M10:54
我们今天先从最新一家强调速度的公司开始。
We kick off today with the latest company to put some emphasis on speed.
过去几个月我们越来越频繁地看到这一点。
This is something we've seen increasingly throughout the last couple of months.
显然，谷歌在它们的Flash模型上强调过这一点，但最近OpenAI也在推他们的Ultra Fast模式，而Anthropic现在通过Fast模式的新功能，已经能以极快的速度提供Opus 5.5的服务了。
Obviously, it's something that Google has emphasized with their Flash models, but more recently, OpenAI has also been pushing their Ultra Fast mode, and Anthropic has now figured out how to serve Opus 5.5 at a blistering pace with a new addition to Fast mode.
他们似乎在这个版本上加快了速度，现在Opus最高能以经认证的每秒285个token的速度提供。
They seem to have stepped up the pace for this one, now serving Opus at a verified 285 tokens per second at max.
OpenAI没有公布GPT-6.1 Sol在Fast模式下的速度测试，但他们的文档表明理论最高每秒280个token。
OpenAI hasn't published a speed test for GPT-6.1 Sol on their Fast mode, but their docs suggest a theoretical maximum of 280 tokens per second.
而且，OpenAI对高级速度收费高出50%，虽然人们在争论哪个更好，但实际上这两者都是非常昂贵的产品版本，应用场景相当小众。
What's more, OpenAI is charging 50% more for the premium speed, and while people debate which is better, realistically, both are very expensive versions of the product with fairly niche applications.
对于Anthropic的版本，正常订阅不包含使用量，有些人这周末就吃了亏才发现的。
For Anthropic's version, usage isn't included in regular subscriptions, as some found out the hard way this weekend.
不过，对于那些想省着钱快速消耗大量token的人来说，Chubby指出Anthropic每美元能多提供50%的token。
Still, for those looking to burn through a ton of tokens quickly on a budget, Chubby noted that Anthropic can serve 50% more tokens per dollar.
我觉得目前这些快速和超快速模式对大多数人来说还有点新鲜，但随着它们变得更普遍和必不可少，模型延迟的巨大变化将改变我们与它们互动的方式，而这些方式在真正实现之前我们很难完全预知。
I think that right now, these fast and ultra-fast modes are a little bit of a novelty to most people, but as they become more common and de rigueur, big shifts in the latency of models will change how we interact with them in ways that I don't think that we'll quite know until we get there.
这最终作为AI竞争的一个方面有多重要，我们拭目以待，但眼下所有实验室肯定都在探索这个方向。
How important this ultimately is as a vector of AI competition, we will see, but it's certainly something that all the labs are exploring at the moment.
说到Anthropic，有人在测试中发现了新的Anthropic模型。
Now speaking of Anthropic, new Anthropic models have been spotted in testing.
用户报告说查询被路由到一个新模型而不是Opus 5.5，结果非常惊艳。
Users have reported queries being routed to a new model instead of Opus 5.5 with very impressive results.
人们在展示基于代码的动画，这些本来是Opus 5.5引入的，但后来似乎被推到了这个新模型上。
People were showing off code-based animations that were introduced with Opus 5.5, but then seemingly pushed to this new model.
还有人称看到Fable的提示也被路由到了一个新模型，这开启了一种双模型同时发布的可能性。
Others reported seeing Fable prompts routed to a new model as well, opening the possibility of a double-headed release.
Chubby在X上写道：‘发布应该不远了。’
Chubby on X wrote, A release should be near.
因为Anthropic的IPO也很快了，他们希望新模型能提前几周发布，这样市场就能把它计入估值。
Since Anthropic's IPO is also pretty close, they want their new model to be released a few weeks ahead so the market can really count that into the valuation.
长话短说，Fable 5.5应该很快就会到来，而且看起来比我们预想的进步还要大得多。
Long story short, Fable 5.5 should arrive soon, and it looks to be a much bigger leap than we even expected.
除了竞争态势，这次发布也可能是一次重大考验，看各大实验室今后愿意发布什么样的内容。
In addition to the competitive dynamics, this release could also be a big test of what the major labs are comfortable releasing moving forward.
OpenAI和Anthropic都限制了对它们最先进内部模型的测试，并可能致力于通过专注于Opus和SOL这类中型模型的发布放缓前沿进展。
Both OpenAI and Anthropic have restricted testing of their most advanced internal models and could commit to pacing the frontier by focusing on mid-sized model releases like Opus and SOL.
或者我们可能会看到新一轮加速，随着更大规模的前沿模型被发布出来。
Or we could see another round of acceleration as new larger frontier models are unleashed.
至于AI业内人士对此的预期，谷歌的Patrick Tulme写道：‘我昨晚在纽约市参加了一个AI晚宴。’
As to what AI insiders expect from this, Google's Patrick Tulme wrote, I was at an AI dinner last night in New York City.
讨论了一些有趣的话题。
Some interesting topics discussed.
其中一个话题是，用他的话说，共识是Fable 5.5会惊艳，但OpenAI将在12月用Belle击败它。
One of those topics was, as he put it, consensus was Fable 5.5 will be stunning, but OpenAI will beat it in December with Belle.
也就是说，如果你以为假期可以放松一下，那再想想吧，朋友们。
If you thought, in other words, about slowing down for the holidays, think again, my friends.
说到OpenAI，据报道，该公司正在探索通过合作来复兴Sora。
Speaking of OpenAI, the company is reportedly exploring a Sora revival as part of a partnership.
很久以前，这款TikTok风格的视频生成应用曾短暂风光过一阵。
Oh so long ago, the TikTok-style video generation app enjoyed a brief moment in the sun.
2025年底，Sora在App Store排行榜上跃升至第一位，取代了ChatGPT。
In late 2025, Sora grabbed the number 1 spot on the App Store chart, displacing ChatGPT.
大约五分钟内，Sora梗就成了热门趋势。
For about 5 minutes, Sora memes became the hot trend.
然而，到今年年初，在面对如何分配算力的艰难选择时，像编程这类企业用例胜出，Sora被搁置了。
However, by the beginning of this year, when faced with difficult choices around where to deploy compute, enterprise use cases like coding won out, and Sora was mothballed.
四月份关停Sora，就是OpenAI所谓“砍掉支线任务”方针的最佳例证。
The shutdown in April was OpenAI's best example of its mandate to end what it called side quests.
不过现在，复出可能正在酝酿中。
Now, however, a comeback could be in the works.
据The Information报道，OpenAI正在与一家未具名的初创公司洽谈，计划授权Sora模型并围绕它重新推出服务。
The Information reports that OpenAI is in talks with an unnamed startup to license the Sora model and relaunch services around it.
