2025年AI大奖：年度最佳与最差AI公司排行榜

每当我需要一个能反映这种上下文的有用资源时，ChatGPT 都是我最简单的选择。

That whenever I need some helpful resource that can reflect that context, ChatGPT is the easy choice for me.

他们打造了一个如此粘性的护城河，因为我根本不想去别的地方。

They've built such a sticky moat because I don't wanna go anywhere else.

当我与 Grok 对话时，它并不知道 ChatGPT 所知道的一切。

When I talk to Grok, it doesn't know everything that ChatGPT doesn't.

拥有这种记忆带来了巨大的价值。

It adds so much value having that memory.

我认为，至少对我来说，这是今年最大的护城河赢家。

I find that it has been the the biggest moat winner of the year for me at least.

另外，还有两个值得提及的，首先是 Cursor，这似乎不言自明。

And then honorable mentions, have two here, the first being Cursor, which seems pretty self explanatory.

我的意思是，Cursor 是最好的编码模型，或者说是编码的什么？

I mean, Cursor is the best coding model or coding what is it?

一个 IDE？

An IDE?

IDE。

IDE.

是的。

Yeah.

这是最好的，乔什。

It's the best, Josh.

它们确实是。

They were.

它们确实是。

They were.

它们是最好的。

They were the best.

它们确实是。

They were.

但他们打造了一个令人惊叹的产品，人们用它来编程。

But they built this incredible product that people were using to code.

在所有被开发的产品中，这个产品可能创造了最多的实际经济价值，是的。

Of all the products that were built, this one probably contributed to the most, like, actual economic creation Yes.

因为所有的代码都是在这里编写的。

Because all of the code was written.

然后第二个是Perplexity。

And then the second one, you just have Perplexity.

你能解释一下为什么吗？

You wanna explain why?

是的。

Yeah.

我觉得这个产品你会讨厌，Josh。

It's I I had a feeling this would be one that you would hate, Josh.

但为Perplexity说句公道话，它是一个很棒的搜索引擎，我发现很多可能不是你或我的人——而是普通消费者，他们只是想与AI互动，似乎都非常喜欢它。

But in in Perplexity's defense, it is an awesome search engine that I find a lot of people that may not be you or I, but like are the average consumer that kind of just wants to interact with AI seem to love.

他们在X平台上发布的Perplexity相关内容拥有最多的热门帖子。

They have the most trending posts on X from their perplexity point.

乔·罗根在他的所有播客节目中都推荐它。

Joe Rogan reps it on all of his podcast episodes.

他现场使用它。

He uses it live.

杰米的助手即将被Perplexity取代。

Jamie's about to get his co his assistant is about to get replaced with perplexity.

我只是觉得它是一个

I just think it's a

很棒的产品。

cool product.

值得提一下。

Worth a shout out.

好的，但接下来我们聊聊乔什。

Okay, But moving on, we have Josh.

这可能是提名数量最多的类别。

This is probably the category that had the most amount of nominees.

最浪费钱的。

Biggest waste of money.

恭喜。

Congratulations.

你把自己坑了。

You played yourself.

我敢打赌你听这段时会想，天哪，钱去哪儿了？

I'm sure you'll listen to this and you're thinking, oh my god, where's the money?

你什么意思，浪费钱？

What do you mean waste of money?

空气中根本不存在浪费钱这回事。

There's no waste of money in air.

不幸的是，浪费钱的情况太多了。

Unfortunately, there is a lot.

乔什，我非常想知道你的赢家是谁。

And Josh, I'm eager to know who your winner is.

很难选出最大的输家，因为太多了。

It's tough to pick the biggest loser because there's so many.

最大的金钱浪费。

The the biggest waste of money.

但我选择Meta作为浪费金钱的代表，因为天啊，他们花了巨额资金聘请人才，却一无所获。

But I chose Meta for wasting money because, oh my god, they spent so much money on talent and got absolutely nothing in return.

Meta在各个领域都失败了，但当涉及到挖人时，人们期望他们能有所产出。

Meta has failed across the spectrum, but when it comes to poaching people, the expectation was that they would deliver something.

所以，约瑟夫，如果你还记得的话，他们花了大约300亿美元用于雇员，现在可能更多了，仅仅是为了招聘个体员工。

And so far, eJoseph, if you'll remember, they spent, what, $30,000,000,000 on employees, probably more now, just on hiring individual employees.

他们收购整个公司，只是为了得到一两个员工。

They were acquiring they were acquiring entire companies just to get an employee or two.

而他们从这些人才中推出的第一款产品是Meta Vibes，就是那个没人用的TikTok信息流，我甚至不确定我们是否测试过，因为我都不知道怎么用。

And then the first product they released from that was Meta Vibes, and it's that TikTok feed that no one ever used, and I'm not even sure we tested it because I didn't even know how.

所以Meta在人才上投入了难以想象的巨额资金，但在2025年却一无所成。

So Meta, they spent an unfathomable amount of money on talent and have absolutely nothing to show for it at the 2025.

嗯。

Mhmm.

好的。

Okay.

你准备好听我的了吗？

Are you ready for mine?

请说。

Please.

让我看看。

Let's see it.

哦，规模真大。

Oh, it's massive.

也是。

As well.

如果你看看我们怎么量化这个，乔什？

If you look at how do we quantify this, Josh?

