← 返回任务列表

What Gemini 4 Argon Is Actually Good For

318 段 · 1 位说话人 · 原片 30:24
F1
F10:00

Google 的新模型 Gemini 4 Argon 在发布两周后,现在终于开放测试了,大家疑问特别多。

2 Weeks after being announced, Google's new Gemini 4 Argon model is here, available for testing, and the questions abound.

它真能达到那么高的基准测试分数吗?

Will it live up to its high benchmarks?

Google 这次在 agentic coding 方面能不能终于追上来?

Can Google finally compete when it comes to agentic coding?

如果不是 agentic coding,那有没有其他场景 Google 现在反而领先了?

If not agentic coding, are there other use cases where Google might now be in the lead?

今天我们来看看这个我们能接触到的最新前沿模型——Gemini 4 Argon,以及它的定位。

Today we're looking at the newest frontier model we have access to, Gemini 4 Argon, and where it fits.

AI Daily Brief 是一档每日播客和视频节目,关注 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.

好了朋友们,在深入话题之前先做几个快速预告。

All right, friends, quick announcements before we dive in.
F1
F11:15

最后说一句,虽然也不是非看不可,但这一期用视频形式看会更有收获,因为有很多 Gemini 4 Argon 的演示是视频的。

Last note today, it's not completely essential, but there are some benefits to consuming this one as a video, given that a lot of the demos of Gemini 4 Argon are video.

这期节目在 YouTube 和 Spotify 上都有视频版,Apple 那边也很快会上线视频。

The episode will be in video on both YouTube and Spotify, and video will be coming to Apple soon as well.

今天我们先从一个内部八卦说起。

We kick off today with some palace intrigue.

周二下午,OpenAI 新闻室账号发帖说:我们发现了一个藏在办公室里的高级录音设备。

On Tuesday afternoon, the OpenAI Newsroom account posted, we discovered a sophisticated recording device that was hidden inside one of our offices.

我们用自家模型分析了这个设备以及它的来源,最终确认是某位员工几天前放进去的。

We used our models to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier.

我们把这个案子移交给了 FBI,非常感谢探员和检察官们迅速行动,抓到了嫌疑人。

We referred the matter to the FBI and were grateful to the agents and prosecutors who moved quickly to make an arrest.

长话短说,一名 OpenAI 员工因为在一间高管会议室发现窃听器而被捕并被起诉。

TL;DR, an OpenAI employee has been arrested and charged after a listening device was discovered in an executive conference room.

根据司法部起诉书,大约两周前,OpenAI 在旧金山办公室的一间会议室里发现了一个伪装在厕纸里的录音设备,插在电视背后。

According to a DOJ complaint, around 2 weeks ago, OpenAI discovered an audio recording device disguised in toilet paper and plugged into the back of a TV in a conference room at their San Francisco offices.

不难想象,那间会议室经常讨论高度机密的商业、技术和战略信息。

As you might imagine, meetings in that room routinely involved discussion of highly confidential technical, commercial, and strategic information.

后来监控录像显示,员工 Abdulrahman Al-Fozan 在设备被发现之前多次出现在那间会议室附近。

As would come out later, security footage showed employee Abdulrahman Al-Fozan near the conference room on multiple occasions prior to the discovery of the device.

Abdulrahman 并没有权限进入那间会议室或参加里面的会议。

Abdulrahman had no authority to access the room or participate in the meetings held there.

Al-Fozan 在 OpenAI 工作了大约一年,职位是 agentic 产品的工程负责人。

Al-Fozan had been employed by OpenAI for around a year as an engineering lead working on agentic products.

他之前在 Meta 工作了五年。

He had previously worked for Meta for 5 years.

起诉书称,Al-Fozan 用一款效率工具查询了 OpenAI 的会议室布局、高管位置以及与外部伙伴开会的细节。

According to the complaint, Al-Fozan used a productivity tool to make inquiries about OpenAI's meeting room layout, executive locations, and details of meetings with external partners.

Al-Fozan 配偶名下的一个聊天机器人也被用于搜索如何搭建连续录音设备、定制 Wi-Fi 连接、改善音频和实时监听、加密 SD 卡等信息。

A chatbot owned by Al-Fozan's spouse was also used for queries related to building a continuous audio recorder, customizing Wi-Fi connectivity, improving audio and live listening, and encrypting the SD card, among other things.

该账号还搜索了被发现的具体设备型号,通过客户标头和唯一的 MAC 地址确认了身份。

The account was also used for queries about the specific device that was found, identified by its customer header and unique MAC address.

这些搜索都发生在设备被发现前的一周内。

These queries all took place in the week prior to the device being found.

最后在 10 月 5 日,Al-Fozan 进入会议室检查了电视背后,这一幕被摄像头拍了下来。

Finally, on October 5th, Al-Faza'an entered the conference room and checked behind the TV, a moment that was captured on camera.

同一天,OpenAI 安保人员搜查了一个无人看管的包,在里面发现了 Al-Fozan 的钱包和身份证件,以及另一台类似型号的录音设备。

The same day, upon searching an unattended bag, OpenAI security personnel found Al-Faza'an's wallet and ID, as well as a second recording device of a similar make.

Al-Fozan 被指控试图使用电子设备窃听口头通讯,FBI 正在进一步调查。

Al-Faza'an has been charged with attempting to use an electronic device to intercept oral communications, and the FBI is conducting further investigations.

如果这只是普通的商业间谍案,那也够我们这些吃瓜群众津津乐道了。

If this were just corporate espionage, there would be plenty to keep all of us in the peanut gallery interested.

但这案子之所以更上一层,是因为 Al-Fozan 是沙特公民,人们怀疑这不只是商业间谍,而是直接的国际间谍活动。

What takes this case to the next level, however, is Al-Faza'an's status as a Saudi national, the suspicion being that this is not only corporate espionage, but straight-up international espionage.

司法部的公告直言不讳地称这可能是一起国家安全事件。

The DOJ announcement was overt in calling this a potential national security incident.

联邦调查局局长Kash Patel表示:“根据今天解封的起诉书,Abdulrahman Al-Faza'an涉嫌在其雇主的行政会议室里藏匿了一个隐蔽的录音设备,秘密捕捉关于公司超级智能技术的敏感通讯。”

FBI Director Kash Patel said, "According to the complaint unsealed today, Abdulrahman Al-Faza'an allegedly concealed a covert recording device inside his employer's executive conference room to secretly capture sensitive communications about the company's superintelligence technology."

保护美国的超级智能创新对于维护我们国家的经济实力和国家安全至关重要,联邦调查局将这一使命视为我们最高反间谍优先事项之一。

Protecting American SI innovation is essential to safeguarding our nation's economic strength and national security, and the FBI views that mission as one of our highest counterintelligence priorities.

Patel指出,国家情报总监Jay Clayton一直在密切参与此案——当然,Clayton也是新任命的管理层AI特别工作组负责人。

Patel noted that National Intelligence Director Jay Clayton has been working closely on the case— Clayton, of course, also being the newly appointed head of the administration's AI task force.

美国检察官Craig H. Misikian补充道:“湾区和硅谷在开发未来技术方面引领世界,我们的办公室将继续站在保护美国创新的最前沿。”

U.S. Attorney Craig H. Misikian added, "The Bay Area and Silicon Valley lead the world in developing tomorrow's technology, and our office will continue to stand at the forefront of protecting that American innovation."