据报道，这家隐身初创公司由前OpenAI研究员、曾领导Sora项目的Bill Peebles在近几个月创立，联合创始人包括前迪士尼董事长、好莱坞传奇人物Jeffrey Katzenberg，以及前Dropbox首席财务官Sujay Jaswa。
The stealth startup was reportedly founded in recent months by Bill Peebles, the former OpenAI researcher who led the Sora project, and co-founders include former Disney chairman and Hollywood legend Jeffrey Katzenberg as well as former Dropbox CFO Sujay Jaswa.
消息人士称，这笔交易仍在推进中，尚不清楚这家初创公司能否获得Sora的完整权重，但他们将有足够的权限对该模型进行重度定制。
Sources said the deal is still coming together and it's unclear whether the startup would get full access to Sora's weights, but they would have enough access to heavily customize the model.
所以，那些怀念猫咪疯狂梗图的朋友们，你们可能很快又能玩起来了。
So for those who miss those crazy cat memes, you might be back in business soon.
在收购新闻方面，亚马逊正在洽谈收购Decart AI，这可能是他们在AI时代最大的一笔收购。
In acquisition news, Amazon is in talks to acquire Decart AI in what could be their largest acquisition of the AI era.
据《华尔街日报》报道，这笔交易正以70亿美元的估值进行谈判，可能很快达成。
The Wall Street Journal reports that the deal is being negotiated at a $7 billion valuation and could come together soon.
Decart对亚马逊来说是一个有趣的目标，也能在很大程度上说明他们如何看待自己在AI竞赛中的地位。
Decart is an interesting target for Amazon and could say a lot about how they view their position in the AI race.
这家初创公司成立于2023年，一直专注于开发用于实时视频的世界模型。
The startup was founded in 2023 and has focused on developing world models for use in live video.
他们的模型可以生成实时生成式视频，用于诸如直播中角色实时变身或生成式游戏等场景。
Their models can produce live generative video for use cases like live character transformations for streams or generative gaming.
尽管技术很酷，但Decart并没有获得太多市场关注。
While the tech is cool, Decart hasn't found all that much traction.
与此同时，亚马逊自去年12月以来就没有发布过新的模型系列，并在上一轮裁员中解雇了大部分AI研究人员。
Amazon, meanwhile, hasn't released a new series of models since last December, and let go most of their AI researchers in the last round of layoffs.
在大多数人看来，他们没有兴趣在前沿或通用大语言模型的模型层竞争。
It appeared to most that they had no interest in competing on the model layer for either frontier or commodity LLMs.
那么，这步棋看来是冲着硬件优化去的。
The play then seems to be about hardware optimization.
Decart去年获得了亚马逊Trainium 3芯片的早期使用权，并开发了一个名为Decart优化系统（简称DoS）的技术栈。
Decart was given early access to Amazon's Trainium 3 chips last year and developed a technology stack they called the Decart Optimization System, or DoS.
他们声称，这项技术能让推理速度提高多达8倍，而成本仅为原来的八分之一。
They claim this technology allows inference to operate up to 8 times as fast and at 1/8 the cost.
DoS与硬件无关，因此可以帮助针对Nvidia GPU或Google TPU优化的AI模型高效部署在亚马逊的硬件上。
DoS is hardware agnostic, so could help AI models optimized for Nvidia GPUs or Google's TPUs deploy efficiently on Amazon's hardware.
到目前为止，DoS似乎只用于Decart的世界模型，因此尚不清楚这项技术能否普遍应用于大语言模型，但亚马逊似乎愿意赌一把。
So far, it seems like DoS has only been used for Decart world models, so it's unclear whether the technology can apply generally to LLMs, but Amazon appears to be willing to take the gamble.
上个月，据说亚马逊曾考虑以60亿美元收购，但在进行尽职调查后放弃了。
Last month, Amazon was said to be exploring an acquisition at $6 billion but walked away after performing due diligence.
Decart去年5月完成上一轮融资，估值达到40亿美元，这意味着亚马逊将支付较上一轮估值75%的溢价。
Decart last raised funds in May, reaching a valuation of $4 billion, pointing to Amazon paying a 75% premium on their last mark.
如果这笔交易达成，目的显然是推理优化。
If the deal comes together, it's clearly about inference optimization.
如果这项技术能大规模落地，将让AWS在竞争中取得巨大优势。
If the technology works at scale, it would give AWS a huge edge over their rivals.
鉴于我们看到行业关注点正从单纯的模型和产品转向如何尽可能廉价高效地提供token服务，亚马逊似乎正在为一个仅以硬件效率为唯一重要基准的世界做准备。
Given how much we're seeing industry attention shift from models and products alone to the business of serving tokens as cheaply and efficiently as possible, Amazon potentially appears to be preparing for a world where hardware efficiency is the only benchmark that matters.
另一个有意思的收购案正在酝酿中——据报道，英伟达正在洽谈收购美国开源模型开发商 Reflection AI。
One other interesting acquisition in the works — Nvidia is reportedly in talks to acquire US-based open-model developer Reflection AI.
你可能还记得，经过漫长的等待后，Reflection AI 上周终于发布了他们的第一款模型，叫 Beam。
You might remember that after a long wait, Reflection AI unveiled their first model called Beam last week.
这个模型的目标是跟中国开发者最好的开源权重模型竞争，虽然基准测试还没完全达标，但 Beam 明显优于其他美国开源模型，比如英伟达的 Nemotron。
The model was positioned to rival the best open-weight models from Chinese developers, and while it wasn't quite there on the benchmarks, Beam was still clearly ahead of other US open models like Nvidia's Nemotron.
《金融时报》报道称，在 Reflection AI 发布模型后，英伟达希望加大投资力度。
The Financial Times now reports that Nvidia wants to double down on Reflection AI following their model reveal.
英伟达已经是 Reflection AI 的最大投资者之一，持有 8 亿美元的股份，并且据报道正在考虑多种方案，包括追加现金或算力投资，以及全面收购加人才吸纳。
Nvidia is already one of the largest investors in Reflection with an $800 million stake and is reportedly exploring options including further investment in cash or compute as well as a full acqui-hire.
Reflection AI 在去年三月的估值是 250 亿美元，所以这桩交易可能会超过英伟达去年十二月收购 Groq 时支付的 200 亿美元，成为他们史上最大的一笔收购。
Reflection was last valued at $25 billion in March, so this deal could exceed the $20 billion Nvidia paid for Groq in December to become their largest acquisition ever.
《金融时报》称，这笔交易可能在未来几周内达成，但也有可能黄掉。
The FT said a deal could be reached in the coming weeks but could still fall apart.
现在看来非常明确，英伟达想在西方打造最好的开源模型栈。
It is very clear at this point that Nvidia wants to build the best open model stack in the West.
英伟达买下了 Hugging Face，收购了 Poolside 和 Essential AI，现在又盯上了 Reflection AI。
Nvidia bought Hugging Face, acquired Poolside and Essential AI, and is now eyeing Reflection AI.
