允许你描述你想构建的网站或任何东西。

But, in the past, Level would have something where you could, you know, describe the website or whatever you're trying to build.

它会生成设计稿。

It would generate the design.

然后你本应能点击它进行直接编辑。

And then you're supposed to be able to click on it and edit directly.

但从来都不管用。

It never worked.

而且有时候我在那之后进行聊天时，它会把那些修改撤销掉。

And what sometimes when I do the chat after I do that, it would, like, undo them.

这真的挺糟糕的。

It was just kind of bad.

所以我觉得Claude在这方面做得更好一些。

So I think Claude has cracked this a little bit better.

也许Lumbwell也有类似功能。

Maybe Lumbwell's there as well.

但你可以将结果导出为PDF、URL、PPTX文件，并直接发送到Canva。

But you're able to export as PDFs, URLs, PPTX files, and you can send the outputs straight to Canva.

所以Canva和他们有深度集成，你可以把所有协作内容都保留在那里。

So Canva has a big integration with them, and you can keep all of your collaborations there.

它还能读取你公司的代码库和设计文件，从而在所有输出中保持一致的设计体系，从技术角度来看，这其实更有趣。

It can also read your company's code base and design files to apply a consistent design system across all of your outputs, which is actually, I think, the more interesting piece if you kinda look at this technically.

Anthropic将此定位为对Canva的补充。

Anthropic is positioning this as kind of complementary to Canva.

他们并不是在说，你看。

They're not they're saying, like, look.

我们不会和他们竞争。

We're not gonna compete with them.

这是一种赞美。

It's it's a compliment.

目标用户明确是那些非设计师的人，比如创始人、产品经理或需要快速做出专业外观的初创公司运营者。

The target audience is specifically people who aren't designers, so founders or PMs, startup operators who need to make something look presentable really fast.

我认为这一点表明，Anthropic 正在持续向上游拓展。

What I think about this is that Anthropic is continuing to move up the stack.

今年早些时候，他们推出了 Clot Co Work，然后为特定部门推出了智能代理插件，我想现在又推出了这个。

Earlier this year, they launched clot co work, then agentic plugins for specific departments, and I think now this.

我觉得他们的发展方向是，不仅仅想做一个 API 公司。

I think where they're going with this is that they're not just trying to be an API company.

他们真正想掌控实际的工作流程和使用场景。

They wanna actually own actual workflows and surface area.

这和OpenAI一直在做的策略是一样的，我想等我们稍后深入探讨时，你会更明白为什么这很重要。

It's the same play that OpenAI has been making, and I think you're gonna hear more why this matters when we go into the deep dive later on in the episode.

但也要向谷歌致敬，他们几乎在每一个垂直领域都在这么做。

But also just show shout out to Google who's been doing this in basically every vertical.

谷歌有Stitch，也是一个非常相似的设计工具。

Google has Stitch, which is a very similar design tool as well.

所以，我认为我们会看到这些玩家更多地深入到软件本身，而不仅仅是模型，这相当有趣。

So, yeah, I think we're gonna see a lot of these players get more into the software itself beyond just the models, which is pretty interesting.

好的。

Okay.

接下来我想谈的是令牌消耗问题。

The next thing I wanna talk about is token maxing.

最近在TechCrunch上有一个有趣的故事，讲的是令牌消耗的问题。

So there's a funny story on on TechCrunch recently where they're talking about token maxing.

基本上，这是一种模式：公司和开发者在炫耀他们的AI编程工具消耗了多少令牌，仿佛用的令牌越多，就意味着他们越高效。

Basically, it's the pattern of companies and developers when they're bragging about how many tokens their AI coding tools burn through, as if, you know, the more tokens that they're using means that they're more productive.

我认为他们实际处理的数据非常不同。

I think the actual data that they've been crunching is very different.

像ClaudeCode、Cursor Codex这样的工具，在初次生成时就能获得高达80%到90%的代码接受率。

So tools like ClaudeCode, Cursor Codex, they're all generating way more accepted code at the on the first pass, 80 to 90% of initial acceptance rates.