对我们来说,这是国家安全问题。

For us, it is a matter of national security.

整个湾区,特别是AI行业,长期以来一直是国际间谍活动的目标。

The Bay Area in general and the AI industry in specific have long been a target of international espionage.

这次情况有些不同在于,沙特阿拉伯——名义上是美国盟友——被牵连其中。

What makes this somewhat different is Saudi Arabia, nominally a U.S. ally, being implicated.

X上的记者们开始深入挖掘背后的故事。

Journalists on X got to work unpacking the backstory.

Newsground的Scott Steadman表示,根据他的消息来源,Al-Faza'an是OpenAI唯一的沙特员工。

Scott Steadman from Newsground said that, according to his sources, Al-Faza'an was the only Saudi employee at OpenAI.

Steadman为这个案子再添一条线索,指出Al-Faza'an也是Raed Ventures的合伙人,在那里他与Khalid bin Bader Al Saud王子共事。

Adding another piece of red string to the corkboard, Steadman noted that Al-Faza'an is also a venture partner at Raed Ventures, where he works alongside Prince Khalid bin Bader Al Saud.

Steadman指出,Khalid王子的父亲是沙特情报部门的成员,但正式身份是现任驻英国大使。

Steadman identified Prince Khalid's father as a member of Saudi intelligence, but officially he's currently the ambassador to the UK.

抛开具体细节不谈,这里的重点是,Al-Faza'an与沙特王室和沙特政府有大量联系,足以构成嫌疑。

Holding aside the specifics, essentially the point here is that Al-Faza'an has ample ties to the House of Saud and the Saudi government to ground suspicion.

就影响而言,OpenAI声称他们的运营是安全的。

As far as implications go, OpenAI claims their operations are secure.

他们写道:“目前,我们没有迹象表明此人不当访问或窃取了模型权重、源代码或客户数据。”

They wrote that "At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data."

尽管如此,外交关系委员会的Chris McGuire认为这对沙特阿拉伯来说是一个非常糟糕的局面,如果沙特政府确实参与,这将对美沙AI关系产生重大影响。

Still, Chris McGuire from the Council on Foreign Relations believes this is an extremely bad look for Saudi Arabia, and if the Saudi government is indeed involved, this should have significant implications for the U.S.-Saudi AI relationship.

McGuire提出了一个长篇分析,如果你感兴趣,确实值得完整阅读,但就重点来说,McGuire认为AI实验室需要加强他们的安全措施,并且要假定他们已经是更不笨拙的间谍活动的目标——这是他的原话。

McGuire presented a lengthy analysis that's actually worth reading in full if you are interested, but to hit the high points, McGuire believes that AI labs will need to harden their security, and presume they are already the target of less ham-handed— his words— espionage attempts.

对于沙特与美国的关系,他认为这是一个非常不好的信号。

For Saudi-U.S. relations, he believes this is a very bad sign.

该调查由司法部的反间谍部门负责,而不是通常处理商业间谍案件的刑事部门。

The investigation is being handled by the DOJ's Counterintelligence Division, not the Criminal Division who would handle regular commercial espionage cases.

如果Al-Faza'an是与沙特政府协调行动的,我看不出在这次事件之后,还有哪家公司能在沙特托管前沿AI模型。

If Al-Faza'an was acting in coordination with the Saudi government, I do not see how any company could possibly host a Frontier AI model in Saudi Arabia after this incident.

遗憾的是,这远远超出了本期节目的范围,但目前美沙关系中发生了很多事情。

Sadly, it is way beyond the scope of this particular show, but there is a lot going on right now in the U.S.-Saudi relationship.

尽管有很多解释不涉及沙特政府,但这一事件可能对全球AI政治产生相当重大的影响。

And while there are plenty of explanations for this that don't implicate the Saudi government, it is an incident that could have some fairly significant implications for global AI politics.

现在转向我们新的智能体时代更为平凡的现实,摩根大通计划阻止第三方智能体访问客户账户。

Moving over to the much more banal reality of our new agentic era, JP Morgan plans to block third-party agents from accessing customer accounts.

在周二的财报电话会议上,智能体风险是摩根大通高管们最关注的问题。

During Tuesday's earnings call, agentic risk was top of mind for JP Morgan executives.

首席执行官Jamie Dimon表示:“我们不会允许任何外部智能体进入公司,因为这对客户和摩根大通来说都太冒险了。”

CEO Jamie Dimon said, "We're not going to allow just any outside agent to come inside the company because it's just too risky for clients and for JP Morgan."

你不想要恶意的智能体,而且可能有成千上万个。

You don't want rogue agents and it could be thousands of them.

尽管Dimon似乎将围绕Hugging Face的令人担忧的新闻与像Muse和Instinct这样的消费级智能体混为一谈,该银行仍在改变其政策。

While Dimon seemed to be conflating the scary headlines around Hugging Face with consumer agents like Muse and Instinct, the bank is still changing its policy.

首席财务官Jeremy Barnum表示,在能够阻止这些行为的范围内,我们正在阻止它们,以保护客户和我们自己。

CFO Jeremy Barnum said, to the extent that we're able to block those, we are blocking them to protect customers and to protect ourselves.

电话会议结束后,一位发言人告诉CNBC,保护我们的客户是我们的首要任务。

After the call, a spokesperson told CNBC, protecting our customers is our top priority.

因为这不安全,我们不允许第三方通过获取客户的银行凭证并代表他们登录来访问账户。

Because it is not safe, we do not permit third parties to access accounts by obtaining customers' banking credentials and logging in on their behalf.

我们支持通过开放银行提供商进行安全、客户授权的数据共享,并致力于安全地实现智能体连接。

We support secure, customer-authorized data sharing through open banking providers, and are committed to enabling agentic connections securely.

这意味着智能体未来很可能能够访问账户信息,执行一些基本的只读任务,比如做预算或者查找要取消的旧订阅;但摩根大通的客户基本不可能让智能体帮他们支付账单或结算发票。

This means that agents will likely at some point be able to access account information for basic read-only tasks like making a budget or finding old subscriptions to cancel, but JPMorgan customers are unlikely to be able to get their agents to pay bills or settle invoices.

戴蒙直言不讳地说,客户并不了解其中的风险。他评论道,如果大家要用这些工具,就应该搞清楚风险。

Dimon said frankly that customers don't understand the risks they're running, commenting, if people are using them, they should understand the risk.

他们需要明白,这里面存在一个责任转移的问题。

They should understand that there's a liability transfer.

本质上来说,银行不会像处理欺诈那样替有故障的智能体兜底赔偿。

Essentially, the bank won't indemnify faulty agents in the same way they do for fraud.

自从Muse把个人智能体带入主流视野以后,智能体金融就成了一个敏感话题。

Now, agentic banking has been a touchy subject since the introduction of Muse brought personal agents to the mainstream.

上个月底,阿波罗首席经济学家Torsten Slok警告说,如果智能体为了帮用户找更高收益的储蓄账户,可能会引发银行挤兑。

Late last month, Apollo Chief Economist Torsten Slok warned that agents seeking a higher yield on savings accounts for their users could trigger a bank run.

如果真信这个场景,那摩根大通就是震中——他们储蓄账户的平均利率才0.01%。

If one puts much credence in that scenario, JPMorgan would be ground zero, with an average savings account rate of 0.01%.