Jensen 正在把这些新实验室整合起来，训练美国最好的开源模型。
Jensen is assembling these new labs to train the US's best open-source models.
至于竞争的问题，很明显英伟达现在要和前沿实验室正面交锋了。
As to the question of competition, it's pretty clear that Nvidia will compete head-on with the frontier labs now.
Jensen 很不爽，因为那些实验室都在开发自己的 ASIC，抢英伟达的生意。
Jensen is pissed because labs are developing their own ASICs and undercutting Nvidia.
我不觉得英伟达收购的目的是为了跟 OpenAI 或 Anthropic 竞争。
I don't think Nvidia's intent with this acquisition is to compete with OpenAI or Anthropic.
如果是那样，英伟达干嘛还要参与 OpenAI 现在的融资轮？
If it were, why would Nvidia invest in OpenAI's current funding round?
更可能是为了改善 Nemotron 那糟糕的表现。
It's probably about improving Nemotron's terrible performance.
他接着写道，假设这次收购是为了让英伟达自己变成一个人工智能实验室。
Continuing, he wrote, let's say this acquisition is intended to turn Nvidia into an AI lab itself.
那只会促使 OpenAI 和 Anthropic 加速脱离英伟达，这反而更不符合英伟达的利益。
That would only push OpenAI and Anthropic to accelerate their shift away from Nvidia, which would be even less in Nvidia's interest.
不管背后的逻辑是什么，很明显英伟达非常看好开源这一块市场。
Whatever the logic, it is very clear that Nvidia is very interested in the open-source side of the market.
而且就在这个话题热度正高的时候，微软恰到好处地发布了他们下一代自研芯片，并计划大规模部署。
And with perfect timing to add intrigue to exactly this conversation, Microsoft has unveiled the next generation of their homegrown chips with a plan to deploy them at scale.
CEO 萨提亚·纳德拉周一公布了新的 MAIA 300 芯片，称这是微软在定制芯片上最大的一次押注。
CEO Satya Nadella announced the new MAIA 300 chip on Monday, calling it Microsoft's biggest bet on custom silicon to date.
他声称，新芯片每瓦特能处理的 token 数量比领先的 GPU 硬件多 40%。
He claimed the new chips can deliver 40% more tokens per watt than leading GPU-based hardware.
他写道，从明年开始，我们会把 MAIA 系列扩展到多吉瓦级别的部署，让它成为我们自有产品和服务的主要推理系统。
Starting next year, he wrote, we will scale our MAIA family to multi-gigawatt deployments, making it the primary inference system for our own products and services.
MAIA 200 芯片已经在小规模部署了，但这一代预计会在性能上有大幅跃升。
The MAIA 200 chips are already deployed at a smaller scale, but this generation is expected to be a big leap in performance.
投资者 Chamath Palihapitiya 写道，你应该预料到芯片公司会向上层发展，而应用公司会向下层发展。
Wrote investor Chamath Palihapitiya, you should expect the silicon chip companies to move up the stack as the application companies move down the stack.
在人工智能领域，越来越明显的是，每个人都在跟每个人竞争。
In AI, it is increasingly true that everyone competes with everyone.
很快，每个人都会提供自己的大模型、芯片、云服务和应用程序。
Everyone will soon offer their own LLMs, silicon, cloud, and applications.
问题是，你能围绕什么制约因素建立起长期的市场和利润空间？
The question is, what will the constraint be that you can build a long-term market and margins around?
是分销渠道吗？
Is it distribution?
是数据护城河吗？
Is it a data moat?
还是像电力这样的物理制约因素？
Is it a physical constraint like power?
未来几个月肯定还会有更多这样的讨论，不过今天就先到这里，以上是今天的头条新闻。
Expect a lot more discussion along those lines in the months to come, but for now, that's gonna do it for today's headlines.
接下来，是本期的主节目。
Next up, the main episode.
大家好。
Hello, everyone.
AI带来的一个重大变化是，我们的思考方式从“如何给页面排名”变成了“如何成为AI足够信任、愿意用来回答问题的信源”。
One big change around AI is we've shifted our thinking from how we rank our pages to how do we become the source that AI trusts enough to answer with.
在KPMG，他们亲身感受到了这一点。
At KPMG, they're seeing this firsthand.
AI生成的结果现在会直接展示答案，很多时候连一次点击都不需要。
AI-generated results now surface answers directly, often without a single click.
所以他们在越来越关注生成式引擎优化，也就是GEO，通过对内容进行结构化处理，让AI系统能够检索、理解，并将其作为可信来源引用。
That's why they are increasingly focused on generative engine optimization, or GEO, structuring content so AI systems can retrieve it, understand it, and cite it as trusted authority.
这不仅仅是SEO的演变，更是一种可见性的必然要求。
This is not just an SEO evolution, but a visibility mandate.
事实上，KPMG提出的GEO要求很简单。
And indeed, the GEO mandate from KPMG is simple.
如果AI在影响决策，那么你的专业见解就必须出现在答案里。
If AI is shaping decisions, your expertise needs to show up inside the answer.
更多详情请访问 kpmg.com/us/geo。
Read all about it at kpmg.com/us/geo.
再说一遍，网址是 kpmg.com/us/geo。
Again, that is kpmg.com/us/geo.
本期节目由CData赞助播出。
This episode is brought to you by CData.
正在部署AI的公司，接二连三碰到同一堵墙。
Companies rolling out AI keep hitting the same wall.
Token的消耗在攀升，安全问题拖住了AI代理的后腿，而且没人知道AI到底是怎么被使用的。
Token spend is climbing, security is holding agents back, and no one knows how AI is truly being used.
根本原因在于：AI与你业务所依赖的系统和上下文是隔绝的。
The root cause: AI is cut off from the systems and context your business runs on.
所以CData构建了Connect AI Gateway，在AI和你的系统之间设立一个统一控制点。
That's why CData built Connect AI Gateway, one control point between your AI and your systems.
它基于支撑Palantir、Google、Salesforce等数百家公司的连接层构建，每个AI代理都能精准获得所需的数据和工具。
Built on the connectivity layer powering Palantir, Google, Salesforce, and hundreds more, each agent gets exactly the data and tools it needs.
它可以精细到单条记录级别的管控，让你终于能在同一个地方看到整个组织的完整AI使用情况。
It governs down to the record and lets you finally see your org's complete AI usage in one place.
准确的答案、极低的token成本，以及一个每次交互都在学习你业务的网关。
Accurate answers, a fraction of the token cost, and a gateway that learns your business with every interaction.
在cdata.ai就能了解详情。
See how at cdata.ai.
当我开会的时候，我会完全投入其中。
When I'm in a meeting, I'm fully in it.
我思考的是，要真正推动目标前进，需要哪些迭代和创造性的反复交流。
I'm thinking about the iteration and creative back and forth it takes to actually push a goal forward.