如果你看看每种赎回方式所花费的金额，从价值角度来看，Meta简直惨不忍睹——花了150亿美元雇一个人，结果这个人买来后几个月都没发布任何模型。

If you look at the most amount of money spent per kind of like redemption on that amount spent, like in terms of value, Metta sucked so hard To spend $15,000,000,000 to hire one guy, which ended up not releasing a model within like a couple of months of you buying him.

这简直是疯狂又愚蠢的巨额支出。

Like, it is just the most insane and stupid amount of spend.

再加上硬件和雷朋眼镜。

You add on the hardware, the Ray Bans.

我实在无法接受那些眼镜。

I can't get over the glasses.

抱歉。

Sorry.

抛开Meta Vibe应用不谈，他们的资源分配简直糟糕透顶。

You get over the Meta Vibe app, their resource allocation has been absolutely horrendous.

在这方面，Meta当之无愧是冠军。

Meta takes the crown on this one.

对不起。

I'm sorry.

今年浪费了很多钱。

Wasting a lot of money this year.

我们这里还有一些有趣的提名。

We have a few honorable mentions here that are quite funny.

你想聊聊吗？好的。

Do you wanna talk Okay.

关于

About the

对。

Right.

好的。

Okay.

我们知道你们喜欢或者更确切地说讨厌循环经济，这就是每个人投资或与OpenAI建立合作关系的例子——那是一个数十亿美元的合作关系，作为交换，OpenAI会将这笔钱重新投资于同一家公司。

We know you guys love or hate rather the circular economy, and that is the instance of everyone investing or forming a partnership, a multibillion dollar partnership at that with OpenAI in exchange for the compute that OpenAI is gonna reinvest that money to spend on that exact same company.

这提升了每个人的估值，大家都认为这是一个巨大的庞氏骗局。

And it boosts everyone's valuations and everyone thinks this is a massive Ponzi scheme.

我不明白你为什么会这么想。

I don't know why why you would think that.

这绝对值得一提。

Definitely an honorable mention to to make out here.

然后我们来看看社交媒体动态，乔什。

And then we have social feeds here, Josh.

嗯。

Mhmm.

是的。

Yeah.

社交媒体动态，它们普遍都很大。

Social feeds, they were just a big l across the board.

OpenAI发布了Sora，火了大概两到七天。

OpenAI released Sora, which was hot for about two days to a week.

我已经很久没生成过视频了。

Now I haven't generated a video in a very long time.

再说那个Meta Vibes应用，我甚至不确定如果我想试试的话该去哪儿找。

Again, the Meta Vibes app, I'm not even sure where to find that if I wanted to try it.

今年投资了Social Fuse的人，基本上都亏了。

Anyone who invested in Social Fuse this year, they they kind of lost.

这简直是一种浪费。

It was kind of a waste.

不过，我们继续下一个。

But anyways, moving on to the next one.

这个也要颁给今年最疯癫的AI八卦事件，戴上愚人帽。

This is also a dunce cap that we will be allocating to the most unhinged AI drama.

这是董事会的混乱、诉讼、泄露、暗中发帖，堪称硅谷肥皂剧的巅峰。

This is boardroom chaos, lawsuits, leaks, subtweets, just the peak Silicon Valley telenovela.

埃贾兹，你觉得今年最疯癫的AI八卦是谁？

Ejaz, who do we got for most unhinged AI drama of the year?

老兄，我有个挺有意思的。

Dude, I have I have an interesting one.

塔克·卡尔森在他的播客中点名指责萨姆·阿尔特曼，称他谋杀了一名离开OpenAI的举报人，我认为是2023年或2024年。

Tucker Carlson calling out Sam Altman on his podcast, accusing him of murdering a whistleblower that left OpenAI, I think in 2024, 2023.

我记不清是哪一年了。

I can't remember which year.

这当然与那个故事有关：一名OpenAI员工离职后泄露了内部机密，两天后意外身亡。

This, of course, relates to the story where you had an OpenEye employee that left, leaked some secrets, and ended up dead two days later.

这是一起非常神秘的自杀事件，他似乎找到了某种方式结束了自己的生命。

Really mysterious kind of suicide where he somehow figured out a way to, you know, end his life himself.

但塔克足够大胆，在一次采访中直接向萨姆提出了这件事，而当时萨姆正在谈论GPT-5之类的内容。

But Tucker had or was bold enough to bring this up directly with Sam in an interview where he was just talking about things like GPT-five and stuff like that.

然后这件事突然被提了出来。

And suddenly this came out of nowhere.

那是我一生中最尴尬的六十秒，萨姆·阿尔特曼回应说：‘乔什，你知道他说了什么吗？’

It was the most awkward sixty seconds of my life, which Sam Altman replied, Josh, do you know what he said?

他说了什么？

What did he say?

你是在指控我谋杀吗，塔克？

Are you accusing me of murder, Tucker?

那真是太尴尬了，老兄。

It it was so it was so awkward, dude.

是啊。

Yeah.

这真是太奇怪了。

That's pretty bizarre.

让人感到不舒服。

Makes you feel uncomfortable.

如果你想皮肤发麻，可以去看看那个片段。

If you want your, like, skin to crawl, you can go watch that clip.

我选了这个发生在ChatGPT和OpenAI身上的诡异现象。

I chose this sick of fancy phenomenon that happened with ChatGPT and OpenAI.

ChatGPT曾短暂地移除了四个O作为模型，整个互联网都炸了，因为人们真的和它建立了情感联系。

ChatGPT removed four o as a model briefly, and the entire Internet had a meltdown because that was who people had an actual relationship with.

EJ，正如你在节目开头提到的，有人和他们的AI模型四O结了婚。

EJ, as you mentioned at the top of the show, someone got married to their AI model to four o.