摩根大通似乎并不担心这个特定结果,他们更担心的是个人智能体在没有得到适当授权的情况下乱花钱,带来的潜在责任问题。

JPMorgan doesn't seem concerned about that particular outcome, but rather the potential liability of personal agents spending without proper authorization.

基本上,通过这次政策调整,他们已经说得很明白了:如果Muse开始大肆消费,他们是不会给客户退款的。

Basically, with this policy change, they've made it clear that they won't be refunding customers if Muse goes on a spending spree.

AI for Finance的创始人Bob Su认为,虽然这种情况完全不可避免,但这也绝不是解决真实问题的正确方式。

AI for Finance founder Bob Su thinks that while this was completely inevitable, it's also not the right approach to a real problem.

Bob写道,银行封锁AI智能体,是2026年最可预测的操作。

Bob wrote, banks blocking AI agents is the most predictable move of 2026.

他们关于风险的判断是对的,但解决方法是错的。

They're right about the risk, wrong about the fix.

用户需要的不是把智能体锁在门外。

Users don't need agents locked out.

他们需要的是把智能体牢牢管起来。

They need agents locked down.

作用域权限、硬性额度限制、一个紧急切断开关。

Scoped permissions, hard limits, a kill switch.

与此同时,在光谱的另一端,Shopify则对各种智能体敞开了大门。

Meanwhile, on the other end of the spectrum, Shopify has thrown open their doors to agents of all stripes.

上个月,Shopify和Muse合作,允许智能体在他们的平台上浏览和购物。其中的关键技术细节是,Shopify会直接开放后端的API接口。

Last month, Shopify partnered with Muse to allow agents to browse and shop across their platform, and the key technical detail was that Shopify would provide API access directly to their backend.

这对双方来说都是互利的:Muse能获得更好的结果,而开放API意味着Shopify不用再承担智能体爬取前端带来的网络托管费用。

This was a mutually beneficial arrangement— it allowed Muse to produce better results, while allowing API access meant Shopify wasn't paying web hosting costs associated with agents scraping their frontend.

现在,Shopify已经把这种安排扩展到了所有提供商的智能体。

Now, Shopify has extended this arrangement to agents from any provider.

用户现在可以通过Shopify应用连接自己的智能体,获得完全受支持的智能体购物界面。不管他们用的是Muse、Towne、Hermes、Instinct、Dots,还是其他几十种个人智能体之一。

Users can now connect their agent via the Shopify app for a fully supported agentic shopping interface, regardless of whether they're using Muse, Towne, Hermes, Instinct, Dots, or any of the other dozens of personal agents on offer.

这一举措反映了更广泛的大趋势——在服务智能体这件事上,有两种截然不同的做法正在出现。

The move is part of a broader trend that's seeing two diverging approaches to servicing agents.

像亚马逊、eBay,还有我们刚才说的摩根大通,这些公司把自己的平台封闭起来,逼着客户用他们自己的专有智能体。

Some, like Amazon and eBay, and JPMorgan as we just heard, are locking their platforms down and forcing customers to use their proprietary agents.

这种做法能保持对客户关系和相关数据的掌控,但风险是可能会把使用智能体的客户推到其他平台去。

That approach maintains control of the customer relationship and associated data, but it risks driving agentic customers to other platforms.

另一种选择则像Shopify这样,直接为智能体提供一个无头界面,方便它们进行交互。

The other option is to facilitate agentic interactions directly by offering a headless interface for agents, as Shopify has done.

同样的战略分化也出现在SaaS领域。像Salesforce这样的平台,通过向第三方智能体开放API,已经倾向于无头产品。

The same strategic divergence is playing out in SaaS, where platforms like Salesforce have leaned into headless product by opening their APIs to third-party agents.

而另一边,SAP则加强了对API的控制,用牺牲客户选择权的方式来巩固自己的护城河。

SAP, on the other hand, has tightened controls on APIs, securing their moat at the cost of customer choice.

随着智能体的普及,这种开放平台和封闭平台之间的裂痕成了一个关键战场,到底哪种策略会胜出还远未可知。

This schism between open and closed platforms is one of the key battlegrounds as agent adoption grows, and it's still very unclear which strategy will win out.

最后,今天的新闻是Character AI申请了破产,这款曾经很受欢迎的消费者AI产品走到了尽头。

Lastly today, Character AI has filed for bankruptcy, marking the end of the road for a once-popular consumer AI product.

回到2024年那个古老的时代,Character AI是除了ChatGPT之外的第一个爆款,在A16Z的顶级消费者AI应用榜单上排到第三名。

Back in 2024, the olden days, Character AI was the first breakout hit beyond ChatGPT, ranking at 3rd place in A16Z's list of top consumer AI apps.

它允许用户用聊天机器人创建自己的角色扮演场景。

It allowed customers to create their own character-based roleplay using chatbots.

很多人以为,当谷歌以数十亿美元的人才收购交易挖走该公司大部分顶尖人才时,Character 的日子就屈指可数了,但有一段时间,这家公司依然拥有忠实的用户群。

Many assumed that when Google acquired most of the company's top talent in a multi-billion-dollar acquihire deal, that Character would be living on borrowed time, but for a while, the company maintained a dedicated following.

这次破产背后有两个故事。

The bankruptcy tells 2 stories.

首先,就是在技术和商业决策上草率行事带来的巨大法律风险。

First, there's the incredible legal liability from playing fast and loose with technology and business decisions.

除了因用户自残和聊天机器人模仿医生而引发的诉讼之外,Character AI 还在2025年因为添加了对未经授权角色的支持而惹怒了迪士尼。

Beyond lawsuits over customer self-harm and chatbots mimicking doctors, Character AI ran afoul of Disney in 2025 after adding support for unlicensed characters.

破产申请文件显示,Character AI 有5000万到1亿美元的债务,他们正在寻求通过破产来免除这些债务。

The bankruptcy filing states that Character AI has between $50 and $100 million in liabilities that they're seeking to discharge.

迪士尼作为“假马竞标者”出现,打算从破产中挽救资产——他们同意支付2000万美元购买部分知识产权,最近还聘请了 Character AI 的 CEO 来推动技术转让。

Disney has emerged as the stalking horse bidder to salvage assets from the bankruptcy— they've agreed to pay $20 million for certain intellectual property, and recently hired Character AI's CEO to facilitate the technology transfer.

除此之外,还有一个更大的问题:以消费者为中心的商业模式在2026年是否还能成立。

Beyond that, there's a broader question about whether a consumer-focused business model can function in 2026.

先说明一下,404 Media 对 AI 和整个科技行业都挺敌视的,他们根据破产文件大胆宣称,AI 的经济模式根本不成立。

404 Media, who, it should be known, is pretty wildly antagonistic to AI and tech in general, made the bold claim that the economics of AI don't work based on the bankruptcy filings.

根据 CEO Colin Adams 的宣誓书,Character AI 每月在算力上花费1000万美元,而月度订阅收入峰值只有280万美元。

According to an affidavit from CEO Colin Adams, Character AI was spending $10 million per month on compute, with peak monthly subscription revenue of $2.8 million.

其他月份收入更少,有时甚至跌破每月100万美元。

Other months saw less revenue, even dipping below $1 million a month at times.