但当你两周后回头看同样的代码时，实际接受率会下降到10%到30%，因为工程师们一直在重写它。

But when you look at the same code two weeks later, the effective acceptance drops to 10 to 30% because engineers were constantly rewriting it.

对吧？

Right?

所以，实际情况是，当他们说：看，我们50%的代码都是由AI生成的。

So basically, what's going on here is when they're like, look, we're we're writing, like, all of our code, 50% of our code.

很多公司都说，我们50%的代码是Claude写的，听起来很厉害。

A lot of companies are like, 50% of our code's written by, like, Claude, and it sounds amazing.

他们还说，这些代码全都完美无缺，或者至少大部分都很优秀。

And they're like, and it's all perfect and great or, you know, a high level of it.

也许确实如此。

And maybe that's true.

对吧？

Right?

我不是说这一定就是坏事。

And I'm not saying that that's necessarily bad.

我也在我的初创公司里大量使用Claude Code。

I use Claude code heavily as well with my startup.

但有趣的是，那些把这当作生产力指标的人其实是在误导。

But what's interesting is I I think people that pretend it's, you know, basically a marker of productivity.

因为GitClear的研究发现，AI用户代码的 churn 量是非AI用户的9.4倍。

Because what they found, GitClear, I was doing this specifically, and they found that AI users have 9.4 times higher code churn than non AI users.

FARO AI也发现，在AI高度应用的情况下，代码 churn 增加了861%。

And FARO AI found code churn increased 861% under high AI adoption.

而Jellyfish对7500名工程师的研究发现，拥有最大token预算的团队，吞吐量是两倍，但token成本却高出十倍。

And then you also had Jellyfish, which was looking at 7,500 engineers and found that the teams with the biggest token budgets got two times the throughputs at 10 times the token cost.

我认为这本质上是一种现实的提醒。

And I think basically it's a just a reality check.

对吧？

Right?

AI编码带来的生产力提升是真实的，但它们只是原始输出数字所显示的一小部分。

The productivity gains from AI coding are real, but they're also a fraction of what the raw output numbers suggest.

对吧？

Right?

如果你觉得，我现在一天能写一百万行代码，而以前只能写这么一点。

If you're like, I can write a million, you know, lines of code a day now, and I used to only be able to write this amount.

好吧。

Okay.

当然，你确实获得了真实的生产力提升。

Well, the you definitely are getting real productivity gains.

这并不是在反驳，但我只是认为，我们需要自我审视一下，那一百万行代码并非都是优质的，因为如果你三周后再看，很大一部分都需要重写或修复，这没关系。

It's not an argument, but I just think it's important to to check ourselves and and, you know, the the million lines of code are not all good because if you look at it three weeks later, a big chunk of it has to be rewritten or fixed, which is fine.

我的意思是，我们正常的开发者写代码后，会阅读、处理、优化和修复它。

I mean, our our normal developer writes code and read you know, and and works on it and and optimizes it and fixes it.

有趣的是，资深工程师对AI生成代码的接受度反而比初级工程师更低，可能因为他们更清楚哪些部分存在细微错误。

I think senior engineers, interestingly, are less accepting and than AI of AI code than juniors, and probably because they know which parts are subtly wrong.

对吧？

Right?

所以当有代码推送时，他们更不太可能接受这些推送。

So when there's like a code push, then they're less likely to accept that code push.

如果你是一名管理者，正在思考如何衡量AI的投资回报率，我认为重要的是统计被合并和部署的代码量，而不仅仅是生成了多少代码。

If you're a manager thinking about how to measure AI ROI, I think counting merged and shipped, is important and not just how much is generated.

好的。

Okay.

让我们来谈谈物理智能。

Let's get into physical intelligence.

这是一家机器人基础模型初创公司。

This is a robotics foundation model startup.

它刚刚发布了关于名为PIE 0.7的新模型的研究。

It just published research on a new model called PIE 0.7.

我认为这可能是今天最富有创新性的技术故事，基本上，他们目前声称的是，PIE 0.7 能够通过组合在其他情境中学到的技能，完成从未专门训练过的任务。

And I think this might be the most novel kind of technical story of the day, but basically, what they're claiming right now is that PIE 0.7 can perform tasks it was never specifically trained on by composing skills it learned in other con other contexts.