我不会去想的是记录要点、追踪待办事项之类的事情。
What I'm not thinking about is capturing takeaways, tracking to-dos, or any of that.
这时候Granola就派上用场了。
And that's where Granola comes in.
Granola是一个AI驱动的笔记本，它能记录会议中发生的内容，并转换成干净、结构化的笔记，把决策和行动项单独提取出来，方便查找。
Granola is an AI-powered notepad that captures what happens in your meetings and turns it into clean, structured notes, with the decisions and action items pulled out and easy to find.
不需要安装，也不需要配置。
There's no setup and no configuration.
它正好融入你现有的工作方式。
It just fits into how you already work.
对我来说，这意味着我可以一直处于创意模式，而Granola能确保这些想法真正变成行动。
For me, it means I get to stay in idea mode and Granola makes sure those ideas actually become action.
一旦你在第一次会议上试用过 Granola，就很难再离开它了。
Once you try Granola on a first meeting, it is hard to go without.
你可以去 granola.ai/brief 完全免费试用。
You can try it totally free at granola.ai/brief.
就是 granola.ai/brief，帮你把时间拿回来。
That's granola.ai/brief to get your time back.
Blitzy 在写代码之前会深入理解你的整个代码库。
Blitzy deeply understands your codebase before it writes code.
第一个体现价值的地方就是：AI 时代的安全。
Here's the first place that pays off: security in the age of AI.
漏洞不是孤立存在的，它们深埋在数百万行相互关联的代码里，补一个东西往往会悄悄弄坏另外三个。
Vulnerabilities don't live in isolation— they live buried inside millions of lines of interconnected code, where patching one thing quietly breaks 3 others.
这就是为什么表面扫描根本不管用。
That's why surface-level scans fail.
Blitzy 从整个应用的知识图谱出发，识别并暴露全系统范围内的 CVE，主动推荐补丁，还能直接执行 PR。
Blitzy starts from its knowledge graph of your entire application, identifies and surfaces CVEs across the full estate, proactively recommends patches, and can execute the PR.
每一个修复都基于系统之间的连接和验证方式，确保不会引入新的问题，知识图谱还会动态更新，让你始终领先于不断加速的威胁形势。
Each fix is grounded in how your systems connect and validate so nothing new breaks, and the knowledge graph dynamically updates, keeping you ahead of an ever-accelerating threat landscape.
有一个 Blitzy 客户在 4 天内解决了横跨 6 个核心微服务的 21 个活跃 CVE。
One Blitzy customer resolved 21 active CVEs across 6 core microservices in 4 days.
零编译错误，所有验证扫描全部通过，原本计划几个月的工作不到一周就完成了。
Zero compile errors, every validation scan clean, months of planned work fixed in less than a week.
以真实架构上下文为基础，以计算速度进行安全修复。
Security remediation grounded in real architectural context at the speed of compute.
去 blitzy.com 加固你的代码库。
Harden your codebase at blitzy.com.
就是 blitzy.com。
That's blitzy.com.
欢迎回到 AI Daily Brief！
Welcome back to the AI Daily Brief!
就在不久之前，肯定不到一年前，大多数关注这件事的人多多少少都觉得 Meta 已经退出 AI 竞赛了。
It wasn't very long ago at all, less than a year certainly, that most people paying any sort of attention were more or less writing Meta out of the AI race.
实际上在 2025 年，关于 Meta 和 AI 的大多数头条都是扎克伯格花巨额资金吸引研究人员加入 Meta，然后当然还有这些举动之后引发的各种内部纷争。
In fact, in 2025, most of the headlines around Meta and AI were about Mark Zuckerberg spending an extraordinary amount of money to lure researchers to Meta, and then of course all of the internal strife that would follow those moves.
然而到了 2026 年，agent 之年，2026 年秋天当然是个人 agent 的季节，Meta 完全回到了前沿实验室的行列。
And yet, in 2026, the year of the agent, with fall of 2026 of course being the season of the personal agent, Meta is fully back in the mix as a frontier lab.
我们从 Muse Spark 模型的发布开始看到这一点，这些模型即使不是最先进的，也绝对不丢人，而且足以支撑他们消费者应用 Muse 的许多功能。
We started to see this with the release of the Muse Spark models, which were, if not state-of-the-art, certainly not embarrassing and good enough to power many of the functions for their consumer app Muse.
然而在周一，Meta 推出了 Muse Wonder，这是一个新的更大规模的模型系列。
On Monday, however, Meta introduced Muse Wonder, a new larger class of models.
这次发布包含两个模型变体：Wonder 和 Wonderful。
The release included 2 model variants, Wonder and Wonderful.
Wonder 是为速度优化的精简版，Wonderful 是通过 API 提供的完整版。
Wonder is the slimmed-down version optimized for speed, and Wonderful is the full version available through the API.
今天我们要聊聊这些模型的表现、社区的早期看法和使用体验、这对 AI 竞赛意味着什么改变，以及它对前沿节奏的启示。
Today we're going to talk about how the models perform, early takes and experiences from the community, how this changes the AI race, and what it says about pacing the frontier.
在基准测试中，Wonder 的两个版本都高居榜单前列。
On the benchmarks, both versions of Wonder show up near the top of the rankings.
Wonderful 在 DeepSuite 1.1 上拿到 79%，比 GPT-6 Astra 和 Opus 5.5 高出将近 5 分；在 TerminalBench 4.0 上，只比那些前沿对手低 2 分。
Wonderful scored 79% on DeepSuite 1.1, which is almost 5 points ahead of GPT-6 Astra and Opus 5.5, and on TerminalBench 4.0, it was behind those frontier rivals by just 2 points.
在 agent 任务上，Wonderful 似乎略微领先。
For agentic tasks, Wonderful seemed to have a slight lead.
它在 Meta 内部的个人超级智能指数上排名第一，比 Astra 和 Opus 高出大约 4 分。
It topped Meta's internal Personal Superintelligence Index, beating Astra and Opus by around 4 points.
它还在法律工作的 Harvey Lab、Val 的金融 agent 和 Automation Bench 上达到了最先进水平。
It was also state-of-the-art across Harvey Lab for legal work, Val's finance agent, and Automation Bench.
在测试计算机使用的 OSWorld 上，Wonder 比前沿对手低几分，但依然明确处于第一梯队。
On OSWorld, which tests computer use, Wonder was a few points behind frontier rivals, but still decidedly in the mix.
Artificial Analysis 也验证了 Meta 的结果，给 Wonder Ultra 的智能指数打了 56 分，Wonder 打了 52 分。
Artificial Analysis backed up Meta's results, giving it a 56 on the intelligence index for Wonder Ultra and 52 for Wonder.
这让 Wonder Ultra 与 Sonnet 5.5 并列第二，落后 Opus 5.5 两分。
That makes Wonder Ultra tied for second against Sonnet 5.5 and 2 points behind Opus 5.5.
