20VC：Mercor首席产品官谈来自前沿实验室的收入集中问题 | 为何大型企业不愿与前沿实验室合作 | 为何小型专业模型是未来——与奥斯瓦尔德·尼茨基对话

你还是得把每件事都再核对一遍。

You still have to, like, double check everything.

我常跟团队说，别把你的决策权，也就是你真正的本职工作，全交给模型。不然你会慢慢丧失这种能力，最后甚至会出现认知偏差。

And that's what I I I tell my team is that, like, don't delegate your decision making like your actual job to a model because you're gonna you're gonna lose that ability and then you're gonna get psychosis.

回顾你做过的那些实验，数据跟最终结果之间有对应关系吗？

When you look at the experiments that you've run, does the data correlate to the outcome?

我平时做投资时，经常这么想。

I often think in investing.

有时候我啥也不干，也不怎么努力，却能赚一大笔钱。

Sometimes I do no work and no diligence, and I make loads of money.

有时候我忙得不可开交，结果却做了些糟糕的投资，把赚来的钱全赔光了。

And sometimes I do lots, and I make terrible investments to lose all the money.

付出的努力和得到的回报之间，真的存在关联吗？

The inputs correlate to the outputs?

这得看情况。

Varies.

这要看情况，因为我们做了很多实验。

It varies because we run a lot of experiments.

有时候有关系，有时候没关联。

Sometimes they do, sometimes they don't.

我们更希望多做一些能带来好结果、真正推动业务发展的实验。

We wanna get more that actually show good results and move the business forward.

团队的核心任务就是找出该做哪些实验，并且讲清楚这些产品改动是如何正面影响业务的。

That's really the job of the team is to find the right experiments to run and make the narrative around these changes to our product have impacted the business in a positive way.

这现在是工作的重中之重。

That's a lot of the core job right now.

每周、每月的结果都不一样，但我们努力让趋势朝着正确的方向发展，大家也慢慢摸清了产品的规律和用户的行为，越做越准。

So, you know, it's week to week, month to month, we get different results, but we try to trend in the right direction over time, and people start to learn the dynamics of the product, learn the dynamics of the users better and better to improve over time.

当你想到产品和工程团队在做实验，尤其是做高质量实验的时候，你们现在的团队是怎么组织的？

When you think about product and engine running experiments that you mentioned there and running good experiments, how do you structure the teams today?

这些会议又是怎么进行的？

And what does that meeting look like?

我们有几个不同的团队专门负责实验工作。

We have a few different groups that do experimentation.

所以有两个主要的产品领域跟这个关系最密切，一个是我们的市场平台，用来匹配专家和工作任务；另一个是标注和评估平台，专家会登录这里进行标注，用于评估或训练数据集，我们的运营团队也会登录这里来运行这些项目，而客户则登录这里查看数据并进行评估。

So we have two major product areas where this is most relevant, our marketplace, which matches experts to jobs, and our annotation and eval platform, which is where experts log in to do annotation for eval or training datasets, where our operations team also logs in to run those projects and our customers will log in to see their data and run evals.

标注平台我们叫它 Studio，市场平台就叫 Marketplace。

So annotation platform, we call it Studio, Marketplace, called the Marketplace.

这两个团队在某种程度上是相对独立的，各自专注于让自己负责的产品变得更好。

These two groups, they're kind of like self contained in trying to do to make their individual product offering better.

在人工数据这块，我们主要有两种合作模式。

And we have two main modes of engagement within human data.

一种是纯人力外包，也就是我们直接派专家去客户那边，由他们来主导项目。

Talent only, which is when we just send experts to our customers, and they'll they'll run the project.

这就好比实验室需要医生或者律师之类的专业人士，客户会说，我们就用这个人，谢谢你们帮我们找到最合适的人选。

So this is like a a lab needs a doctor or a lawyer or whatever, and they're like, we're we're just gonna use them and, like, thanks for finding the best person for the job.

客户需要负责支付报酬并进行绩效管理，但项目执行还是由我们来负责。

You'll need to pay them, performance manage them, but, like, we'll do the we'll run the project.

还有一种是托管服务项目，我们直接把数据交给客户。

And then a managed service project where we give our customers data.

所以在纯人才模式下，我们只用市场平台。

So for the talent only model, we just use the marketplace.