因此，他们强调的一个例子是，他们有一个空气炸锅，而机器人在训练中只短暂见过这个空气炸锅。

So the example that they're kind of highlighting all of this, they have like an air fryer, and the robot had only briefly seen this air fryer in training.

所以，它并没有被训练过如何操作空气炸锅。

So, you know, it wasn't trained to operate the air fryer in any way.

它在训练中只是瞥见过几次空气炸锅，我想大概只有两个短短的视频片段。

It briefly seen it in training, and, I think it only seen like two short clips of an air fryer.

然后他们给了它一个逐步的口头指令，它就弄明白了如何操作它。

Then they gave it a step by step verbal instruction, and it figured out how to operate it.

在更广泛的测试中，这个通用模型在制作咖啡、叠衣服和组装纸箱等任务上的表现，甚至与专用模型不相上下。

And in in like some broader testing, the generalist model actually matched specialized models on jobs like making coffee, folding laundry, and assembling boxes.

研究人员表示，这种泛化能力让他们感到非常惊讶。

Researchers said that the generalization ability was really surprising to them.

对吧？

Right?

所以基本上，你专门训练机器人来叠衣服，它就能做得很好。

So basically, you train a special you you train the robot specifically to fold laundry, and, yeah, it does good.

然后他们有了一个新模型，这个模型在各种任务上都具备通用能力。

And then they have a new model where it's just kind of a generalist at everything.

它并没有被专门训练来叠衣服，但只要你一步步地指导它如何叠衣服，它就能做得和专门训练过的机器人一样好，或者几乎一样好。

It's not trained to fold laundry, but you explain step by step how to fold laundry, and it does just as good as the robot or almost just as good as the robot that was specifically trained on this.

这太令人着迷了。

That is fascinating.

而且我认为，当你看看像OpenAI或Anthropic这样的所谓通用模型时，它们能普遍出色地完成多种任务，这对它们来说是个好消息，因为对于物理机器人这类应用，你可能不再需要为每个特定任务单独训练模型了。

And I think especially when you look at, you know, I would say, quote unquote, general models like OpenAI or Anthropic that they're building that can do a lot of different things generally good, it's kind of good news for them because, you know, you you may not have to have models specifically trained on just a specific task when he just talks about, you know, kind of like physical robotics and stuff.

在商业方面，Physical Intelligence公司已经筹集了超过10亿美元。

So on the business side, physical intelligence has already raised over $1,000,000,000.

它们上一次估值为56亿美元。

They were last valued at 5,600,000,000.

据报道，它们目前正在谈判，希望将估值几乎翻倍至110亿美元。

They're reportedly in talks to nearly double that to 11,000,000,000.

他们的联合创始人莱西或拉奇·格鲁姆有投资Figma、Notion和Ramp的成功记录。

Their cofounder, Lacey or, Latchy Groom has a track record of backing Figma, Notion, and Ramp.

所以我认为，这就是风险资本资金流向拉奇的原因。

So I think, this is why VC dollars are going to latchy.

我认为有一点值得注意，如果我要诚实地说的话，实际上pi 0.7仍然无法处理很多多步骤任务。

I think something that is interesting, a caveat that I'll put on this if I'm trying to be honest, like, basically, pi 0.7 still can't handle a lot of multistep tasks.

对吧？

Right?

它并不是在没有任何指导的情况下自主完成这些任务的。

It's not doing this autonomously without any coaching.

我认为机器人领域并没有像大语言模型那样清晰的评估标准。

I think the robotics field doesn't really have a lot of clean benchmarks like LLMs do.

你知道，我们在人文、期末考试等方面有各种各样的评估标准，可以准确衡量AI模型在工程、数学和其他领域的表现。

You know, we have like humanities last exam and like all these different benchmarks that we give AI models on, you know, like engineering and math and other areas, we can tell exactly how good they are at those tasks.

但在机器人领域，这样的评估标准还很少。

There's not a lot of that with robotics.

我相信随着我们更深入地了解，还会出现更多这样的情况。

I'm sure there's gonna be more as we we get more into it.