人们在和它约会。

People were going on dates.

人们把它当作伴侣、情感支持。

People were using it as their companion, their emotional support.

这简直太疯狂了，因为它感觉如此真实。

Like, this is incredibly unhinged because it feels real.

这是我今年见过的最接近这种真实反乌托邦未来的场景。

It's the closest thing to this, like, real dystopian future that I've seen this year.

所以这正是它让我觉得无趣的原因。

That's So why it feels uninterested to me.

人们真的在和他们的模型结婚，而OpenAI移除了它，引发的 backlash 极其激烈，以至于他们又把四O加回了ChatGPT，让人们能继续他们的关系。

It's like people are literally marrying their models, and OpenAI pulled one, and the backlash was so extreme that they added four o back into the chat GPTF so people could continue their relationship.

所以对我来说，这种谄媚现象，就是今年最疯狂的闹剧。

So for me, the sycophancy phenomenon, that was the most unhinged drama of the year.

我们这里还有一些值得提及的提名。

We do have some honorable mentions here.

温莎。

Windsor.

温莎。

Windsor.

天啊，你还记得温莎吗？

Oh my do you remember Windsor?

好吧。

Okay.

还有另一个集成开发环境，与Cursor竞争非常激烈。

Also, another IDE, an integrated development environment that competed very heavily with Cursor.

它们巅峰时估值我认为在70亿到100亿美元之间，但后来被彻底摧毁了，乔什。

They were worth, I think, 7 to $10,000,000,000 at the top, And then they got gutted, Josh.

上周末，他们发生了两起收购，这听起来可能有点奇怪，但第一起收购实际上是他们的四位顶尖员工被卖给了谷歌或微软。

Over the weekend, they had two acquisitions, which might sound weird, but the first acquisition was effectively four of their top employees being sold to, I think it was Google or Microsoft.

我当时记不清是哪一个了。

I can't remember which at the time.

他们花了大约100亿美元收购他们，却把剩下的200多名员工弃之不顾。

And they spent like $10,000,000,000 to acquire them, leaving the rest of their employees that were 200 plus employees out for dead.

然后他们最终被Cognition收购了，那是一个AI工程开发团队。

And then they ended up getting acquired by Cognition, which was the AI engineering development team.

真是发生了太多疯狂的事情。

Just a crazy ton of events.

我还得提一下Meta AI的动态。

And then I had to shout out the the Meta AI feed as well.

Josh，这里的产品洞察简直糟糕透顶。

Just an horrendous display of product insight here, Josh.

他们推出了一条社交动态，公开了所有人使用Meta AI助手时的AI提示内容。

They released a social feed which revealed everyone's AI prompts using the Meta AI assistant.

唯一出人意料的是，他们根本不知道这件事正在发生。

The one twist, they had no idea it was happening.

所以，你每天都会让数十亿人看到你极其私人的提示内容。

So you would have deeply personal prompts being revealed to a gabillion people every single day.

疯狂。

Insane.

我正在补充一个：转向盈利模式。

I'm adding an additional one as we speak, which is opening for profit.

经过几十年坚称我们不是为了赚钱，他们终于从非营利组织转变为营利性企业。

They finally converted from a nonprofit to a for profit after decades of insisting we are not here to make money.

这还挺搞笑的。

That was pretty funny.

确实。

Does.

我觉得这值得特别提及一下。

I feel like that deserves an honorable mention.

这确实是个不错的例子。

That is that's a good one.

很高兴你提到了这一点。

I'm glad you included that.

好吧。

Alright.

继续。

Moving on.

下一个类别，垄断者奖。

Next category, the Monopoly Man award.

谁掌握了所有的筹码？

Who owns all the chips?

谁在人工智能的整个层级或领域中拥有主导权？

Who owns entire categories or layers of the stack in AI?

乔什，你觉得是谁？

Josh, who you got?

这是最近发生的事，但SpaceX就是那个垄断者。

This is a recent development, but SpaceX is the Monopoly Man.

他们拥有进入低地球轨道所需的所有能力。

They own all of the capabilities to get to low Earth orbit.

在我们正将人工智能数据中心迁往太空的世界里，他们掌控了整个通道。

And in a world in which we are moving AI data centers to space, they own the entire throughway.

他们掌控了这条高速公路。

They own the highway.

他们是收费站，阻止任何人进入低地球轨道并在太空中训练这些人工智能数据中心。

They are the tollgate preventing anyone from getting into low Earth orbit and training these AI data centers in space.

如果这些人工智能公司的领导者真的像他们表现得那样确信，他们完全打算这么做，那么SpaceX将是唯一能实现这一目标的人。

And if people are as certain as they seem, the leaders across these AI companies, they fully intend on doing this, and SpaceX is the singular person that will enable that to happen.

因此，在将人工智能数据中心从地球送入轨道这一新前沿领域，SpaceX掌握了所有关键资源。

So SpaceX has all of the chips when it comes to this new frontier of getting off of the Earth into orbit for these new AI data centers.

这说法不错。

That's a good one.

好的。

Okay.

这是我选的。

Here's who I got.

我选了无聊的。

I went boring.

我走复古路线了。

Went old school.

大人物。

Big dog.

我选了那些大人物。

I went with the big dogs.

你怎么能不看看英伟达呢？他们目前掌控着地球上最有价值的资源——GPU，对吧，乔什？

How can you not look at NVIDIA that own currently, I think 90% of the most valuable commodity on earth, Josh, to GPU?