虽然 404 把这说成是对整个 AI 商业模式的批判,但我觉得更有意思的问题是——这些问题不一定套用在所有东西上才有意思——消费者愿意为什么买单、愿意付多少,以及这和现有产品到底合不合拍。

While 404 presented this as a condemnation of the AI business model as a whole, I think the more interesting questions— which don't need to apply to absolutely everything to be interesting— are about what consumers are going to be willing to pay for and how much they're going to be willing to pay, and how that does or doesn't line up with products on offer.

发现人们不愿意为某种特定的角色扮演聊天机器人支付足够多的钱,并不意味着消费级 AI 就没有商业模式。

Discovering that people aren't willing to pay enough for a specific implementation of roleplay chatbots doesn't mean that there's no business model in consumer AI.

它只意味着,人们不愿意花足够多的钱,去支撑那一套特定的角色扮演聊天机器人。

It means that people didn't wanna spend enough to justify that particular set of roleplaying chatbots.

Pleiometrics 总结道。

Pleiometrics summed it up.

一家奇怪的公司。

A strange company.

他们沉迷于免费的用户数据指标,审核成人内容的时候拖拖拉拉,然后卡在了一个“到底做不做”的循环里。

They got addicted to free user analytics numbers, dragged their feet on moderation for adult content, and got stuck in this will-they-won't-they loop.

到处都是半吊子。

Half measures all around.

计算机科学家 Jon Kantman 写道:“他们错失良机太多次了。”

They missed the boat one too many times, wrote computer scientist Jon Kantman.

这就是 AI 界的 Pets.com。

This is the Pets.com of AI.

人们真的完全不懂 AI 行业。

People genuinely do not understand the AI industry.

我不敢相信真的有人认为 Character AI 破产意味着下一个就是 Anthropic。

I can't believe people genuinely think that Character AI filing for bankruptcy means Anthropic is next.

不过,随着 AI 的早期阶段让位给别的东西,这仍然是个有趣的时刻。

Still an interesting moment as the early phases of AI cede to something different.

不过,今天要说的新闻头条就这些。

For now, however, that is gonna do it for today's headlines.

接下来,是本期的主节目。

Next up, the main episode.

如果你在企业里负责 AI 业务,你早就知道,现在的差距不是能力,而是执行。

If you're leading AI inside an enterprise, you already know that the gap right now isn't capability but execution.
F1
F113:14

这里有一个残酷的事实。

Here's a harsh truth.

你的公司很可能正在花几千甚至几百万美元买一堆 AI 工具,结果却根本没怎么用起来。

Your company is probably spending thousands or millions of dollars on AI tools that are being massively underutilized.

一半的公司有 AI 工具,但只有12%的公司用它们创造了实际业务价值。

Half of companies have AI tools, but only 12% use them for business value.

大多数员工还在用 AI 整理会议纪要。

Most employees are still using AI to summarize meeting notes.
F1
F114:29

这就是大多数遗留系统现代化项目失败的原因。

Here's why most legacy modernization projects fail.

干活的那个AI没法理解大规模代码库。

The AI doing the work can't understand codebases at scale.

它只能看到很小一段上下文,检查一下语法,却看不到全球应用生态系统里积攒了多年的决策。

It sees a small slice of context, examines syntax, and misses years of decisions distributed across the global application ecosystem.
F1
F115:24

欢迎回到AI Daily Brief。

Welcome back to the AI Daily Brief.

今天,我们连续第二天来看一个大型新模型的发布。

Today, for the second day in a row, we are looking at a big new model release.

有意思的是,和昨天一样,这次发布模型的公司在至少上一轮AI竞争中被认为有点落后。

And what's interesting is that just like yesterday, this is a model release from a company that has for at least the last phase of AI been considered a little bit behind the 8-ball.

我说的当然是Google新发布的Gemini 4 Argon。

I'm talking of course about Google's new Gemini 4 Argon.

虽然大家对Google的看法远没有像今年早些时候对Meta那么糟糕,但也是一路坐过山车:先是远远落后,然后又一下子回到竞争行列,2025年大部分时间和OpenAI、Anthropic基本持平,结果几乎完全错过了2026年的智能体和智能体编码浪潮。

And while the perception of Google has been nowhere near as bad as it was for, for example, Meta earlier this year, it's been a bit of a rollercoaster where Google was far behind only to then launch themselves right back into contention, spending most of 2025 right on par more or less with OpenAI and Anthropic, only to almost completely miss the 2026 wave of agents and agentic coding.

Gemini 4 Argon是几周前,也就是九月底宣布的,但当时还不能用,很多人觉得这不是个好兆头。

Gemini 4 Argon was announced a couple of weeks ago at the end of September, but it wasn't available yet, which many took as not such a good sign.

不过,对于我们这些很愿意给Google一个机会、完全不想就这么把它排除掉的人来说,这其实只是一场等待游戏,等着看我们真正拿到手之后,这个模型到底能做什么。

Still, for the very many of us who were quite willing to give Google the benefit of the doubt, and not at all comfortable with writing them off, it was really always just a waiting game to see what the model could actually do when we could get our hands on it.

在我们深入讨论之前,有必要指出,我认为关于任何新前沿模型的正确问题已经变了。

As we dig in, it's important to note that I believe that the right question about any new frontier model has shifted.

以前的问题可能是“它有多好?”潜台词是“我该不该换过去?”而现在的问题更多是“它适合放在哪里?”

Whereas it might have been before, how good is it, with the implication being, should I switch, the question is now much more, where does it fit?

换句话说,这个特定的模型在哪些方面做得好,或者有什么不同,让我值得考虑把它整合进我整体的模型组合里?

In other words, what things does this particular model do well or different that should make me consider integrating it into my overall model stack?

这就是我们今天要探讨的内容。

So that's what we're going to explore today.

简单回顾一下几周前的信息:这是Google今年早些时候砍掉Gemini 3.5 Pro发布之后,基于全新的预训练打造的全新模型。

To recap the basics from a couple of weeks ago, this is a brand new model built on a brand new pre-train after Google scuttled the Gemini 3.5 Pro release earlier this year.

目前还不清楚这个模型具体有多大,但Google将其定位为与Astra和Fable同级别的超大型模型。

It's unclear exactly how large the model is, but Google is presenting it as an ultra-large model in the same class as Astra and Fable.

回顾一下我们之前看到的基准测试,至少从这些基准来看,Argon看起来基本上和竞争对手实验室的前沿模型不相上下。

To recap the benchmarks that we got a preview of, at least by the benchmarks, Argon appears to be broadly a match for frontier models from the rival labs.

在知识工作方面,Argon在VALS Index、Automation Bench、VALS Finance Agent和Harvey's Legal Agent基准上都达到了最先进水平。

For knowledge work, Argon is state-of-the-art on the VALS Index, Automation Bench, VALS Finance Agent, and Harvey's Legal Agent benchmark.

而且除了核心的VALS Index之外,其他每个基准都遥遥领先于竞争对手。

And other than the core VALS Index, each of these benchmarks have a wide gap above competitors.

比如在Harvey的法律基准上,Argon比最接近的竞争对手Fable 5.1高出13分。

On Harvey's legal benchmark, for example, Argon is 13 points ahead of its closest competitor, Fable 5.1.

在编码方面,结果就比较参差不齐了。

On coding, the results are more mixed.

Argon在DeepSWE上是最先进的,比Opus 5.5高出几分,但在FrontierSWE上,Argon落后于三大主要对手,比领先的Astra低了大约10分。

Argon is state-of-the-art on DeepSWE, scoring a few points above Opus 5.5, but on FrontierSWE, Argon is behind the 3 major rivals and about 10 points off of the leader Astra's score.