这一结果让它排在Claude Haiku 5.1和GPT-6 Astra的前面。
It places it ahead of Claude Haiku 5.1 and GPT-6 Astra.
Wonder Slim稍微靠后一点，排在第七位，与GPT-6.1 Sol和Gemini 4 Argon并列，后者当然还没有发布。
Wonder Slim is a little further behind in 7th place, tied with GPT-6.1 Sol and Gemini 4 Argon, which is of course yet to be released.
Artificial Analysis强调，用他们的话说，这比Muse Spark 1.3有“实质性提升”，而且距离那个模型发布才六周。
Artificial Analysis highlighted that this is a, quote, material gain over Muse Spark 1.3 and comes just 6 weeks after that model's release.
主要的提升出现在智能体专业工作和编码任务上，包括在GDPVal-AA上排名第二，在Double-A Briefcase上排名第三。
The big gains were on agentic professional work and coding tasks, including a 2nd place ranking on GDPVal-AA, and 3rd place on Double-A Briefcase.
Muse Wonder相当贵，每个任务要4.52美元，价格几乎是Muse Spark 1.3的三倍。
Muse Wonder was pretty pricey at $4.52 per task, which represents almost a 3x increase from Muse Spark 1.3.
这样算下来，每个任务比GPT-6 Astra或Grok 4.7贵大约一美元，比Opus 5.5便宜大约1.5美元。
That puts it about a dollar per task more expensive than GPT-6 Astra or Grok 4.7, and about a buck 50 cheaper per task compared to Opus 5.5.
如果单看基准测试，Meta是时隔多年重新回到了前沿。
If we then went purely from the benchmarks, Meta is back at the frontier for the first time in years.
对一些人来说，这本身就已经让人惊掉下巴了。
For some, that was jaw-dropping all on its own.
Ollie Lehmann写道，我们到底处在什么时间线？扎克伯格现在都能压制Astra和Haiku了？
Ollie Lehmann writes, what timeline are we in where Zuck is mogging Astra and Haiku now?
无穷的钱加无穷的发行渠道，现在又加上前沿智能？
Infinite money plus infinite distribution and now Frontier Intelligence?
看多Meta。
Long Meta.
Shopify的Tobi Lutke写道，这个模型非常非常棒。
Shopify's Tobi Lutke writes, this model is very, very good.
我费了好大劲都找不到一件它做不了的事，哪怕其他顶级模型已经失败了。
I was struggling to find anything to throw at it that it couldn't do, even when other top models failed.
Meta回来了。
Meta is back.
但基准测试永远不是全部，没过多久大家就开始对这股兴奋的宣传挑刺了。
And yet the benchmarks are never the full story, and it wasn't long before folks started poking some holes in the excited press.
Max Blade给了Muse Wonder和Opus 5.5同一个提示词，让它们构建一个AAA级割草机模拟器，他写道，很明显Muse在3D任务上离前沿水平差得很远，比如电子游戏，甚至可打印的CAD设计。
Giving Muse Wonder and Opus 5.5 the same prompt to build a AAA-level lawnmower simulator, Max Blade wrote, it's very clear that Muse is not anywhere near close to frontier in 3D tasks like video games or even printable CAD designs.
这预示着我们接下来会看到的另一个抱怨，他继续说，更吓人的是它烧掉了我50美元的预算，外加4.11美元的API等价代币。
Presaging another complaint that we will see, he also continued, even scarier is it burned my $50 spending plan plus $4.11 in API equivalent tokens.
我不得不升级到200美元的套餐，才能完成这次运行。
I had to upgrade to the $200 plan just to finish this run.
我就不给你举数百万个Muse Wonder在这种测试上表现不佳的例子了，但我想让你看看Theo的评论，他写道，已经看到很多人不把Muse Wonder当回事，因为它不像Astra和Opus那样在3D方面做过强化学习训练。
I will spare you the million more examples of Muse Wonder underperforming on that type of test, but point you to this comment from Theo who writes, Already seeing a lot of people discount Muse Wonder because it wasn't RL'd in 3D stuff like Astra and Opus.
3D游戏演示很酷，但根本不能代表真正的编码工作。
3D game demos are cool, but they don't represent real code work at all.
我特别特别同意这一点。
I could not agree more strenuously with this.
事实上，我认为这对整个行业和整个世界都是一个极其负面的净影响，因为3D游戏和渲染在社交媒体上看起来很好看，前沿实验室就觉得他们都必须针对这种用例进行优化，因为他们知道X上的人会根据那个用例来评价。
In fact, I think that it is actually extremely net negative to the entire industry and world around it that because 3D games and renderings look good on social media, the frontier labs feel like they all have to optimize for that use case because they know that people on X are going to judge it based on that use case.
Meta的首席AI官Alexander Wang甚至补充说，整个团队现在都觉得自己很傻，但确实如此。
Meta's Chief AI Officer Alexander Wang even added, Whole team feels dumb now, but it's true.
我们并没有明确地用3D演示游戏来训练我们的模型。
We did not explicitly train our model on 3D demo games.
显然，回头想想，我们大概也应该做一点，不过一点也不后悔。
Obviously, in retrospect, we probably should have done a bit of that, but no regrets.
请你在真实的编码工作中试试Muse Wonder。
Please try Muse Wonder on your actual coding work.
不幸的是，问题不止出在那些3D任务上，有些人在别的地方也发现了问题。
Unfortunately, it wasn't just those 3D tasks where some people found issues.
AI研究员Mehul Malhan写道，抱歉，但Muse Wonder完全没让我觉得惊艳。
AI researcher Mehul Malhan writes, sorry, but not impressed with Muse Wonderful at all.
远不及Opus 5.5的表现。
Nowhere near Opus 5.5 performance.
我觉得甚至还不如GPT-6.1 Sol。
I'd argue not even GPT-6.1 Sol.
我不确定其他人用的是什么模型，但那些让它能和 Opus 相比的基准测试数字看起来很奇怪。
I'm not sure what model others are using, but the benchmark numbers making it comparable to Opus seem weird.
不是价格的问题，就是质量本身很差。
Not pricing, just quality itself is bad.
这些是真实的编程任务，用的是我自己的 agent。
These are real programming tasks and my own agents.
Muse Wonder 在我们的资深工程师基准测试中的表现令人失望。
Disappointing show for Muse Wonder on our senior engineer benchmark.
它要求模型从头重写一个 vibe-coded 的垃圾数据库。
It asks the model to rewrite a vibe-coded slop database from scratch.
这是衡量它对现实工程有用性的一个很好的指标。
A good measure of its usefulness for real-world engineering.
Muse Wonder 得了 52%，甚至远落后于 GPT-6.1 Sol 的 71%。
Muse Wonder gets 52%, far behind even GPT-6.1 Sol at 71%.