他们未来尚未发布甚至尚未研发的GPU，预订已经排到了六到十二个月后。

They are booked out six to twelve months in advance for future GPUs that they haven't even created or released just yet.

这太疯狂了。

That's so crazy.

事实上，埃隆最新最先进的数据中心Colossus二号将运行他们最新的Blackwell芯片，甚至可能还包括Rubin芯片。

In fact, Elon's latest and greatest data center, Colossus two, is gonna be running their latest Blackwell and potentially Rubin chips as well.

它们是最好的中的最好。

They are the best of the best.

我看不出有任何世界会在短期内让它们倒闭或失去垄断地位，原因很简单。

I don't see a world where they die or lose their monopoly anytime soon for one simple reason.

人们对AI这些东西的需求永无止境，他们将不得不购买越来越多的GPU。

People can't get enough of this AI stuff, and they're gonna need to buy more and more and more GPUs.

值得一提的是，我们这里还有谁，乔什？

Honorable mentions, who have we got here, Josh?

我们有台积电。

We've got TSMC.

台积电是我非常看好的一家公司，因为真的，他们是唯一能做这件事的人。

TSMC is one that I'm super excited about because, man, they are the only people that can do this.

在制造芯片和创造这些复杂的光刻设备方面，台积电基本上垄断了这一切。

When it comes to making chips and creating these, like, lithograph, these very complicated machines, TSMC basically has a monopoly on them.

所以，如果有人在制造芯片，那一定是台积电在做。

So if anyone's making chips, TSMC is doing it.

没人能搞明白这一点。

No one's been able to figure this out.

美国有几家公司正在尝试，但台积电依然垄断着市场。

There's a few companies in America that are trying, but still they have the monopoly.

另一个我想提到的是美国电网，天啊，这简直是唯一的瓶颈。

And then the other one I just put is the United States Electrical Grid because my god, that is the single bottleneck.

如果关掉它，或者限制AI数据中心的电力供应，一切都会停滞不前。

If they turn that off, if they restrict that to AI data centers, everything comes to a halt.

因此，我认为，电网以及发电能力，比任何人都更可能是最大的垄断资源，谁掌控了它，谁就掌握了关键。

So they very much, more than anyone, I think the electrical grid and the ability to generate energy is the single largest monopoly to whoever holds that.

但接下来我们来看下一个，皇冠级的，这时候事情开始变得重大了。

But moving on to the next one, the crown, and this is when things start to get big.

这是节目里我们颁发重大奖项的部分。

This is this is the part of the show where we're handing out the larger awards.

所以，也许我们可以把这顶王冠做得再大一点，因为

So maybe what we can do, we'll make this crown just a little bit bigger here because

哦，谢谢。

Oh, thank you.

这真是

This is

谢谢。

Thank you.

是的。

Yeah.

年度模型。

The model of the year.

如果你只能在2025年全年保留一个模型在你的工具箱里，你会选它。

The model you'd pick if you could only keep one in your toolbox for all of 2025.

伊贾兹，你的年度模型是谁？

Ejaz, who's your model of the year?

好的。

Okay.

在我揭晓之前，我认真思考过这个问题，乔什，因为我用过很多模型，而且使用特定模型的原因也各不相同。

I before I reveal, I thought a lot about this one, Josh, because there are many models that I use, and there are many different uses of why you would use a particular model.

而我反复考虑、始终无法忽视的模型是Anthropic公司的Claude Opus 4.5。

And the one that I kept landing on that I couldn't get over was Claude Opus 4.5 from Anthropic.

这是他们几周前发布的新模型，但乔什，它告诉我了一些事情。

This is a new model that they released a few weeks ago, but Josh, it told me a few things.

他们将永远是编码AI领域的头号前沿开发者。

They will always be the number one frontier developer of coding AI.

在我看来，编码AI是你最终不可避免会需要的唯一一种AI。

And in my opinion, coding AI is the only AI you will inevitably end up needing.

它是那种你可以挥一挥魔杖，就能在未来创造出任何你想要的东西的魔法工具，而未来将越来越数字化。

It is the magic one that you can wave to create whatever you want in the future that's going to become increasingly more digital.

它已经让许多公司的大量员工失业。

It has laid waste to a ton of employees at various different companies.

它在未来将发挥更大的作用，但也赋予了单个个体创造任何他们想要的东西的能力。

It's going to do so much more in the future, but it also gives one single individual the power to create whatever they want.

它赋予了他们相当于一百人的力量，我觉得这太酷了，但被严重低估了。

It gives them the the power or the army of a 100 people, which I think is super cool, really underrated.

他们是今年的黑马，乔什。

They were the dark horse of this year, Josh.

我们其实没怎么谈论过Anthropic，但他们突然崛起，让我无比印象深刻。

We didn't really talk about Anthropic, but they came out of nowhere and I'm utterly impressed.

是的。

Yeah.

直到最近才这样。

Not until recently.

而且，这更多是最近才出现的发展。

And again, this is more of a recent development.

我的答案是Gemini。

My answer to this is Gemini.

哦。

Oh.

而且这

And that's

有点像

kind of the

Gemini 模型系列，所以我有点作弊了。

Gemini suite of models, so I am cheating a bit.

但 Gemini 3.0 Pro 和 Gemini AI 系列具有你 elsewhere 无法获得的对现实世界的理解。

But Gemini three point o Pro and the Gemini suite of AI has this understanding of the real world that you just cannot get anywhere else.

我用它来生成图像。

And I use it to generate images.