同样,在TerminalBench 4.0上,Argon比领先的Opus 5.5低9分。

Similarly, on TerminalBench 4.0, Argon is 9 points behind Opus 5.5, who has the leading score.

不过公平地说,这样呈现可能有点误导。

Although to be fair, that presentation is probably a little misleading.

随着基准测试最近的更新,编码分数的差距拉大了。虽然Argon落后,但它仍然稳居前沿范围,只是不在最前面。

The spread of coding scores has widened with recent updates to the benchmark, and although they are behind, Argon is firmly in the frontier range, just not at the head of the pack.

Google还重点介绍了智能体科学、网络安全和多模态等多个其他基准,根据Google选择的基准,Argon在每个类别上都达到了最先进水平。

Google highlighted a variety of other benchmarks across agentic science, cybersecurity, and multimodal, with Argon state-of-the-art across each category based on the benchmark selected by Google.

还有一个值得强调的基准,考虑到很多问题都会涉及智能体,那就是Argon在OSWorld上的得分,OSWorld是标准的计算机使用基准。

One other benchmark worth highlighting, given that a lot of the questions are going to be about agents, is Argon's score on OSWorld, which is the standard computer use benchmark.

Argon得分69.2%,略低于Astra的72.6%,而Astra当时发布时被认为是AI计算机使用领域的范式转变。

Argon scored 69.2%, which is just shy of Astra's 72.6%, which when announced was considered a paradigm shift in AI computer use.

Artificial Analysis的独立基准测试同样将Argon列于前沿位置。

Independent benchmarking from Artificial Analysis similarly placed Argon squarely at the frontier.

它的总分为53,与Fable-5.1和GPT-6 Astra并列第四。

It scored 53 overall, putting it tied for 4th place with Fable-5.1 and GPT-6 Astra.

Muse、Wonder、Full和Sonnet-5.5以56分并列第二,Opus-5.5以58分领先。

Muse, Wonder, Full, and Sonnet-5.5 are tied for 2nd with 56 points, while Opus-5.5 is in the lead with 58 points.

说实在的,为了不过分纠结具体数字,现在美国四个实验室里有六到八个模型,都算得上是前沿水平的有力竞争者——当然,这取决于你怎么看 GPT-6.1 Sol 和 Grok 4.7。

Realistically, and so we don't get too hung up on the exact numbers, there is now a group of between 6 and 8 models across 4 US labs with a legitimate claim to being competitive at the frontier— depending of course how you feel about GPT-6.1 Sol and Grok 4.7.

但当然,我们不再只活在性能的世界里,我们活在高效性能的世界里——而成本方面,由于谷歌推出这个模型的方式,要仔细拆解就有点难了。

But of course we no longer live in just a world of performance, we live in the world of efficient performance— and cost is a little more difficult to dissect due to the way that Google has rolled this model out.

在 API 上,Argon 在 11 月 30 日之前都有 50% 的降价。

On the API, Argon is available with a 50% cost reduction until November 30th.

这样一来,这个模型跑完 AA 基准测试,每个任务成本 1.99 美元,跟 GLM-5.3 的成本非常接近,而且除了 GPT-6.1 Sol 之外,比所有前沿竞争对手都便宜得多。

That allowed the model to complete the AA benchmark run at $1.99 per task, very similar to the cost of GLM-5.3 and significantly cheaper than all frontier rivals other than GPT-6.1 Sol.

一旦补贴取消,Argon 会比 Astra 更贵,但价格仍然大约是 Anthropic 模型的三分之二。

Once the subsidy goes away, Argon will be more expensive than Astra, but still around 2/3 the price of the Anthropic models.

总结一下,单从数字来看,Argon 看起来就像是个打折的前沿 AI,按理说这应该能让谷歌重新回到牌桌上。

Trying to sum it up, going purely from the numbers, Argon looks like a frontier AI at a discount, which presumably would put Google back in the game.

然而,对很多人来说,Argon 发布后最需要回答的问题是:它能写代码吗?

And yet, for many, the biggest question to be answered by the release of Argon is, can it code?

过去一年,编程一直是谷歌的软肋,而且一直有猜测说,糟糕的编程表现导致 Gemini 3.5 Pro 今年夏天被搁置了。

Coding has been Google's Achilles heel over the past year, and there's been speculation that poor coding performance led Gemini 3.5 Pro to be shelved over the summer.

Pavel Huren 让 Argon 跑了他的找 bug 任务,发现它的表现还算说得过去。

Pavel Huren ran Argon through his bug hunting task and found it to be adequate.

它的表现和 MuseSpark 1.3 差不多,但明显落后于 Astra、Fable 和 Opus。

It performed about as well as MuseSpark 1.3 and significantly behind Astra, Fable, and Opus.

Huren 还是肯定了谷歌的进步,写道:“一个比 Gemini 3.8 Flash 好得多的模型。”

Huren still gave Google credit for making progress, writing, 'a way better model than Gemini 3.8 Flash.'

看起来谷歌回来了。

Looks like Google is back.

来自 Railway 的 Noah Dunnigan 就没那么买账,发帖说:“这个模型在面对真实世界的代码时毫无用处。”

Noah Dunnigan from Railway was less impressed, posting, 'this model is useless when it comes to real-world code.'

我不生气,只是有点失望。

I'm not mad, just disappointed.

不过,这其实对谷歌来说也说得通。

This however actually kinda makes sense for Google.

我觉得他们已经退出了编程模型的竞赛,转而瞄准了知识工作者。

I think they have given up the programming model race and are instead aiming for knowledge workers.

其他人也发现了类似的不太理想的结果。

Others found similarly not-so-great results.

开发者、内容创作者兼创业者 Theo 写道:“我刚刚在代码库分析基准测试上跑了 Argon。”

Developer, content creator, and entrepreneur Theo wrote, 'just ran Argon on my codebase analysis bench.'

看起来不太妙。

Not looking good.

唯一一个虚假声称率超过 50% 的模型。

Only model with a greater than 50% false claim rate.

来自 Every 的 Dan Shipper 写道:“Gemini 4 Argon 在我们的资深工程师基准测试上真的很吃力。”

Dan Shipper from Every writes, 'Gemini 4 Argon really struggles on our senior engineer benchmark test.'

得分只有 21%,远低于 GPT-6.1 Sol,甚至不如 Muse。

Comes in at 21%, far below GPT-6.1 Sol and even Muse.

分析师 Max Weinbach 倒是回应了 Dan,说:“这真的取决于你用的是哪个工具环境。”

Analyst Max Weinbach did respond to Dan saying, 'it really depends on the harness you're using.'

我在 Antigravity 里不太喜欢它,但发现它在 Claude Code 和 OpenCode 里真的很好用。

I didn't love it in Antigravity, but found it was really good in Claude Code and OpenCode.

不过,也有人指出,在其他工具环境里使用这个模型,严格来说是违反服务条款的。

However, as some pointed out, using the model in other harnesses is technically a violation of the terms of service.

总结一下,开发者 Rishi 写道:在数据、代码和分析方面花了几十亿 token 之后,我可以这么说。

Summing it up, developer Rishi writes, a few billion tokens later on data, code, and analysis, here's what I can say.

Gemini 4 大致相当于 Opus 4.8。

Gemini 4 roughly equals Opus 4.8.