在托管服务中，我们通过市场平台把人派到我们的标注平台上，然后我们负责运行项目，并把完整的数据集交给他们。

For the managed service, we use the marketplace to send people to our annotation platform, and then we'll give them we'll run the project and give them the whole dataset.

完全正确。

Totally.

这两个是独立的产品团队吗？

Are they two separate product teams?

对，是两个独立的产品团队。

They're two separate product teams.

产品团队规模有多大？

How big are the product teams?

每个产品线大概两到三个人，另外还有数据科学家和设计团队。

Around two to three per product area with also data scientists and a design team.

每个领域都有专门的数据科学家，以及一个在这几个领域间灵活调配的设计团队，规模都很小。

Data scientists dedicated to each and a design team that flexes between them of just a few.

对。

Yeah.

所以你们大概是四到五个人组成一个小组？

So you have like pods of like four, five?

差不多吧。

That's fair.

对。

Yeah.

你觉得产品经理和设计团队的比例以后会变化吗？

Will those ratios change over time, do you think, between PMs and design?

这个比例会一直保持不变吗？

Will that stay the same?

我觉得随着工程效率的提高，比如有了更好的编码助手，每个产品经理对应的工程师数量会减少，所以这个比例会发生变化。

I think that the ratio of PM to end will change over time to have fewer engineers per PM as engineering velocity increases with better coding agents.

我们会受限于对业务需求和用户需求的理解。

And we will be bottlenecked by understanding business needs, user needs.

这更多是产品经理的职责。

That's more of a PM job.

作为一家高速成长的公司，我们必须谨慎地让团队同步扩张，因为过去一年我们的总人数增长了十倍以上，我们不想让某一方增长得太快。

We need to be very careful as a hyper growth company to grow the teams in lockstep because as the headcounts increased, you know, like more than 10 x in the last year, we just wanna be careful not to grow one faster than the other.

不过，趋势会是产品经理与工程师的比例越来越高。

The trend will be higher PM to end ratio, though.

当我们谈到好的实验以及确保我们有一个严谨的流程时，这些会议具体是怎么安排的呢？

And so when we talk about the the good experiments and and making sure that we're running a really tight process, what does that look like in terms of the meetings?

你们每周都会开一次产品团队会议。

You have a weekly product team meeting.

目前，产品团队会议该以什么频率召开，又该怎么组织才最有效？

What is the right way to approach cadence of product team meetings and how to run them today?

我们把这些工作拆解开来，在每个产品领域里，我们每周都会开一次产品团队会，参与的不只是产品和工程，还有其他很多相关方。

We break it down into so within these product areas, we'll have a whole PA, like weekly, it's product and Ang and a lot of other stakeholders as well.

这一部分就像大家所知的那样，被划分成了若干个小团队。

And this one is just like everyone, you know, kind of like it's broken down into pods.

就像我刚才提到的，在这个产品领域内，可能会有三位产品经理。

So as I mentioned, within that product area, there might be like, let's say, three product managers.

他们各自负责产品的一部分，自然地形成了独立的小组来开展工作。

They'll all have like a pod of these parts of the product that naturally kind of like segment work into.

以我们的市场平台为例，它分为面向专家端和面向招聘经理端两个部分。

Our marketplace, for example, has a expert facing side and a hiring manager facing side.

这其实是两个完全独立的团队。

These are naturally two distinct pods.

里面还有一些其他的团队，比如负责专家体验的，确保每个人都能享受到优质的客户服务。

There are some other pods within here as well, like managing the expert experience, making sure that everyone has great customer support.

在Mercor工作从来不会有什么问题。

There's never any issues with with any you know, with working from Mercor.

每个团队都会各自进行自己的冲刺规划。

Each of these pods will do their own sprint planning.

他们每周会聚在一起开产品团队会议。

They'll come together in the weekly kind of like product area meeting.

我们尽量让会议高效，同时保持各小组之间的高度透明，因为大家的路线图必须对齐。

And we try to keep it efficient but maintain like a lot of visibility between the pods because they all need to have their roadmaps well aligned.

是的。

Yeah.

但我们需要每周开会来确保责任落实。

But we need to have weekly meetings to maintain accountability.