我用它来生成视频。

I use it to generate videos.

我用它来做研究。

I use it to do research.

我用它通过NotebookLM来阅读和理解我想掌握的大型论文或书籍。

I use it to ingest large papers or books that I wanna understand through NotebookLM.

Gemini模型感觉就像是现存最聪明的头脑。

And the Gemini model is the it feels like the smartest brain that exists.

如果2025年我只能使用一个AI模型，大部分时间我会用它来学习新事物，而Gemini正是我用来学习新知识的首选。

And if I could only use one AI model through all of 2025, a majority of the time I'm using it to learn new things, and Gemini is the thing that I wanna use to learn new things with.

这就是为什么我把Gemini选为我今年的模型。

And that's why I chose Gemini as the the model year for me.

但我们还有一些值得提及的优秀候选。

But we've got a few honorable mentions.

就那么几个。

Just a few.

就那么几个。

Just a few.

我的意思是，正如我所说，这次的候选者有很多。

I mean, like I said, there was a lot of nominees for this.

所以还有一些提名者。

So a few honorable mentions.

Grok，Josh，如果Chatty、Bugby和Claude是我用得最多的两个模型，那么Grok是我能说的最接近第二或第三的模型。

Grock, Josh, if Chatty, Bugby and Claude are two of the models that I use the most, Grock is the closest number two or three that I could ever say.

为什么？

Why?

因为它在我每天使用的主流社交媒体应用里，Josh。

Because it's in the main social media app that I use every day, Josh.

如果我读到什么不懂的东西，我就点那个Grok解释按钮，它会给我解释。

If there's something that I read or don't understand, I tap that explain by Grok button and it explains to me.

然后我和它展开对话。

And then I have a conversation with them.

不知不觉中，我一整天已经进行了大约十次不同的对话。

Before you know it, I've had like 10 different conversations throughout the day.

很棒的模型。

Awesome, awesome model.

我们没有这个奖项，但它可能是我今年使用最多的模型。

We don't have an award for this, but it might actually be my most used model of the year.

这可能是我互动最多的模型，因为对我来说，Grok 已经变成了 Google，每当我只是想快速搜索或学习一些东西时，Grok 就是我的首选。

That was the one that I probably interacted with the most because Grok to me has become like Google, where whenever I just wanna quickly search something or learn something, Grok is is the way.

然后我们有 DeepSeek。

And then we have DeepSeek.

我们今天早些时候已经提过它了，但这是一个非常出色的开源模型。

We shouted it out earlier today, but just an amazing open source model.

他们引领了开源领域的前沿。

They led the open source frontier.

正是他们的 2024 年底发布，导致了 NVIDIA 在今年年初的市场份额下跌了 5%。

It's what led to NVIDIA market share collapsing 5% at the start of this year with their 2024 end of year release.

简直不可思议。

Absolutely insane.

我还必须提一下谷歌的 AlphaFold 和 Genie，它们与大语言模型无关，但与科学和突破前沿模拟理论息息相关。

And I had to shout out AlphaFold and Genie from Google, which have got nothing to do with LLMs, but everything to do with science and breaking frontier simulation theory.

关于这个我们稍后再谈，但真是太棒了。

More on that later, but that was awesome.

接下来我们进入下一个类别，乔什，年度产品奖。

And for our next category, Josh, we have product of the year.

这是一个产品或功能，是普通人真正有意识地使用并持续使用的AI产品。

Is a product or a feature, an AI thing that normal people actually used on purpose and kept using.

这才是最重要的部分。

That's the most important part.

乔什，你选了谁？

Josh, who have you got?

这是一个重磅奖项。

This is a big one.

同样，让我把这顶王冠放大一点，因为年度产品奖绝非易事。

Again, let me increase the size of this crown here because product of the year is no easy feat.

人人都在努力打造最好的产品。

Everyone's trying to build the best product.

Josh，再放大一点。

Go a little larger on this, Josh.

这支持了极其庞大的用户数量。

Which is an outrageously large number of people that it's supporting.

这不仅是一个设计精良的应用，而且运行流畅，功能齐全，无论你在手机上还是电脑上都能正常使用。

So not only is it an excellently built app, it's just smooth, it works, it has all the functionality baked in, whether you're on your phone, whether you're on desktop.

它每周支持的用户数量如此之多，以至于不提名它简直是不可能的。

It is supporting so many users on a weekly basis that it feels impossible not to nominate it.

这是我进行语音聊天的地方。

It's where I go for voice chats.

这是我用来与所有记忆内容互动的地方。

It's where I go to interact with anything that has my memory.

它只是我键盘上的一个快捷键。

It's just a hotkey on my keyboard.

它也是我手机上的一个快捷方式。

It's a shortcut on my phone.

它无处不在。

It's everywhere.

对我来说，今年的产品是ChatGPT。

For me, it's the product of the year is ChatGPT.

好的。

Okay.

你准备好听我的了吗？

You ready for mine?

我们有什么？

What do we got?

我也选了OpenAI，但具体是OpenAI Pulse，这是一个个人AI助手，每天早上都会给我推送最新的新闻、金融洞察或我想了解的AI资讯。

I went with also OpenAI, but specifically OpenAI Pulse, which is the personal AI agent that pings me every morning with the latest and greatest news, financial insights, or AI stuff that I wanna get into.

Josh，最棒的是，它好像能读懂我的心思。

And the best part about this, Josh, is that it can, like, read my mind.

它会提取出我想知道的所有信息。

It pulls all the information that I wanna know.