所以对于机器人来说，你基本上只能相信它们的演示效果。

So basically for robotics, so you kind of have to just trust whatever their demo is.

但我认为，如果这种通用行为将成为我们关注的重点，那将大大推动我们迈向能够在真正混乱的现实环境中工作的机器人。

But I think if this kind of generalized behavior is gonna be something that we're looking at, it's pretty significantly stepping us towards robots that actually work in really messy real world environments.

因此，未来六个月左右，我会密切关注这一点。

So this is something I'll be closely closely watching over the next six months or so.

好的。

Okay.

OpenAI 刚刚为其桌面应用发布了一系列新功能，这款应用叫做 Codex。我过去曾尝试过，但最近几周，我选择了 Anthropix、Cloud Code 和 Cloud Cowork。

OpenAI has just released a new a whole bunch of new features to their desktop app, which is called Codex, something I've tried in the past but have opted for Anthropix, Cloud Code, and Cloud Cowork in recent, you know, weeks, in the last month.

我认为 OpenAI 意识到了这一点，他们非常希望大力推广，吸引用户重新使用 OpenAI Codex，或者让首次尝试的用户来体验它。

And I think OpenAI sees that, and they really want to make a big push to win people back or to have people try OpenAI Codex for the first time.

Codex 进行了重大升级。

So huge upgrade to Codex.

我来介绍一下几个新功能，因为我认为OpenAI这是直接针对Anthropic的Cloud Code，而后者最近确实表现得非常出色。

I'll I'll walk through a couple of things that are new because I think OpenAI is basically swinging directly at Anthropix Cloud Code, which honestly has been really crushing it.

对吧？

Right?

这就是他们新增的内容。

So this is what they added.

首先，Codex现在可以在你的Mac后台运行，这太棒了。

First, Codex can now run-in the background on your Mac, which is phenomenal.

对吧？

Right?

它可以打开应用程序。

It can open up applications.

它可以进行点击操作。

It can click around.

它可以在你处理其他事情的同时，在桌面上输入内容。

It can type into your desktop while you keep working on something else.

这其实是我喜欢的一点。

This is actually something I like.

Claude 大致也能做到这一点，但说实话，即使使用 Claude Co-Work，很多时候当我有自动化任务运行时——比如每天早上9点做一件事，中午12点做另一件事，它会抓取分析数据、获取信息或生成报告——

Claude sort of does this, but I'm gonna be honest, even with Claude co work a lot of times, if I have, like, an automated task running, like I've got a bunch of things that I'm just like, you know, every day at 9AM do this, every day at noon do this, and it's like grabs like analytics or grabs data or goes and gets me a report on something.

所以我最常用它的地方是：当某个服务没有 API 时，我会让它登录账户，自己去获取我需要的数据，然后返回给我。

So, actually the thing that I love using it for is if there's no API for a service, I'll just have it go log in to the account and go grab the data I need and bring it back to me.

希望这些公司未来能提供 API，但目前这就是我的做法。

Hopefully those companies offer APIs in the future, but for now that's what I do.

无论如何，Cloud Co-Work 的一个烦人之处是，当这些自动化任务开始执行时，Chrome 浏览器经常会突然弹到我屏幕上——如果它无法在后台完成操作，就会直接在我眼前点击各种东西，我得像赶苍蝇一样把它赶走，同时还得继续做别的事。

In any case, it is annoying with Cloud Cowork that lots of times when those automated tasks start happening, all of sudden this, Chrome browser pops up on my screen if there's if it can't do it like in the background and all of a sudden it's like clicking on things right in front of me and I'm like swatting flies, like trying to get this thing away while I keep working on something different.

我该给它单独配一台电脑吗？

Should I have its own computer?

有可能，但很多时候我只是让它在旁边运行。

Like possibly, but a lot of the times I just have it running on the side.

无论如何，OpenAI 正在努力解决这个问题，让这些操作能在后台进行。

In any case, OpenAI is trying to, combat that and have it work on things in the background.

所以这不仅仅是写代码，而是真正操作你的整个机器。

So it's not just writing code in an editor, it's actually operating your entire machine.