安排在周五下午，别太早，免得大家周末提前走人。

Do them on Friday, bit later in the day, make sure no one's leaving early on the weekend.

在你们的产品会议里，有哪些事情是你们不该做的，但做了反而能提升效果？

What do you not do in your product meetings that you should do to make them better?

这

That's a

是个好问题。

good question.

这得看具体的产品领域。

It varies by product area.

所以像市场平台和标注平台面临的挑战就不太一样。

So the challenges in like the marketplace versus like the annotation platform are a bit different.

随着我们快速扩张，主要的挑战在于如何与其他团队保持恰当程度的沟通和反馈。

The main challenge as we grow quickly is having the right amount of communication and feedback from other teams.

比如我们的市场团队和工作室团队就需要互相同步信息。

So like our marketplace and our studio team need to get information from each other.

对吧？

Right?

有时候一个地方出了问题，会影响到另一个地方。

There's cases where like something's wrong in one and it's effect it's it's popping up in the other.

某个产品出了故障，结果反而影响了用户在另一个产品上的体验。

Something's wrong with one product and it's affecting like the expert experience when they're like on the other one somehow.

因为团队人数增长得太快，沟通渠道一下子就变得混乱不堪。

That communication just like the the because the headcount and the team's grown so quickly, like the communication channels, like, just explode very quickly.

所以我们需要加强跨产品领域的协作。

So we need to do more kind of like cross product area collaboration.

随着团队规模扩大，保持高效真的很难，因为人员变动频繁，而且人越来越多。

Keeping it efficient is just really hard as the team grows because we're you know, the nodes just keep moving around and there's more of them.

你们在扩大市场供应方面有什么秘诀吗？效率这么高。

What has been the secret to scaling supply on the marketplace side so efficiently?

你们是怎么做到这么出色的？

How have you guys done that so well?

我觉得主要可以归结为三点。

I'd probably put it down to three things.

第一点是给专家提供很好的体验。

The first one is a great expert experience.

专家能按时拿到报酬。

Experts get paid on time.

而且报酬透明、合理。

They get paid well transparently.

所有参与专家体验相关工作的人都非常关心专家是否面临挑战，以及工作是否有尊严、报酬是否优厚和公平。

Everybody involved in what the expert experience is cares deeply about not whether they're having any challenges and whether or not the work is dignified and well paid and fairly paid.

这也是推荐计划做得好的前提，因为没人愿意把朋友或同事推荐到那种待遇很差的工作上去。

And that is a requirement for a great referrals program because nobody's going to refer their friends to, know, their colleagues to some kind of job that's like, that sucks.

所以，大家共同重视专家体验，直接推动了推荐计划的成功。

So everybody caring about expert experience drives a great referral program.

另外，一支强大的招聘团队能在全球范围内找到拥有特定技能的人才，这帮我们在面对某项技能需求激增时，及时填补了空缺。

Additionally, a great sourcing team that's able to find people in every corner of the world with very specific skills helps us fill the gaps when, you know, we have spiky demand for a specific skill set.

那些目光短浅的人，就是只想拿最高薪水。

A people who shortsighted is just being wanting to be paid the most.

我听说 Mercor 给的薪水最高。

I've heard that Mercor pays the most.

我倒不觉得这是目光短浅，因为我们也想留住顶尖专家。

I wouldn't say it's like, you know, shortsightedness because we want to retain the top experts as well.

如果你在一个项目上拿了很多钱，我知道这行里有很多竞争对手会搞一些疯狂的短期奖金或者冲刺奖励之类的。

If you get paid a lot on like one project and it's like I know there are a lot of other competitors in the space who will do some crazy like bonus payouts and stuff and for short term sprints.

光靠钱留不住人，真正让人愿意回来的，是良好的工作体验、充足的工作量、对未来任务的清晰预期，还有那种感觉自己技能在提升、能在几个工作里自由选择、做的是有趣的事，以及项目管理者沟通顺畅到位。

That doesn't get you to come back as much as a great experience with a lot of work, visibility into like what future work is coming up, the feeling of like I'm growing my skill set, I have the ability to pick between a few different jobs, I'm doing interesting work, I have great communications, right, from the people running the project.

找线上工作最难的就是，刚注册好，结果迎面甩过来一份上百页的操作指南，太劝退了。

It's really hard to sign up for online work and then you just like have get hit with this like 100 page instruction document.