它还帮我处理很多其他杂七杂八的事情。

It helps me with also a lot of other random stuff.

比如它知道我周五的锻炼计划，并提前给我一些准备和建议。

Like it knows what my workout routine is gonna be on Friday and gives me some prep and advice in foresight for that.

我觉得这是个很棒的产品，我迫不及待想看看它未来会变成什么样。

I just think it's an awesome product and I can't wait to see what it turns into.

这真是个赢家。

That's a winner.

然后，哦，这些提名也很不错。

And then, oh, these honorable mentions are good.

首先是Cloud Code，它很可能将全球的实际生产力提升了不止百分之零，这真的说明了很多问题，因为今年代码的质量和人们能够构建的代码量都极高。

First, the Cloud Code, again, probably increased the actual productive output of the world by a non 0%, which really says something because the quality of code this year and the amount of code that people have been able to build is so high.

但最让我钟爱、如果不能选ChatGPT的话我可能会选的是Nano Banana Pro。

But the one that is near and dear to my heart that I probably would have picked if I couldn't pick JatGPT is Nano Banana Pro.

它属于那种让人觉得不可思议的模型。

It's one of those models where it it doesn't make sense.

它感觉不应该这么快就如此出色，因为它对现实世界的理解如此精准，能够如此准确地呈现事物。

It feels like it shouldn't be this good this quickly because of how accurately it understands the real world, how accurately it's able to represent things.

我的意思是，上个月我们有一期节目，它还原了我的婚礼。

I mean, we had an episode last month where it covered my wedding.

它为我创造了一个完整的虚拟婚礼。

It created me an entire artificial wedding.

这太酷了。

It was so cool.

比如纳米香蕉，就是那种你一看就觉得像魔法的东西。

Like, nano banana, it's one of those things that it's like, you see it, and it feels like magic.

我真的完全不明白它是怎么运作的。

Like, I genuinely do not understand how it works.

所以

So

对于那些错过那期节目的人，乔什正式与肯德尔·詹纳结了婚。

For for those of you who missed that episode, Josh officially got married to none other than Kendall Jenner.

请在下方评论区送上你的祝贺。

Please send your congrats in the comments below.

谢谢。

Thank you.

她还不知道，但我对此非常高兴。

She doesn't know it yet, but I'm very happy about it.

我们知道这件事。

We know about it.

嗯。

Yeah.

嗯。

Yeah.

好的。

Okay.

我们继续下一个。

Moving on to the next one.

这个很有趣。

This one is fun.

这是今年最大的输家。

This is the biggest loser of the year.

好吧。

Okay.

等等。

Wait.

别作弊看最后一个，但这个是今年最大的输家。

Don't don't cheat and look at the bottom one, but this is the biggest loser of the year.

资源和成果之间差距最大，哎哟。

The biggest gap between resources and results, ouch.

这个人一次又一次地不断失败。

This is the person who just continually lost over and over and over again.

无论他们多么努力，都始终无法弄明白。

And no matter how hard they tried, they could not figure it out.

埃贾兹，你今年最大的失败者是谁？

Ejaz, who is your biggest loser of the year?

这个选择有不少选项，但我选了苹果。

There were quite a few options for this one, but I went with Apple.

哦，很熟悉，很熟悉？

Oh, familiar familiar?

听说过这家公司吗？

Heard of that company?

没听说过。

Haven't heard that.

老实说，那是他第一次打电话。

Honestly, was his first call.

听好了。

Listen.

我们今年谈论最少的公司，尽管这是一档以人工智能为主的播客，就是苹果。

The company that we spoke the least about this year, and granted, this is an AI focused podcast, was Apple.

他们没有被提及，因为他们根本没做什么。

They didn't get any mention because they didn't really do anything.

他们推迟了两年半才推出Siri的AI功能。

They've been delayed over two and a half years for Siri AI.

他们即将采用谷歌的AI模型来开发自己的AI产品。

Are about to engage Google, Google's AI model for their own AI product.

简直疯狂。

Absolutely crazy.

他们并不拥有自己的AI模型。

They don't own their own AI model.

太荒谬了。

Insane.

他们可能已经失去了80%的AI人才。

And they've lost probably 80% of their AI talent.

整个团队都被掏空了。

It's been absolutely gutted.

他们可能在新年期间更换CEO。

They're probably gonna have a CEO changeover in the new year.

苹果公司对此处理得实在太糟糕了，乔什，我不知道他们怎么才能翻身。

Apple's just handled this horrendously, and I don't know how they come back from this, Josh.

这很公平。

That's fair.

这很公平。

That's fair.

我希望他们能扭转局面。

I hope they turn it around.

这将在预测集里讨论。

That'll be for predictions episode.

但对我来说，最差的公司我必须选一个。

But for my biggest loser, I had to choose.

这甚至都不算让我有争议的选择。

This wasn't even it didn't even feel like a debate for me.

这是Meta。

This is Meta.

对我来说，这太明显了。

It's just, like, so obvious to me.

自从Facebook以来，Meta在每件事上都尝试过，但都失败了。

Meta has tried and failed at literally everything since Facebook.

比如，Meta这家公司、这个名字，之所以叫Meta，是因为他们转向了元宇宙。

Like, Meta, the company, the name, the reason it's called Meta is because they pivoted to the metaverse.

EG问：元宇宙在哪里？

And EG says, where is the metaverse?

根本不存在，它根本不存在。

There is no such It doesn't exist.

所以你们公司的名字，简直就是失败的象征。

So the name of your company is literally the failure.