现在回到 Antigravity,这次发布的一个特殊之处,可能就是导致体验差异的原因——谷歌决定在编程功能上区分 Pro 和 Ultra 订阅用户。

Now going back to Antigravity, one of the quirks of this release that could be causing the difference in experience is that Google decided to differentiate between Pro and Ultra subscribers on coding.

Pro 订阅用户只能在 Gemini 工具环境里使用 Argon。

Pro subscribers only have access to Argon in the Gemini harness.

你还是能用 Gemini CLI,但它和谷歌的 Antigravity 编程工具不是一回事——后者是他们对标 Codex 或 Claude Code 的产品。

You still have access to the Gemini CLI, but that's not the same as Google's Antigravity coding harness, which is their equivalent of Codex or Claude Code.

要在Antigravity里用Argon,用户得升级到Ultra档位才行。

To use Argon in Antigravity, users need to upgrade to the Ultra tier.

Pavel Huren再次提到,这两个设置之间的性能差距非常大。

Pavel Huren again noted a huge difference in performance between the 2 setups.

他又跑了一次bug hunt测试,写道:用Antigravity命令行界面测试Gemini 4 Argon做实际工作,而不是用Gemini命令行界面,分数发生了巨大变化。

Running his bug hunt test again, he wrote, testing Gemini 4 Argon on real work with Antigravity CLI, not Gemini CLI, changed the score dramatically.

Argon从一个好模型变成了前沿模型,非常接近Opus 5 5 Max。

Argon went from good to a frontier model, landing very close to Opus 5 5 Max.

我跑了三次基准测试来确认结果。

I ran the benchmark 3 times to confirm the results.

之前它对我没效果,我也没预料到测试框架会这么重要。

It hadn't worked for me before, and I didn't expect the harness to matter so much.

Marcelo P. Lima总结了这种挫败感。

Marcelo P. Lima summed up the frustration.

为什么谷歌让人这么难试用他们的新模型?

Why does Google make it so hard to try out their new model?

我有一个Workspace账户。

I have a Workspace account.

我不能花100美元买AI Ultra附加包,因为那只对个人用户开放。

I can't pay $100 to get AI Ultra add-on because that's only available to individuals.

要想获得AI Ultra的访问权限,我还得跟销售谈?

In order to get AI Ultra access, I have to talk to sales?

这太疯狂了。

This is crazy.

让人们能无摩擦地试用你的产品吧。

Make it frictionless for people to try your product.

你们的模型这么聪明。

Your models have all this intelligence.

你肯定能让Argon用vibe coding写一个购买页面,好让我买这个附加包吧。

Certainly you can ask Argon to vibe code a sales page that'll let me buy this add-on.

现在,这个节目的老听众不会惊讶,因为不同的测试框架确实会导致巨大差异。

Now, regular listeners of this show will not be surprised that there is a big difference depending on the harness.

坦率地说,对谷歌来说,如果模型性能只是设置问题,而不是底层模型能力问题,那会是个好消息,因为代理式编程这个用例太重要了。

And frankly, for Google, it would be great news if model performance was a setup issue, not underlying model capability, given how important that agentic coding use case remains.

不过,谷歌模型一直擅长的领域之一是多模态。

Still, one area that Google's models have always shined is on the multimodal.

考虑到现在每个模型都必须在社交媒体这个舞台上竞争,而视频和3D渲染作品吸引了所有注意力,你可能会认为这会是谷歌的强项。

And given how much every model now has to compete in the arena of social media, where videos and 3D renders capture all the attention, you would think this would be an area for strength for Google.

确实,这家公司随模型一起发布了一些令人惊叹的示例。

Certainly, the company released some breathtakingly impressive examples alongside the model.

Demis Hassabis展示了一个现代版战斗棋类游戏作为示例,结果看起来像AAA级游戏画面,很多方面都超过了竞争对手的表现。

Demis Hassabis showcased a modern version of battle chess as an example, with the results looking like AAA game graphics and way beyond in many cases what we've seen from rivals.

Cetus Lua被谷歌的一个演示深深震撼了,这个演示从多个角度展示了一个巨大的宝塔和寺庙建筑群,包括完整的穿越体验。

Cetus Lua was blown away by one of the demos from Google, which depicted a massive pagoda and temple complex from multiple angles, including a full walkthrough.

其他人指出,这并非一次生成,而是一次大规模的多智能体运行,成本会相当高,但输出结果仍然令人印象深刻。

Others noted that this wasn't a one-shot, but a massive multi-agent run that would have been quite expensive, but still the output is impressive.

Ethan Mollick教授能够用他一直在测试不同模型的Imminence游戏版本,同样得到了令人印象深刻的视觉效果。

Professor Ethan Mollick was able to produce a version of his Imminence game that he's been using to test different models with similarly impressive visual results.

但也有人发现结果平淡得多。

But others have found much more lackluster results.

各位,我真的对Gemini 4 Argon感到困惑。

Guys, I'm genuinely confused about Gemini 4 Argon.

刚用我老套的vibe check提示跑了Argon High,我真的不知道该怎么评价它。

Just ran Argon High through my old vibe check prompt, and I really don't know how to evaluate it.

现在,Max的测试是一个3D渲染的车辆,可以从不同角度观察。

Now, Max's test is a 3D render of a vehicle that you can look at from different angles.

Max继续说,它在车形方面做得非常好。

Max continued, it did a really good job with the car shape.

比Opus 5.5的参考模型看起来更像,但颜色完全褪色了。

Looks much more like the reference than Opus 5.5's, but the colors look completely washed out.

那些纹理真的就是飘在空中的。

The textures are literally floating.

车库场景还是太简陋了,界面也还停留在2024年。

The garage scene is still way too basic, and the UI is still stuck in 2024.

它确实很快,但说真的,有这么多缺点,你真的会选它而不是Opus的输出吗?

It was fast, but be real— would you actually choose this over an Opus output with all of these cons?

Max后来甚至还在想,他们是不是拿那些被炒得很热的公开测试来训练这个模型的。

Max actually even later wondered whether they trained the model on hyped public tests.

他说,它离Opus 5.5或Astra还差得远,而且说实话,Sol和Sonnet有时候看起来要好得多。

He said, it's not close to Opus 5.5 or Astra, and honestly Sol and Sonnet sometimes look way better.

它擅长火箭、鹈鹕和体素这种炒作测试,但一旦需要做任何真正有创意的事情,它输出的就是垃圾。

It's good at rockets, pelicans, and voxel hype tests, but the second it needs to do anything actually creative, it serves junk.

那我们该怎么看待这一切呢?

So what are we supposed to make of all this?

当初Astra发布时我说过,它给人的感觉是一个真正在要求你把自己推得更远的模型。

One of the things that I said about Astra when it came out is that it felt like a model that was really asking you to push yourself farther.

不是简单地把同样的事情做得稍微好一点,而是真正去尝试做完全不同的事情。

Not to simply do the same stuff a little bit better, but to really try and do totally different things.

考虑到Argon的一些多模态和3D能力,也许我们也应该用类似的视角来看待这个模型。

And given some of the multimodal and 3D capabilities of Argon, maybe that's a similar way that we should be looking at this model.

当然,那些印象最深的人,往往都是在探索这些动态的、多模态的、先进的或科学的使用场景。

Certainly the people who were most impressed tended to be exploring these types of dynamic, multimodal, advanced, or scientific use cases.