这就是我之所以对OpenAI的计算机使用功能感到兴奋的原因，他们其实早就开始做这件事了。

So this is what I'm excited about, computer use from OpenAI, which they sort of have done for a long time.

在Anthropic开始在这方面有所突破之前，他们就是先驱，但那时他们的表现真的显得陈旧而糟糕。

They were the OGs way before Anthropic was shifting things in this, but they really just felt stale and bad.

我过去试过很多次。

I've I've tried a lot in the past.

相信我，如果几年前OpenAI的智能体像现在Anthropic做得这么好，我早就到处宣传了。

And trust me, if OpenAI agents back in the day were, you know, like a like six months, a year ago, were as good as what Anthropic's doing now, I'd be, you know, shouting them from the rooftops.

但看起来现在他们正在卷土重来。

But it seems like now they're making a comeback.

因此，它们还可以在不干扰你桌面的情况下并行运行多个智能体，这意味着你可以同时让一个修复bug，另一个运行测试，还有一个撰写文档。

So they can also run multiple agents in parallel without interfering in your desktop, which means that you can have one fixing a bug, and you can also have one running tests, one writing docs all at the same time.

它们还内置了一个新的应用内浏览器，可以直接访问网页应用。

They also have a new in app browser, so it can hit web applications directly.

他们有111个插件集成，比如Coderebbit、GitHub或GitLab问题。

They have a 111 plug in integrations, so Coderebbit, GitHub, or GitLab issues.

他们的记忆功能带来了一系列令人兴奋的新功能，可以记住之前的会话。

They have a bunch of new exciting things with their memory feature, so it can remember previous sessions.

他们现在在Codex中集成了图像生成功能，老实说，Claude并没有任何图像生成功能。

They have an image generation that is now inside of Codex, which to be fair, Claude does not have any sort of image generation.

他们还为企业和商业客户推出了按需付费的定价模式。

They also rolled out a pay as you go pricing specifically for enterprise and business customers.

所以虽然我认为Anthropic目前肯定领先，但我会说，插件生态系统可能是整个公告中最被低估的部分之一，因为他们发布时就有111个不同的插件。

So while I think Anthropic is definitely ahead right now, I would say that the plug in ecosystem is probably part of one of the most underrated pieces of this entire announcement because they they have, like, a 111 different plug ins at launch.

他们还会继续增加更多。

They're gonna be adding more.

至于Claude Cowork，它确实很棒，但我这里可能只有四样东西，比如我的Google日历和Chrome同步了。

And with Claude Cowork, I mean, it's awesome, but I have, like, maybe, like, four things, you know, my Google Calendar and Chrome synced up.

所以是一些Google工具和GitHub。

So a bunch of Google tools and GitHub.

但除此之外，我用的很多工具都和它集成得不太好。

But beyond that, there's so many different tools I use that don't integrate very well with it.

我得打开Chrome浏览器才能使用它们。

It's gotta go and use my Chrome browser to access them.

所以，我认为OpenAI引入的这些集成功能将会非常有用。

So anyways, I think a lot of these integrations that OpenAI is pulling in are going to be very useful.

好了。

Alright.

这就是本期节目。

That is the show.

如果你从这些节目中获得了价值，请在Apple Podcasts上留下评论，或在Spotify上给几颗星。

If you're getting value from these episodes, please drop a comment over on Apple Podcasts or leave a couple stars over on Spotify.

你可以在Spotify的‘关于’页面提交评价。

You hit the about tab on Spotify to to drop a review.

基本上，评价有助于让节目触达更多人。

Basically, the reviews help the show reach way more people.

这会让算法推荐提升。

It boosts in the algorithm.

这对它帮助非常大。

It helps it out a ton.

如果你还没做的话，我会非常感激。

If you haven't done it already, I would be eternally grateful.

另外，如果你想整合你已经付费的订阅服务，去看看 aibox.ai 吧。

Also, if you wanna consolidate the subscriptions you're already paying for, go check out aibox.ai.

描述里有链接。

There is a link in the description.

80 多个模型，用通俗语言设置自动化，每月 8.99 美元。

80 plus models, plain English automations, $8.99 a month.

我们下一期再见。

I'll catch you guys all in the next episode.