这活儿确实挺陌生的。

It's a very foreign kind of job.

这也是整体体验的一部分。

That's part of the experience as well.

知道自己在擅长的领域能长期拿到高薪，这才是让人保持热情的关键。

Knowing that you're gonna get paid highly for a long time for something that you can do for a long time is what keeps people interested.

你有没有发现你们的利润率随着时间推移有提升？

Have you seen your margin improve over time?

还是说，考虑到业务复杂度，利润率其实相对固定？

Or is it one where actually margins relatively fixed given the complexity?

所以，利润率在这个行业里是个挺有意思的话题。

So so margins are an interesting thing in this business.

作为产品团队，我们一直在思考如何为客户提供最大的价值。

We try to think about as a product team, how do we deliver the best value for our customers?

这和项目的定价是两码事。

That is independent of how do we price the project.

比如，我们可以通过自动质检和合成数据优化来提升交付质量，也可以通过调整项目的人员配置来改变服务成本。

So there are cases where you could have automatic quality control and synthetic data improvements to make the delivery better, situations where you could, you know, think about the staffing on the project to change the the cost of of the service.

总之，作为产品团队，我们要确保既能给客户最好的价值，也能让专家们的体验达到最佳。

All of that, like, as as a product team, we wanna make sure that we could deliver the best value to our customers, and we can have the best experience for our experts.

利润率是事后根据成本核算来确定的。

Margins are decided after the fact based on, you know, consideration of of costs.

现在这些项目的成本主要来自支付专家费用和用于合成数据及自动质检等大模型的算力支出。

And, you know, now for a lot of these projects, the costs are driven from paying experts and LLM spend on things like synthetic data and automatic quality control.

别人总把收入挂在嘴边，像街头小贩一样吆喝着往人群里扔花生米，这会让你觉得烦吗？

Does the it's not revenue, it's not revenue shouting from the crowds throwing peanuts, Does that annoy you?

还有，有没有什么你没说出口、但觉得大家其实没真正理解的东西？

And is there anything there that hasn't been said that you think people are just, like, not getting?

这并不会让我感到烦恼。

It doesn't annoy me.

不会。

No.

毕竟每周结束时，我们的银行账户里都会多出数百万美元。

Because, you know, we end every week with millions more in the bank.

对吧？

Right?

所以挺有意思的，我之前在其他公司待过，见过各种各样的财务操作和会计手段，人们能搞出好多不同的指标，但我们每周结束时银行账户里的钱都多出很多。

So it's funny how you can have I've been at other companies where I've seen interesting financial engineering and accounting and people can have all these different metrics, but we end every week with so much more money in the bank.

业务非常健康，我们花钱的速度根本跟不上。

Like, the business is is very healthy and we can't we can't spend money fast enough.

所以别人怎么称呼它随他们便，但现金流简直太惊人了。

So what people wanna call it is, you know, up to them, but, like, the the cash flow is insane.

你的收入集中度这么高，这会有影响吗？

Does it matter that you have such high revenue concentration?

你知道的，前沿模型提供商，目前为止你们最大的客户，有人甚至认为这是最大的客户，这种集中度很高。

You know, the frontier model providers, your biggest customers by far, some would say, that's a lot of concentration.

你们是怎么看待这个问题的？

How do you think about that?

我可以从它对产品团队的影响这个角度来回答。

I can answer this from kind of like how it affects the product team.

对。

Yeah.

我们非常希望往下沉市场拓展，目前最大的挑战就在这里，目标是让每家企业都能高效地运行用于评估和训练的人机数据项目。

We would love to move like, biggest challenge is moving down market so that every single enterprise can efficiently run human data projects for eval and training.

这肯定能帮我们分散收入来源，毕竟企业的数量远比实验室多得多。

And that'll diversify our revenue for sure because there's many more enterprises than there are labs.

而且，打造这样的产品难度更大。

And that's a harder product to build.

这也是我们产品和公司的发展方向：实现自助式服务，让项目运行更高效，配备类似AI项目经理的功能，从而让中小客户更容易上手开展这项工作。

And that's the direction that we are taking our products, taking the company is to be able to self serve, run these projects very efficiently, have like AI project managers so that it's a lot easier to do this work for smaller customers.