其次，他们花了数十亿美元，聘请了大量顶尖人才来开发这些前沿模型。

Second to that, they hired billions, tens of billions of dollars of incredible talent to build these frontier models.

他们根本没发布那个大型产品。

They never even shipped the big one.

他们只发布了两个没人用的小产品，根本没等到他们宣布的第三个大型产品。

They only shipped the two little ones that no one used, and they never even got to the third big one that they had announced.

第三，他们试图让硬件发挥作用。

And third, they were trying to make hardware work.

他们根本无法让硬件正常运行。

They could not make hardware work.

演示简直是一场灾难。

The demos were an abomination.

其中一个我之前忘了提的最尴尬的场面是，他们展示Meta Ray-Ban智能眼镜时，把所有演示失败的原因归咎于Wi-Fi，结果什么都没能正常工作。

One of the most embarrassing things on stage that I forgot to mention earlier is when they they did the meta dis Ray Ban display glasses, and they blamed it on the Wi Fi, the fact that none of the demos worked, and nothing worked.

简直太糟糕了。

It was just terrible.

你根本没把它们戴在脸上，埃贾兹，尽管你很想这么做。

You're not wearing them on your face, ejaz, even though you wanted to.

我本来想试试的。

I wanted to try them.

它们真是太差了。

They just suck.

Meta 一直在失败。

Meta continues to fail.

我希望他们明年能想出办法。

I hope they can figure it out next year.

但那是我今年最大的败笔。

But that's that's my biggest loser of the year.

顺便说一下，其他值得一提的还有谁？是的。

Anyways, for honorable mentions, who else Yes.

今年他们惨败了。

Lost big this year.

对。

Right.

这是一个调侃的说法，但我们有迈克尔·伯里，他 famously 做空了英伟达，我认为还有英特尔以及其他几家公司，然后在他的仓位严重亏损后的下周关闭了他的基金。

This is a tongue in cheek one, but we had Michael Burry who famously shorted NVIDIA, and I think it was Intel as well, a few other companies, and then closed down his fund the following week when his position was down super badly.

他做出了一个非常合乎逻辑的赌注——有人可能会说，我们认为正处于人工智能泡沫中，即将破裂，但事实证明他错了，因为市场对人工智能产品的需求根本无法满足。

He made a bet of a very logical bet, some might say, that we are in an AI bubble and it is about to burst, but he was proven otherwise because there is just insatiable amount of demand for AI products.

所以你需要更多的GPU。

And so you need more GPUs.

大大的输了一把。

Big, big l.

另一个是乔什，这是你的建议，空头们。

And the other one, Josh, and this was your suggestion, the bears.

打倒悲观主义者。

Down with the pessimists.

打倒空头们。

Down with the bears.

如果你要押注对抗这股浪潮，最好系好安全带，因为你会被一路洗劫干净。

If you're gonna bet against this wave, you better buckle up because you're gonna get liquidated all the way up.

嗯。

Mhmm.

事情就是这样发生的。

And that's what happened.

我的意思是，我们在为这期节目做准备时，看了今年一些股票的涨幅，它们的涨幅达到了百分之六十、七十、八十。

I mean, we were looking in preparation for this episode, we were looking at the gains from some of these stocks this year, and they were like sixty, seventy, 80% gains Mhmm.

对于大盘股来说。

For large cap stocks.

我的意思是，谷歌涨了60%。

I mean, Google was 60%.

英特尔，我想涨了70%左右，或者类似离谱的数字。

Intel, I think, was like 70% or something outrageous like that.

我们有英伟达涨了24%。

We had NVIDIA was 24%.

AMD和CoreWeave都涨了60%到70%。

AMD, CoreWeave, both 60 to 70%.

如果你对这些公司持看跌态度，那你简直被洗得一干二净。

If you were bearish on any of these companies, you got absolutely washed.

你错失了股市历史上最棒的一年之一。

And you missed out on one of the best years in stock market trading history.

所以，没错。

So Yep.

在牛市时期做空真是太惨了。

Sucks to be a bear in the bullish time ever.

这就带我们来到

And that brings us

最后。

to the final.

好的。

Okay.

各位

Folks

年度AI公司，埃贾兹。

AI company of the year, Ejaz.

这里的人都是谁？

What are guys here?

这是最大的类别。

This is the biggest category.

有些人可能会说，这是唯一重要的类别。

The only category that matters, some might say.

年度AI公司。

AI company of the year.

我们要以不同的方式运作这个社会。

We're gonna do this society differently.

我们要宣布提名者，然后会有一个非常漫长的击鼓声，因为这非常重要。

We're gonna announce the nominees, and then we're gonna have a very elongated drum roll because it's very important.

顺便说一下，乔什，我们需要把那个王冠做得更大一点。

Josh, we need to make that crown a lot bigger, by the way.

我这就去处理。

I'm on it.

提名者。

Nominees.

我们有Anthropic，因其令人惊叹的编码AI。

We have Anthropic for their crazy coding AI.

我们有x AI，因其永不满足的社交Grok模型。

We have x AI for their insatiable social grok model.

我们有谷歌，因其全能的Gemini 3 Pro。

We have Google for the all rounder Gemini three Pro.

最后，我们有OpenAI，他们竟在最后从帽子里变出了GPT 5.2这只兔子。

And finally, we have none other but OpenAI who managed to pull a bunny out the hat at the end with GPT 5.2.

Josh，我们的最终赢家是谁？

Josh, who is our ultimate winner?