用了 Muse Wonder 5 个小时。
Used Muse Wonder for 5 hours.
本来打算做个视频。
Was gonna make a video.
没必要了。
No need.
换回 Codex 和 Claude Code。
Back to Codex and Claude Code.
有人推测，问题实际上是 Muse Wonder 和 Muse Wonderful 之间相当大的差距。
Some speculated that the issue was actually the fairly significant gap between Muse Wonder and Muse Wonderful.
Muse Wonderful 看起来确实很好，但 Muse Wonder 在我的所有测试里表现都不太好。
Muse Wonderful looks really good, but Muse Wonder isn't doing very well in any of my tests.
所以现在大家都从 Muse Wonder 那里得到糟糕的结果，而这让 Muse Wonderful 跟着受影响，因为它们名字一样。
So now everybody is getting bad results from Muse Wonder, and it reflects badly on Muse Wonderful because they have the same name.
老实说，我觉得这两个在不同时间发布会更好。
I honestly think it would have been better to release these at different times.
我要指出一个能体现团队投入程度的事情：Alexander Wang 回复了那条评论，说“公平的反馈，我私信你了解更多。”
I will note, giving a sense of how engaged the team is, Alexander Wang responded to that saying, fair feedback, DMing you to understand more.
我还看到不少其他用户有类似的反馈，他们对 Wonderful 的体验很好，但对更精简的 Wonder 的体验很差。
And I saw a number of other users suggesting something similar where they had good experiences with Wonderful but bad experiences with the slimmer Wonder.
另一个主要抱怨是套餐用量消耗得太快。
The one other big complaint was around how fast the plans got consumed.
Arafat Shuvo 写道：“我连用 Muse Wonder 正常工作都做不到。”
Arafat Shuvo wrote, I can't even do work properly with Muse Wonder.
它太快烧完每周限额了。
It burns through the weekly limit so fast.
Matt Sylve 写道：“我刚升级到 Muse 每月 50 美元的套餐，好试试 Wonderful。”
Matt Sylve wrote, I just upgraded to the Muse $50 a month plan to try Wonderful.
即使一个小时前用量限制刚重置，它在不到 10 分钟内就用掉了我 5 小时用量的 13%。
Even after the usage limit reset an hour ago, it used 13% of my 5-hour usage in less than 10 minutes.
不过 Alexander Wang 确实说过，可能有一个 bug 错误计算了人们用了多少套餐额度，但不管怎么说，无论是什么原因，这都没能改善整体上的第一印象。
Now, Alexander Wang did report that there might have been a bug that was miscounting how much people were using of their plans, but still, whatever the cause, it didn't help the overall initial impressions.
然而 Muse Wonder，或者更准确地说 Muse Wonderful，也确实赢得了一些支持者。
And yet Muse Wonder, or specifically Muse Wonderful, did win some adherents as well.
开发者兼企业家 Theo 再次把 Muse Wonderful 排进了他的 S 级，和 Opus 5.5 并列，其他一切都远低于这个级别。
Developer and entrepreneur Theo again put Muse Wonderful in his S tier alongside Opus 5.5, and everything else being far below that.
举个例子，他写道：“让 Opus 5.5、GPT-6.1 Sol 和 Muse Wonderful 修一个 bug。”
Giving an example, he wrote, Asked Opus 5.5, GPT-6.1 Sol, and Muse Wonderful to fix a bug.
Opus 找到了根本原因，并提交了一个 PR。
Opus found the root cause and filed a PR.
GPT-61-Sol 问了个问题，然后因为异步提问工具中断而崩溃，接着找到了 bug，修好了，写了 300 行测试，还用一种让人完全看不懂的方式解释了它。
GPT-61-Sol asked a question, crashed from the async question tool interrupt, found the bug, fixed it, wrote 300 lines of tests, and explained it in impossible-to-understand terms.
MuseWonderful 发现了别人提交的一个小 PR，然后说：“你们也许应该合并这个。”
MuseWonderful found a small PR someone else had filed and said you should probably merge this instead.
我真的很喜欢这个模型。
I really like this model.
Conductor 的 Charlie Holtz 也很喜欢这个模型。
Conductor's Charlie Holtz also liked the model.
他写道，快速印象：一次就发现了我们代码库里的一个真实安全漏洞，其他代理都没找到。
He wrote, quick impressions: found a real security vulnerability in our codebase in one shot that other agents missed.
它在设计和可视化方面很厉害。
It's great at design and visualizations.
它特别喜欢用子代理。
It really likes subagents.
命令行界面里也有很多不错的细节。
And lots of great details in the CLI.
对这个进展印象非常深刻。
Super impressed by the progress here.
我提前体验了 Muse Wonder，它确实是个令人惊喜的模型。
I had early access to Muse Wonder, and it is a surprising model.
在我们跑过的所有技术基准测试里，它都和 Astra、Fable 和 Opus 不相上下，甚至更好。
On every technical benchmark we ran, it was neck and neck or better than Astra, Fable, and Opus.
它的处理很细腻。
It has a delicate touch.
它会做出很有灵感的创意决策。
It makes inspired creative decisions.
现在还没有衡量灵魂感的基准测试，但 Muse 有这个特质。
There aren't really benchmarks for soulfulness, but Muse has it.
它是一个温暖的模型，一个能干的模型，而且有那种大模型的气息。
It's a warm model, a capable model, and it has big model smell.
它的输出很开心、很天真、也很友好。
Its outputs are happy, innocent, and friendly.
它从来不会像某些模型那样对你居高临下。
It never talks down to you like some of the other models.
它就想帮你做出很酷的东西。
It wants to help you make dope stuff.
它也希望你能赚钱。
It wants you to make money.
它跟你是一条心的。
It's aligned with you.
Meta 在消费级代理这块会大获全胜。
Meta is going to win consumer agents in a landslide.
Claire Vo 按类别细分了她的测试，发现它在前端设计上表现很差，除了写邮件之外的其他写作也不行，它对 SVG、视频和创意设计也不擅长，但用她的话说，在编程上它能跟大模型们平起平坐，而在代理任务上，它直接碾压了，在她测试的 8 个前沿模型里排第一。
Claire Vo broke her tests down by category, finding that on front-end design it was quite bad, writing anything other than emails, it was bad, that it wasn't great for SVGs, videos, and creative design, but that, as she put it, it rolls with the big boys on coding, and on agentic tasks, it crushed, coming in first out of 8 frontier models she tested.
那么一个有趣的问题是，它在 Meta 的战略里会怎么定位。
Now, one interesting question is where this is going to fit in Meta's strategy.
当然，Meta 比其它实验室更专注于消费者体验，因为他们有消费者应用的分发渠道和网络。
Meta is, of course, much more focused on the consumer experience than other labs are, given their distribution and network of consumer apps.