因为为实验室运行一个人工数据项目真的非常难。

Because running a human data project for a lab is incredibly hard.

这是一种需要运营团队投入大量人力的白手套式服务。

It's a white glove service that requires a lot of people on the operations team.

随着我们通过更好的产品和流程提高效率，我们可以承接更多样化、规模更小的项目，服务更多客户。

As we make that more efficient with better products, better processes, we can do smaller projects that are more heterogeneous for more customers.

这正是我们一直努力的方向，这也降低了客户集中度。未来随着每家企业都开始为自己的专有场景进行人工数据处理，我们还会继续朝这个方向前进。

It's the direction we have been heading, which has reduced concentration, and it's the direction that we'll continue to head as every enterprise begins to have human data work for their proprietary use cases.

这到底难在哪里？

What's so hard about it?

把它做得特别简单？

Making it really simple?

把它解释清楚？

Explaining it?

实现普及化有什么难点？

What is the challenge with democratizing?

做人类数据项目本身就很难。

Running a human data project is just hard.

需要从客户、客户的最终用户那里收集大量信息，再传递给专家，而且所有边缘情况都很重要。

There is so much information that needs to be transmitted from the customers, the end users of our customers to experts, and all the edge cases matter.

对吧？

Right?

所以人们会试着写一份指南，比如说明如何制作一个数据点。

So people will try to write a guideline that says, like, here's how you here's how you make a data point.

但专家们总会遇到一些边缘情况，这些情况会被提出来，而且怎么处理这些边缘情况非常关键。

But the experts will have, like, some edge case that gets bubbled up and, like, what you do on that edge case matters a lot.

所以，做一个人类数据项目的过程，基本上就是不断把这些边缘情况找出来。这需要客户、客户的客户、领域内的其他专家，以及负责标注的专家之间保持极其快速的对齐。

So the process of making a human data project is basically, like, continually surfacing these edge cases, which requires insanely fast alignment between customers, maybe their customers, maybe other experts in the field, and the experts who are doing the annotation.

同时，运营团队也需要高度警惕，确保每一个数据点都完美无缺，符合客户的所有指南，项目按时推进，并且及时清除所有瓶颈。

And it also requires a huge amount of paranoia from the operations team to make sure that every data point is perfect, it fits whatever guidelines the customers have, and the projects are running on time, all the bottlenecks are removed.

这确实是一个运营强度很高的过程，因为它必然要处理那些边缘情况、以前没见过的情况，以及超出模型能力范围的问题。

It's just an operationally intense process because it necessarily deals with edge cases and things that haven't seen before and are outside of model capabilities.

数据类型也变化得非常快。

The data types also change very frequently.

所以我们的工作流程已经从监督微调，转向偏好排序，再到基于评分标准的标注，现在又变成了跨多种模态的强化学习环境。

So we've moved from supervised fine tuning to preference ranking to rubric based annotation to now RL environments across a whole bunch of different modalities.

每个项目内部都很复杂，项目之间的协作也很复杂。

There's a lot of complexity within each project and then between projects.

所以我觉得归根结底就是这两点：需要极度谨慎，以及需要非常清晰高效的沟通，正是这两点让工作变得很有挑战性。

So I would boil it down to those two things of, like, the need for paranoia and the need for very crisp communication that make it challenging.

你觉得现在需求不算特别大，但明年会爆火的数据类型是什么？

What data type is not hugely in demand today that you think will be hugely in demand next year?

我们这边增长最快的是环境数据。

The data type that's growing the fastest for us is environments.

你可能在推特上看过不少关于强化学习环境的内容。

You might have seen a lot about these RL environments on Twitter.

这词儿现在有点被炒得太热了。

It's kind of like a hype term.

每家公司对它的定义都不太一样。

Every company kind of like has a different definition for it.

但我们肯定是这个领域的领头羊，把它看作是你希望你的智能体去使用的那些应用的模拟环境。

But we are certainly the leader in the category and view it as basically these simulations of apps that you might want your agent to use.

同时，它也是一个丰富的初始状态，我们称之为“世界”，基本上代表了你在电脑上可能拥有的所有数据，比如你的笔记本电脑里的内容。

And also as a rich start state, we call like the world that is basically representative of all the data you might have on your machine, like your laptop.