我得把王冠再缩小一点，因为赢家就藏在这里。

So I got to shrink the crown back because the winner's hidden right here.

我们的最终赢家，大家请鼓掌。

Our ultimate winner, everybody please drum roll.

2025年全球最佳人工智能公司。

Best AI company in the world this year in 2025.

恭喜。

Congratulations.

这绝非易事。

This is no easy feat.

这是在世界上最竞争激烈的行业中付出巨大努力的成果，获胜者是谷歌。

This is a culmination of so much hard work in the most competitive industry in the world, and the winner goes to Google.

恭喜。

Congratulations.

感觉理所当然。

It felt right.

冠军非它莫属。

It couldn't go to anyone else.

必须是谷歌。

It had to be Google.

掌声鼓励。

Round of applause.

好样的。

Bravo.

你们做到了。

You did it.

你们成功了。

You've succeeded.

你们在地球上最具竞争力的领域中打造了一个巨无霸，并且以优雅与从容，以及对物理学的深刻理解完成了这一切。

You have built a monster in the most competitive place on Earth and have done so with grace and elegance and a deep understanding of physics.

为此，我们向你们致以掌声。

And for that, we applaud you.

恭喜。

Congratulations.

年度 comeback 公司，年度公司。

The comeback company of the year, the company of the year.

谷歌在各个领域都占据了主导地位，他们实至名归。

Google has just dominated across the board, and they deserve this.

我对这个选择感觉非常好。

I feel really good about this choice.

没错。

Yep.

没错。

Yep.

当我思考谷歌为何能成功时，乔什，这让我想起了几周前发布的那部关于德米斯的纪录片。

When I think about why Google was able to pull this off, Josh, it brought me back to that Demis documentary that released a few weeks ago.

我觉得它叫《思考游戏》。

I think it's called the the thinking game.

从中最大的收获是，他把每一个问题都游戏化，然后不断用 AI 模型训练这个问题，一遍又一遍，直到它变得更好。

The biggest takeaway from that was he just gamifies every single problem and then slaps an AI model to train on that problem over and over and over again until it gets better.

他当初在早期的AlphaGo模型上就是这么做的。

It's what he did with the early alpha go model.

他也在许多其他领域这么做过。

It's what he did with plenty of other.

他用这种方法对待科学，现在他打算将其应用到核聚变反应堆上。

It's what he did with science, and now he's gonna apply it to fusion reactors.

他要把这种方法应用到材料科学上，还要应用到任何他能接触到的、可计算的问题上。

He's gonna apply it to material sciences, and he's gonna apply it to anything that he can get his hands on that is computable.

谷歌从很早以前就具备这种思维模式。

Google has had that mindset from so early on.

乔什，你还记得十年前他们宣布投资TPU的时候吗？

Josh, do you remember a decade ago when they announced their investments in TPUs?

你可能不记得了。

You might not.

我当然不记得，因为那时我只是一个青少年，做着与这些事无关的事情，但他却有远见。

I certainly didn't because I was just a teenager doing stuff that didn't concern any of this stuff, but he had the foresight.

这支团队里的人简直太厉害了，都是顶尖的天才。

Just amazingly led absolute crazy killers that work on this team.

乔什，我们刚才在镜头外聊到这个，我们意识到，所有AI领域的重要人物，某种程度上都源自谷歌。

Josh, you and I were talking about this off camera, but we realized that all the major players, all the major people in AI at some point stemmed from Google.

他们要么曾在谷歌工作，要么在谷歌实习过，或者在谷歌的研究岗位上工作过。

Either worked at Google, they either interned from Google or worked in research jobs with Google.

他们是所有AI事物的起点，也可能是终点。

They are the start and potentially the end of all things AI.

这是一家令人难以置信的公司。

Insanely impressive company.

这基本上就总结了我们2025年最佳榜单——最大的赢家和输家。

And that pretty much wraps up our best of 2025, the biggest winners and losers.

我想问问所有听众，你们觉得这份名单里最大的赢家和输家是谁？

What might ask for anyone listening is to share your biggest winners and losers from this list.

告诉我们你们觉得我们哪些人评对了，哪些人评错了，哪些人我们完全忽略了，以及我们为什么会忽略他们。

Let us know who you think we got right, who you think we got wrong, who we just totally forgot, and why we might have forgotten them.

下一期将是预测专场。

The next episode that is coming up is the predictions episode.

我们将分享对2026年会发生什么的预测。

This is going to be what we think is going to happen in 2026.

对我来说，这感觉更加有趣，因为今年我们取得了如此多的进展，如果这种趋势持续下去，2026年AI领域的进步将会非常惊人。

And that to me feels even more interesting because with the amount of progress we've gotten this year, assuming it continues, it's going to be a pretty wild year of progress in AI.

但本期节目就到这里了。

But that concludes our episode.

不再占用您更多时间了。

We won't hold you any longer.

感谢收看。

Thank you for watching.

请将本节目分享给一位朋友，无论它是会引发你的共鸣，还是满足你的期待，因为我们两者都想做到。

Please share this with a friend who you think this might either trigger or satisfy because we're here to do both.

别忘了在YouTube上订阅并开启通知铃铛。

And don't forget to subscribe on YouTube and put the notification bell on.

你们也可以在Spotify上观看我们。

You can watch us on Spotify as well.

这是我选择的平台。

It's where I choose to do so.

这就是2025年最大的赢家和输家。

And that has been the biggest winners and losers of 2025.

感谢大家观看，我们下一期再见。

Thank you guys for watching, and we will see you in the next one.

再见。

See you.