Argon对世界有一种惊人的自然理解力。

Argon has this amazing natural understanding of the world.

它可以从基本原理出发做非常多的事情。

It can do so many things from first principles.

BionX分享了一架完全用Three.js制作的未来感3D战斗直升机,说:你看到的一切都是从头程序化生成的。

BionX shared a futuristic 3D battle helicopter made purely in Three.js, saying, everything you see was procedurally generated from scratch.

这是Argon和Nano Banana多轮自我迭代循环的结果。

This is a result of a multi-round self-iteration loop of Argon and Nano Banana.

风险投资人Justine Moore写道:让Gemini-4 Argon做一个苹果电脑的互动博物馆,而且没有提供任何图片或数据。

Venture investor Justine Moore wrote, asked Gemini-4 Argon for an interactive museum of Apple computers and didn't provide any images or data.

它抓取了Mac的存档照片,写了一个清理图片的工具,拉取了50年的规格参数,还建了一个3D cover flow,界面会随着你所在的年代而变化。

It scraped archival Mac photos, wrote a tool to clean the images, pulled 50 years of specs, and built a 3D cover flow with UI that morphs to match the decade you're in.

换句话说,它把她想要的东西执行得令人叹为观止。

In other words, it is a wildly impressive execution of what she was looking for.

他认为这才是这个模型的真正意义所在,他写道:Gemini 4 Argon是第一个让我感觉图像、视频和语言模型三者合一的模型。

Argued that this is really what the model is about, writing, Gemini 4 Argon is the first model where it felt like an image, video, and language model all at once.

如果是这样的话,一些早期测试感觉只是触及了皮毛,也就不足为奇了。

And if that's the case, it's not surprising that some of the early tests feel like they're just scratching the surface.

他写道:我觉得我们还没有充分看到Gemini 4 Argon真正能做什么。

Wrote, I feel like we haven't seen enough of what Gemini 4 Argon can actually do.

我给了它一张哥特式维多利亚马车的图片,让它把整个东西做成一个交互式3D体验。

I gave it one image of a Gothic Victorian carriage and asked it to build this whole thing as an interactive 3D experience.

就一张图,然后这就是你得到的结果。

Just one image and this is what you get.

一个从零生成的十七万四千面的模型,带有能打开的门、带铰链的窗户、能工作的轮子物理、悬挂系统、闪烁的灯笼,甚至还有定制的骷髅雕刻。

A 174,000-face model from scratch complete with opening doors, hinged windows, working wheel physics, suspension, flickering lanterns, and even custom sculpted skulls.

所有这些都是通过代码生成的,没有用Blender。

All of this generated through code, no Blender.

有一点很清楚:不管它是为什​​么而生的,Gemini 4都不一样。

What's clear is that, whatever it's for, Gemini 4 is different.

Ethan Mollick写道:使用Gemini 4会让你意识到,有多少其他模型一定是在蒸馏Claude或GPT,或者使用了类似的训练流程。

Ethan Mollick writes, Using Gemini 4 makes you realize how many other models must be distilling Claude or GPT or using a similar training pipeline.

它给人的感觉和其他模型非常不同。

It feels quite different than other models.

Theo同意它确实不一样,但也不一定喜欢这一点,他写道:对Gemini来说这是一个巨大的飞跃,而且就这个价格来说性价比很高,但它终究还是一个Gemini模型。

Theo agrees that it's different but doesn't necessarily love that, writing, it's a huge jump for Gemini and pretty good value for the price, but it's still a Gemini model though.

它经常做出奇怪的事情,我觉得我完全没法信任它。

Does weird stuff regularly and I don't feel like I can trust it at all.

我大量测试了它在智能体编程方面的表现,然后写道:我今天一直在用 Gemini 4 Argon 编程,实际上我完全忘了 Gemini 系列模型在编程时那些特定的性格特征。

Tested it extensively with agentic coding and wrote, I've been coding with Gemini 4 Argon today and I actually completely forgot about the character traits Gemini models have when coding.

它们就是喜欢用某种特定的方式来组织代码,并且加注释。

They just love to structure code in a certain way and add comments.

而且注释也不长,大多是简洁的单行注释,这反而是件好事。

Not too long either— mostly neat one-liner comments, which is a good thing, if anything.

我还有点喜欢这样。

I kind of like it.

我甚至还不确定代码能不能运行,这又是经典的 Gemini 风格,但就是看着很舒服,你能理解吗?

I'm not even certain if the code works yet, which is the classic Gemini trait, but it's just nice to look at code, if that makes sense.

跟其他模型一比,它们的习惯差别可大了。

When you compare it to the other models, their habits are wildly different.

跟一个带有自己独特模型血统的模型对话,感觉还挺耳目一新的。

It's kind of refreshing to talk to a model that feels unique to its own model lineage.

再展开说,Google DeepMind 的 AGI 经济学总监 Alex Emas 写道:有一件事一直被低估,那就是在前沿模型上实现真正的输出多样性有多重要。

Expanding out, Google DeepMind's Director of AGI Economics Alex Emas wrote, one thing that's been underemphasized is the importance of having true diversity of output at the frontier.

如果只用单一模型,从写作风格到审美偏好再到代码,每个模型都会有自己明显的烙印。

Using a single model, there are unmistakable signatures of each in everything from writing to taste and code.

这在设计上尤为明显,比如背景选用的色系。

And this is most obvious in design, i.e., the color palette chosen as background.

我已经用 Gemini 4 有一阵子了,可以说它完全独特。

I've been using Gemini 4 for some time now, and I can say it is completely unique.

从大家现在都在分享的可视化结果就能看出来,而且它的写作和互动方式也很独特。

This is obvious from the visualizations everyone is sharing now, but also in the way that it writes and communicates back and forth.

对一个经济学家来说,多样性和独立性的好处几乎渗透到了我们所有的模型里。

For an economist, the benefits of diversity and independence are baked into nearly all of our models.

相关信号、信息和先验知识正是导致脆弱性、学习陷阱和 AI 垃圾的根源。

Correlated signals, information, and priors are how you get fragility, learning traps, and AI slop.

所以我很高兴看到我们的“大宝贝”终于能让大家使用了。

So I'm very glad to see our big boy out there for people to use.

把所有这些总结起来,来自 Every 的 Dan Shipper 甚至发现,至少在某些领域,切换到 Gemini 作为核心工具并不需要什么天马行空的想象。

And bringing it all together, Dan Shipper from Every even found at least one place where it doesn't take some wild imagination to consider switching to Gemini for a core use case.

他写道:Gemini 4 发布了,而且出乎意料地好?

He wrote, Gemini 4 is out and it's surprisingly good?

我承认,看到基准测试的时候我本来准备把它批得一文不值,但我们 Every 的六个人上周在编程、写作、设计和知识工作等方面测了一圈,结果还挺惊喜的。

I admit I was ready to trash this model when I saw the benchmarks, but 6 of us at Every spent the past week testing it across coding, writing, design, and knowledge work, and were pleasantly surprised.

如果你的公司完全依赖 Google 生态,那这算是一个突破;对于喜欢尝鲜前沿模型的人来说,它也是一个不错的备用选择,用来处理某些特定任务。

It's a breakthrough if your company lives in Google's ecosystem, and a strong option for frontier hoppers as a secondary model for certain tasks.