然后我们还有具体的任务，用来训练智能体如何使用这些工具来完成一些有用的事情。

And then we have tasks that train agents how to use those tools to accomplish something that's useful.

标注过程挺复杂的，因为智能体得跟这个模拟世界进行互动。

It's a bit of a it's it's a bit of a complicated annotation process because the agent has to like interact with this simulated world.

我们需要设定初始状态，这可能会涉及成百上千个文件。

We have to make that start state which can be hundreds of files, thousands of files.

现在的变化在于，模型（也就是智能体）进行评估和训练所使用的数据，跟它们在实际部署时看到的数据非常接近。

And the shift here is that the data that the models are now the agents are being evaled and trained on looks a lot closer to what they see in deployment.

对吧？

Right?

所以如果你想学着用 Salesforce 这样的系统，你的评估和训练环境就必须高度还原，完全像真的一样。

So if you want to learn how to use something like Salesforce, you need a pretty high fidelity mock that acts exactly like Salesforce in your eval and training.

但要搭建这样的环境，其实挺复杂的。

And it's complicated to get this set up.

就像几年前，偏好排序的设置也特别麻烦。

Just like years ago, preference ranking was really hard to get set up.

当初 InstructGPT 刚出来的时候，监督微调的配置也是一团乱。

SFT was really hard to get set up when when InstructGPT first came out.

现在这正是最前沿的挑战。

So this is the frontier right now.

实验室们正在摸索，Neo 实验室也在想办法解决。

Labs are figuring out, neo labs are figuring it out.

总有一天，这套流程会变得特别顺畅，连企业都能轻松上手。

Eventually, it'll get so smooth that enterprises can do it too.

实验室在数据获取上对成本敏感吗？

Are labs price sensitive on data acquisition?

说到数据获取

By data acquisition

跟你们合作做项目的时候，他们对价格敏感吗？

When go on a project with you, are they price sensitive?

他们会讨价还价，比如‘埃德温在Surge那边给了我10%的折扣，你能不能也这样？’

Are they haggling going, oh, well, Edwin at Surge gave me a 10% discount, Can I have that?

还是说他们直接说，‘把数据给我就行’？

Or are they like, just give me the fucking data?

当然，谈判和采购团队总是希望能拿到更优惠的价格。

Well, there's always the aspect of negotiation and the procurement team trying to get a better deal.

但我们选对了一个好行业，我们的工作能直接影响客户的业务成果。

But we've chosen a great business where our work directly affects the business outcomes of our customers.

对吧？

Right?

所以我们的处境很好。比如你在实验室里构建评估集，其实就是在评估客户真正想做的事。

So we we have a great setup where if you're making an eval set, like in your lab, you're evaluating something that your customers want to do.

只要你能做得更出色，收入自然就会增加。

If you could just do it better, you would make more revenue.

如果他们买的是训练集，那其实就是在优化那个评估集，毕竟那个评估集代表了他们客户真正的需求。

If they're buying a training set, they're now hill climbing that eval set that they've said, you know, that represents what their customers wanna do.

所以只要花在数据上的钱少于他们能赚到的收入，他们就很乐意加大投入。

So as long as, like, the amount of money they're spending on data is less than the revenue that they're gonna get, they're happy to crank the lever.

大家都想拼命加大投入，因为在 Mercor 上的每一分钱支出，都能直接转化为客户的更高收入。

Like, people wanna crank it harder and harder because spend on Mercor directly translates to more revenue for our customers.

你觉得未来会出现专门提供单一数据的服务商吗？

Do you think we'll have a unbundled data provider world?

你知道，我是一名风险投资人。

You know, I I'm a venture investor.

我看到太多人都在说，我们要像Mercor那样，不过做的是国内机器人领域。

I see so many people that are like, oh, we're like we're like Mercor, but for, like, domestic robotics.

然后你就只能回一句，好吧。

And you're like, okay.

不错。

Cool.

很好。

Good.

好的。

Okay.

明白了。

Get it.

但你觉得我们会看到那种细分领域里有成千上万家专业数据提供商的世界吗？

But do you think we will see this kind of specialized data provider world where niches have thousands of players?

某种程度上，我们已经身处这样的世界了。

To an extent, we're already in this world.