大概一个月前，我注意到 Muse 的一个独特之处是，它是第一个大多数用户根本不在乎背后是什么模型的 AI 产品。
About a month ago, I had observed that one of the things that was distinct about Muse was that it was the first AI product where most of its users just didn't care what model was powering it.
但尽管我相信这是真的，还是有相当多的用户开始遇到了 Muse 的一些限制。
And yet, even though I believe that that is true, there were a fair number of users that started to run up against some limits in what Muse could do.
我在 X 上发了那个帖子一周后，Sergito 写道，有没有人觉得 Muse 不如 Instinct 聪明？
A week after I posted that on X, Sergito wrote, Anyone else finding Muse to be less smart than Instinct?
我给它们俩布置了类似的多步骤任务，Muse 自己瞎编了一些东西，然后试图修复时又不断放大错误。
I gave them both a similar multi-step task and Muse hallucinated some stuff and then kept compounding its error when trying to fix it.
事实上，我们现在的处境有点像之前讨论过的，高阶用户会针对不同任务用不同模型，很多早期用户也在同时用多个不同的个人助手。
And in fact, we've kind of been in a spot where in the same way that I discussed power users using different models for different tasks, a lot of early adopters are using multiple different personal assistants.
Brycent 最近在 X 上发帖说，我意识到根本就没有什么完美的个人助手。
Brycent recently posted on X, what I've realized is that there is no one optimal personal assistant.
目前我同时在用 Instinct、Muse、Grokbot 和 Dots。
At the moment, I'm currently using Instinct, Muse, Grokbot, and Dots.
我真的挺享受在特定场景下逐个使用它们的感觉。
I genuinely enjoy using them all for specific use cases.
事情可能确实会这样发展，但我觉得对很多人来说，在多个个人助手之间切换的转换成本，会比在其他方式下切换模型的成本高得多。
Now that may be how it plays out, but I think for a lot of folks, the context-switching cost of using multiple personal assistants is going to be significantly higher than the context-switching cost for using models in other types of ways.
换句话说，LLM 和编码代理在很多场景下都能提供巨大价值，即使它们对你一无所知。
In other words, there are many use cases for LLMs and coding agents in general where they can provide a ton of value even if they don't know anything about you.
而个人助手则相反，它们对你的了解程度以及能访问的系统越多，就越有用。
Personal assistants, on the other hand, get more useful in proportion to what they know about you and the systems of yours they have access to.
大多数人连让一个个人助手访问所有这些信息都很难，更别说半打了。
And most people are going to have a hard time giving even one personal assistant access to all that information, much less a half dozen.
考虑到这一点，最近社交媒体的页面上到处都能看到对 Muse Wonder 这类工具的需求。
And with that in mind, the need for something like Muse Wonder is written all over the recent pages of social media.
投资者兼创业者 Satya Malik 最近写道，虽然 Muse 在执行即时任务（比如订航班、订酒店）方面很擅长，但用他的话来说，它在长期任务上表现很糟糕。
Investor and entrepreneur Satya Malik recently wrote that while Muse is very good at immediate tasks like booking flights and hotels, it is, in his words, terrible for long-term tasks.
比如他说，今天它提醒我下午 2 点有个会还没安排好。
For example, he said, it reminded me today that a meeting has not been set for 2:00 PM.
结果是个假警报，因为那个会其实是昨天下午 2 点就定好了。
That was a false alarm because the meeting was set for 2:00 PM yesterday.
这种情况已经发生过好几次了，它因为搞错日期就慌慌张张地通知我。
This has happened several times where it panics and tells me something because it gets the day wrong.
投资者 Sam Lessin 写道，昨晚试图用 Muse 订酒店。
Investor Sam Lessin wrote, tried to book a hotel with Muse last night.
希尔顿的网站让我的代理陷入了死循环。
Hilton site put my agent in a doom loop.
试了第三次，我放弃了，自己手动订了。
Third retry, I gave up and booked it myself.
很明显，Meta 团队在打造 Wonder 时就已经想到了这类挑战。
And it's quite clear that this type of challenge was on the mind of the Meta team in building Wonder.
Chubby 在 X 上写道：发布新闻里可能被淹没的一个有趣洞察是——Meta 为那些能代表你行事的代理构建了 Wonder，首先是从它的 Muse 应用开始。
Chubby on X wrote, An interesting insight that was probably drowned in the release news: Meta built Wonder for agents that act on your behalf, starting with its Muse app.
两个功能非常重要：Active Memory（主动记忆），能在对话中持续追踪对你重要的事情；以及 Active Focus（主动聚焦），能在任务遇到困难时继续推进，尝试其他方法，并从错误中恢复。
2 Features are really important: Active Memory, which keeps track of what matters to you across conversations, and Active Focus, which keeps a task going when it gets hard, tries other approaches, and recovers from errors.
不过，看了发布媒体后，我还有一个问题：借助 Wonder 模型系列，Meta 是不是真的开始认真追求不仅仅是个人代理场景，也包括专业场景了？
Another question I had though, looking at the launch media, is whether with the Wonder family of models, Meta is actually getting serious about going after not just personal agentic use cases, but professional ones as well.
当然，Mark Zuckerberg 称其为个人和专业代理工作的顶尖性能。
Certainly, Mark Zuckerberg called it state-of-the-art performance for personal and professional agentic work.
Alexander Wang 在介绍时写道：在个人代理和专业代理方面达到顶尖水平。
Alexander Wang, in introducing it, wrote state-of-the-art at personal agents and professional agents.
AI at Meta 账号也明确提到了个人代理和专业代理，这意味着他们进行了协调一致的信息传播，确保提及专业代理。
And the AI at Meta account also explicitly mentioned both personal and professional agents, meaning that there was a coordinated messaging campaign to make sure to mention professional agents.
这可能意味着 Meta 这次认真对待专业场景了，也可能意味着他们明白个人代理中隐含了专业场景，从公司战略角度划清界限或许说得通，但这并不符合消费者对自己的认知。
Now, that could mean that Meta is serious about professional use cases this time, or it could mean that they understand that implicit in a personal agent are going to be professional use cases, and that drawing a clear line might make sense from a company strategy standpoint, but doesn't match how consumers think about themselves.
换句话说，可能并不是 Meta 突然想跟 Claude Code 在企业级交易上竞争，而是他们相信，在个人层面，如果消费者真的选择并投资于一个单一的个人代理，那么这个个人代理在很多情况下也必须能处理专业事务。
In other words, it may not be that all of a sudden Meta wants to compete with Claude Code for enterprise deals, but that they believe that on an individual level, if consumers really choose and invest in a single personal agent, that personal agent is going to have to do professional stuff in many cases.
当然，还有另一类专业人群对 Meta 来说已经非常重要，那就是小企业主，他们已经在 Facebook 和 Instagram 等其他应用上使用了大量功能。
There is of course another category of professional that is already deeply clearly important to Meta, which is small business owners who already use tons and tons of features on their other apps like Facebook and Instagram.