Dan 发现的一些事情包括,用他的话来说,Gemini 4 在重构方面不太稳,但在全新项目上很快——基本上就是它能又快又好地搭建新东西,但用他的话讲,它在现有代码库的自我验证上偷工减料。

Some of the things that Dan found were that Gemini 4 was, in his words, shaky at refactors but fast on greenfield apps— basically that it could build new stuff pretty quickly and efficiently and well, but that it, in his words, cut corners on self-verification in existing codebases.

他还指出了运行环境的问题,说它是一个“2026 年的模型被困在了 2025 年的壳子里”。

He also noted the problems with the harness, calling it a 2026 model trapped in a 2025 harness.

而且我觉得值得指出的是,Google 不仅错过了八个月的编程模型开发,也基本上错过了八个月运行环境用户体验的迭代。

And I think it's worth noting that in addition to missing 8 months of coding model development, Google has also more or less missed 8 months of Harness UX iteration.

Dan 直言不讳地说,Google 的桌面代理应用 Anti-Gravity 感觉比 Claude Code 和 Codex 落后了半年到一年,缺少云环境、SSH、内置浏览器和智能权限检查,结果就是你要么不停地点击弹窗,要么使用不受限制的 Turbo 模式。

Dan put it bluntly saying, Anti-Gravity, Google's desktop agent app, feels 6 months to a year behind Claude Code and Codex, lacking cloud environments, SSH, a built-in browser, and smart permission checks, leaving you clicking constant pop-ups or using unrestricted turbo mode.

但到了写作方面,Dan 说它是他新的最爱写作伙伴。

But then it comes to writing, where Dan called it his new favorite writing partner.

他写道:DeepMind 雇了专业写手拿着红笔去消除 AI 垃圾,而且确实管用。

He wrote, DeepMind hired professional writers with red pens to stamp out AI slop, and it worked.

我用它来理清结构,打磨比喻。

I use it to untangle structure and sharpen metaphors.

它对结构和详细解释有一种天生的感觉,这是 Astra 和 Opus 所欠缺的。

It's got a natural feel for structure and detailed explanations that Astra and Opus lack.

最终结果是,他说:我日常开着 ChatGPT 和 Claude,但当需要一位编辑伙伴来组织一篇难搞的文章、打磨一个比喻或者精读一本书的时候,我就切换到 Gemini 4。

The net result, he said, I keep ChatGPT and Claude open day-to-day and switch to Gemini 4 when I need an editorial partner to structure a tough essay, sharpen a metaphor, or close-read a book.

用它来进行实时写作协作、快速低成本的原型设计,或者可操控的 UI 和仪表盘设计。

Reach for it for live writing collaboration, fast, low-effort prototypes, or steerable UI and dashboard design.

理想情况下是通过 Cursor 或 API 来使用,这样就能跳过 AntiGravity。

Ideally via cursor or the API to skip antigravity.

当一个模型好到逼你用一个烂的工具时,这本身就能说明问题。

When a model is good enough to force you to use a bad harness, it tells you something.

考虑到其他所有模型在写作上还是一如既往地差,如果Dan的结果真的能站稳脚跟,光这一点就足够让我花点时间研究这个模型了。

Given how bad all the other models continue to be on writing, if Dan's results really stick, that alone would be enough for me to spend some time with this model.

所以,回到Argon到底适合用在哪儿这个问题上,答案似乎是——目前为止,我们还不完全确定。

So, when we come back to the question of where Argon fits, the answer seems to be so far, we're not quite sure yet.

那个写作用例本身已经值得测试一下了,但确实感觉,就像X上那位网友说的那样,我们可能才刚触及它的皮毛。

That writing use case is enough to give it a test, but it does feel like, as that one X poster said, that we may be just scratching the surface of this.

同样地,我们或许应该督促自己去想一些更难、更奇特、更意想不到的事来用它做做看,这样才能真正挖掘它的全部潜力。

And once again, we might want to push ourselves to think about harder, stranger, more unexpected things that we can do with it to really see its full potential.

或者反过来,等别人去探索那些事,这样我就能在节目里聊它们,而你就只管做那些真正管用的就好。

Or alternatively, wait till others do that so I can talk about them on this show and you can just do the stuff that actually works.

不管怎样,我很兴奋,我们在这个十月中旬的周三坐在这里,而Google和Meta都重新回到了模型竞赛的赛道上。

Either way, I'm excited that we are sitting here on Wednesday in the middle of October with both Google and Meta meaningfully back in the model race.

不过现在,今天的AI Daily Brief就到这里。

But for now, that's going to do it for today's AI Daily Brief.

一如既往,感谢你的收听或收看。

Appreciate you listening or watching as always.

下次再见。

And until next time.

下次见,平安!

Till next time, peace!

平安!

Peace!
已剔除 4 处广告(点击展开查看)
F1
F10:35广告 · 已剔除

First of all, thank you to today's sponsors: KPMG, Section, Robots and Pencils, and Blitzy.

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

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

And to find out about everything happening in the broader AIDB ecosystem, go check out the website.

You can find access to our past episodes, each of which is divided into a million small sharing nuggets to make it easy for you to share with friends and colleagues.

You can also get access to our free training programs, or if you want to go deeper, you can check out the paid training programs that are offered via Super Intelligent, like the new AI Champions program that we just announced.

F1
F112:41广告 · 已剔除

That's why KPMG's You Can with AI is back with a new season featuring conversations with leaders like Surojit Chatterjee of EMA, Mei Habib of Ryder, McKesson CIO Ellery Fisher, and others focused on practical execution.

What's working, what's not, and what it actually takes to move from pilots to real scaled impact across strategy, data readiness, governance, workforce, and value.

And of course, it's co-hosted by me, Nathaniel Whittemore.

Go listen and subscribe at www.kpmg.us/aipodcasts.

That's www.kpmg.us/aipodcasts.

F1
F113:29广告 · 已剔除

If you're the one responsible for AI adoption at your company, you need Section.

Section is a platform that helps you manage AI transformation across your entire organization.

It coaches employees on real use cases, tracks who's using AI for business impact, and shows you exactly where AI is and isn't creating value.

The result?

You go from rolling out tools to driving measurable AI value.

Your employees move from meeting summaries to solving actual business problems, and you can prove the ROI.

Stop guessing if your AI investment is working.

Check out Section at sectionai.com.

That's S-E-C-T-I-O-N-A-I dot com.

The best teams don't have a single star carrying everyone else.

They know their own strengths and each other's weaknesses and play to both.

That's the team Robots and Pencils has built on purpose.

Nobody there is grinding through busywork to pad a headcount number.

People come for the hard problems, and they stay because everyone around them is leveling up at the same time.

In a market full of companies that are just trying to hire fast, that's worth a look.

Check out robotsandpencils.com/careers.

F1
F114:42广告 · 已剔除

Blitzy solves this the way it solves everything.

Grounded in your code before any migration begins, Blitzy's agents reverse engineer the entire legacy system into a persistent knowledge graph.

Every dependency, every constraint, every piece of tribal knowledge that used to live in one engineer's head.

From that understanding, Blitzy autonomously executes language migrations, framework upgrades, and monolith-to-microservices transformations, all validated end-to-end.

One Blitzy customer modernized a $10 million monolithic insurance stack in 16 weeks against a 137-week baseline with coding agents.

That's 9x compression.

Retire technical debt while accelerating your roadmap.

See how at blitzy.com.

That's B-L-I-T-Z-Y dot com.