不过对小玩家来说，情况并不总是那么顺利。

It's not that successful though for the small players always.

所以我怎么形容呢，我们正面临一个看似家庭作坊式的局面，创始人都在自己搞标注。

So how I would describe it is we're facing what looks like a cottage industry of founders doing annotation themselves.

对吧？

Right?

所以你看，随着模型越来越强，标注工作的门槛也越来越高，这就出现了一些初创公司，它们的创始人其实就是在直接生产数据。

So you have all of these small startups where as the skill bar for annotation gets higher and higher as models get better, you have startups where the founders are actually just making the data.

而且实验室特别喜欢这种模式，因为这种数据的价格被严重低估了。

And labs love this because it's just like totally mispriced.

你懂的。

Know?

他们拿到一笔融资。

They get like someone raises a bunch of money.

手里握着大把现金，就跑去那些大模型公司，说一定要拿下你们的业务。

They have loads of cash to blow, they're they go to these labs and they're like, I I need to you're like, I need to get your business.

拜托，让我来给你们干活吧。

Like, please let me work for you.

而且他们都很聪明。

And then they're they're smart people.

他们就是创始人。

They're they're founders.

他们以前都是些很厉害的技术员工，但现在都是他们自己在操办项目。

They're formerly, like, great technical employees, But they're running the projects themselves.

标注工作也是他们亲力亲为。

They're doing the annotation themselves.

这简直就是风投补贴的工作，实验室们特别喜欢。

This is just like VC subsidized work that labs love.

问题在于，一旦数据量超过几个点，或者超出了创始人和全职员工能处理的范围，就很难扩展了。

The problem is scaling it beyond a few data points or what what one founder or full time employees can do.

我们现在的处境就是，得跟那种由创始人亲自带队的标注团队竞争。有些团队甚至把它当成现金流生意在做，利润直接揣自己兜里了。

And this is the position that we're in, is we're having to compete against basically founder led annotation where some of them are even running it as cash flow businesses and they're just like taking the profits home themselves.

但这模式根本没法规模化。

It doesn't scale though.

我们的客户，也就是那些供应商，心里都清楚这一点：当你需要吞吐量翻十倍、项目量翻十倍的时候，这种模式是撑不住的。

And vendors are our customers know this, that it won't scale when you want a 10 x the throughput, 10 x the amount of projects.

但这确实反映了这个领域的发展方向，我们需要更高技能的专家。

But it is indicative of the direction the field's heading in, in that we need higher skilled experts.

我们需要世界上最优秀的人来做数据清洗工作。

We need the best people in the world to be doing the sanitation.

别笑。

Don't laugh.

我有点自负，所以我想让自己感觉像个特别的存在。

I'm I'm I have a bit of an ego, and so I like to feel like a special snowflake.

我的意思是，如果Meta、OpenAI或者任何一家大公司从好几个人手里买数据，我会觉得你有点花心，像是在背叛我。

And what I mean by that is, I would be like, oh, when Meta or OpenAI or you name your large company is buying data from multiple people, it feels like you're being promiscuous and cheating on me.

你介意吗？

Do you mind?

你会监控预算吗？比如在你这里花的预算比例，跟在其他供应商那里花的比例对比一下？

And do you monitor budget and percent of budget that gets spent with you versus another provider?

当然，我们做了很多竞品分析。

Of course, we do a lot of competitive intelligence.

客户们挺喜欢我们的，所以他们会经常跟我们分享一些信息。

Our customers like us, so, you know, we they'll often share information with us.

不过大家都希望模型能变得更好。

But everybody just wants models to get better.

对吧？

Right?

所以我们很乐意接受这种竞争压力，它能告诉我们该往哪个方向走。

So we're happy to have this kind of competitive pressure that tells us, like, where to go.

如果别人能做得比我们更好，我们非常乐意听听，然后做得比他们还要好。

If someone else is able to do something better than us, we'd love to hear about it and then do it better than them.

对吧？

Right?

供应商之间的竞争其实挺健康的。

It's it's healthy to have vendor bake offs.

这能促使我们把服务做得更好。

It pushes us to make our services better.

我们一直密切关注，因为我们要给客户提供更好的服务。

We do stay on top of it because we want to deliver better services to our customers.