这正是 Citrini Research 看到可行性的地方，他们写道：我们的结论是，Meta 很有机会抓住小企业的代理市场。
That's where Citrini Research saw the viability, writing, our takeaway is that Meta has a real shot at capturing agents for small business.
它在能力和执行力之间取得了非常好的平衡。
It's a really good balance of capability and execution.
对很多人来说，大新闻不是终点的 Wonder 能做什么，而是变化的速度和幅度。
For a lot of folks, the big story is not the endpoint and anything Wonder can do, but the delta and the rate of change.
Pulseas Bencerra 写道：Meta 正在以闪电般的速度前进。
Pulseas Bencerra wrote, Meta is moving at lightning speed.
Muse 34 天前才发布，现在他们已经加入了前沿行列。
Muse launched 34 days ago, now they've joined the frontier.
Ethan Mollick 表示，从有限的测试来看，它似乎是个不错的模型，但引用他的话说：也许更令人印象深刻的是对比进展。
Ethan Mollick said that it seems to him like a good model from a limited test, but, quote, perhaps more impressive, compare the progress.
Meta 在短时间内取得了非常快的进步。
Very fast progress from Meta in a short time.
Yuchen Jin 写道：一年前，很多人嘲笑 Meta 花数十亿美元聘请顶级 AI 人才并收购 Alexandr Wang。
Yuchen Jin wrote, A year ago, many dunked on Meta for spending billions to hire top AI talent and acquire Alexandr Wang.
老实说，我当时也持怀疑态度。
I was skeptical too, to be honest.
但是Muse应用和模型的成功证明了一件事：雇佣顶尖人才，组建一支超强团队，给他们一个宏大的目标和足够的资源，他们就能创造奇迹。
But the success of the Muse app and models proves one thing: hire top talent, assemble a cracked team, give them an ambitious goal and enough resources, and they can work wonders.
市场显然很买账，消息公布后Meta股价上涨了4%。
Certainly markets liked what they saw, with Meta up 4% on the news.
但这对整个关于pacing the frontier的讨论意味着什么？
But what does this mean for the whole pacing the frontier conversation?
在各大AI首席执行官中，Zuckerberg与其他人的公开表态并不特别一致。
Among the AI CEOs, Zuckerberg has not been particularly aligned with the messaging coming from the others.
他没有签署Pacing the Frontier公开信，而且在这封信发布几周后，他用一篇题为"The Future is for Everyone"的文章做出了反向表态。
He did not sign the Pacing the Frontier letter, and a few weeks after it was published, he countersignaled with his essay, The Future is for Everyone.
他的核心论点是，开放和广泛的访问必须成为AI时代的核心原则，即使这会带来一些额外的风险。
His core thesis is that open and widespread access needs to be the core tenet for the AI era, even if that carries some extra risk.
在Zuckerberg看来，将未发布的模型仅供内部使用所带来的权力集中，可能比发布强大AI模型所伴随的风险更具灾难性。
In Zuckerberg's view, the concentration of power that comes from keeping unreleased models for internal use only could be even more disastrous than the risks associated with releasing powerful AI models.
现在很明显，这批Wonder模型并没有推动前沿，而且似乎确实有非常强大的网络安全护栏。
Now clearly, this set of Wonder models isn't pushing the frontier, and does seem to have very strong cyber guardrails.
也就是说，与已经可用的模型相比，它们并不危险。
Meaning that they are not dangerous compared to what's already available.
但从发展轨迹来看，Zuckerberg对放缓毫无兴趣。
But from a trajectory perspective, Zuckerberg is showing no interest in slowing down.
在接受Financial Times采访时，他宣称Meta已经到达了前沿，并且引述他的话，"已经在着手下一步的跃升了"。
In an interview with the Financial Times, he declared that Meta has reached the frontier and is, quote, already working on that next step up.
他不仅没有放缓的意思，还补充说，模型开发"从这里开始可能会感觉快很多"。
Rather than slowing down, he added that model development will, quote, probably feel like it'll go quite a bit faster from here.
他拒绝给出自己的P-doom数值——也就是AI杀死所有人的概率，这个话题在过去一个月里让全世界都炸了锅。
He declined to give a number for his P-doom, the odds of AI killing everyone that got the entire world riled up over the past month.
相反，他说："我更倾向于相信，AI是一种可以用正常护栏进行工程化、可以负责任地构建的技术，而不是那种拟人化的AI愿景——也就是AI会自己跑去做自己的事情。"
Instead, he said, I'm much closer to the belief that AI is a technology that can be engineered with normal guardrails and built responsibly than the anthropomorphic vision of AI, which is that it's just going to go do its own thing.
不管你认为这有没有可能，Meta与其他实验室在如何看待自己正在构建的东西上，似乎确实存在根本性的差异。
Whether you agree with whether it's possible or not, there does seem to be a fundamental difference in how Meta views what they're building from the other labs.
简而言之，OpenAI和Anthropic要么在努力成为、要么觉得自己不可避免地正在建造令人敬畏的机器之神，而Meta试图构建的只是一种普通技术。
In short, OpenAI and Anthropic are either trying to or feel like they inevitably are building machine gods that inspire awe, while Meta is trying to build what it views as just a normal technology.
谁对谁错仍然是争论的焦点，但这就是根本的分界线。
Who's right or wrong about that remains the debate, but that is the fundamental dividing line.
总的来说，退一步看，我们进入2026年时有3家前沿模型实验室，分别是OpenAI、Anthropic和Google。
Overall, when you take a step back, we headed into 2026 with 3 frontier model labs in OpenAI and Anthropic and Google.
随着coding agents强势崛起，并证明了它们比其它一切都更重要——至少在短期内如此——这个数字降到了2。
As coding agents forced their way to the top and showed that they were more important than everything else, at least in the short term, that number went down to 2.
然而现在，尤其是如果Gemini 4 Argon能达到Google宣称的水平，我们又回到了美国有差不多5家前沿实验室竞争的格局。
And yet now, especially if Gemini 4 Argon can do what Google says it can do, we are back to having 5-ish frontier labs competing in the US.
不过话说回来，正如我们今天在头条新闻中听到的，幕后可能还有更强大的模型即将出现。
Then again, behind closed doors, as we heard in the headlines today, there might be much more powerful models just on the horizon.
所以也许那个前沿双头垄断很快就会回归。
So maybe that frontier duopoly will be right back soon.
就目前而言，Meta似乎已经完全回到了竞争之中。
For now, it appears that Meta is fully back in the game.
这就是今天AI Daily Brief的全部内容。
That's going to do it for today's AI Daily Brief.
一如既往，感谢您的收听或收看。
Appreciate you listening or watching as always.
下次再见，祝您平安。
And until next time, peace.