我们想知道谁做得比我们好，然后想办法超越他们。

We wanna know who's doing better than us and then we wanna surpass them.

只要Mercor能领先，这种竞争完全是好事。

So it's a totally healthy thing to happen as long as Mercor is winning.

我们之前提到过前沿领域。

We said frontier earlier.

对。

Yeah.

我是Ligure的投资者，大家都说不是。

I'm an investor in Ligure, and everyone's like, no.

你真正的对手其实是Anthropic。

Your real competition is is actually Anthropic.

如果Anthropic去抢Cooley和Goodwin的法律业务，那这个世界就出问题了，他们该去解决癌症和气候变化才对。

And I'm like, if Anthropic go after legal and winning Cooley and Goodwin, something's gone very wrong with the world because they should be solving cancer and climate change.

我这么说有多大的道理？

To what extent am I right?

我该怎么平衡 Anthropic 对 Ligure 和 Figma 的威胁，以及他们同时在解决人类面临的前沿问题这件事？

And how do I balance between Anthropic are coming for Ligure and Figma, and Anthropic is also working on the frontier problems that humanity faces today?

这是个好问题。

It's a good question.

你可以参考其他大科技公司，比如谷歌和微软，它们也做过不少尝试，想解决气候变暖和癌症这类问题。

I would look to precedence for from other big tech companies who've had a lot of different efforts like Google, Microsoft, you know, coincidentally also try to, solve climate change and cancer.

但那些都不是它们的核心业务，它们的手伸得太广了。

But it's not their main business, and they have their hands in, like, a lot of different areas.

可竞争者还是不断冒出来。

But competitors still emerge.

你记得谷歌+吗？

So you've seen like, you know, you remember Google plus, right?

对。

Yeah.

那个项目最后没做成，对吧？

That didn't go anywhere, right?

它刚出来的时候，估计把不少人给吓到了。

It probably maybe it freaked some people out when it happened.

你应该还记得 Threads 吧。

You probably remember threads.

我不太清楚 Threads 现在的情况。

I don't know the current state of threads.

据他们的营销团队说，有4亿用户。

Apparently 400,000,000 users according to their marketing team.

这挺有意思的。

That's very interesting.

评论一

Comment I

关于这一点说得太多了，因为

too much on on that because I

哦，真有意思。

Oh, fascinating.

我倒是很想看看互动数据怎么样。

I'd love to see the engagement.

我是真的不了解这方面的情况。

I genuinely don't know anything about this.

不过确实是这样。

But yeah.

所以我觉得，如果看先例的话，大公司通常会尝试新的布局，寻求多元化，但往往输给了那些在市场里极度专注的公司。

So it's in in I think if you look to precedence here, large companies often try to, you know, make new bets, diversify, but they lose to companies that have intense focus on their market.

所以我们看看结果如何吧。

So we'll see how it plays out.

不过我挺好奇的，能不能从谷歌和微软的历史里学到点什么？它们都有很多业务线和各种尝试，但核心业务才是推动它们所有收入的关键。

But I would I would wonder if there's anything to learn from, you know, history with Google and Microsoft having having many business units, many efforts, but a core business that has driven all of their revenue.

你知道的，我很喜欢布兰登。

You know, I love Brandon.

我记得黑客事件发生时给他发过消息。

I remember texting him when there was the hack.

出这种事特别难办，你都不知道该说什么，但就是想让他知道我在。

It's tough when there's a hack because you're like, don't know what to say, but like, I'm here for you.

给你点个赞。

Thumbs up.

那时候我感觉自己真像个投资人，就想着：你别怕，我在这儿。

And I felt like such a VC because you're like, I'm here for you.

加油。

Good luck.

真是糟透了。

Fuck all hell that is.

我想问你，这件事是怎么改变你对产品的想法和做法的？

My question to you, how did that change your mindset and approach to product?

这种经历真的特别难熬。

It's a really hard thing to go through.

我记得你当时承受着巨大的压力和焦虑。

I remember you were under intense pressure and stress.

我真的很难过，因为经历这种事实在太糟糕了。

I seriously, I am sorry for that because it's horrible to go through.

这件事是怎么改变你对产品的看法的？

How did it change your product mindset?

我不是安全方面的专家，但我们请了很多这方面的高手，我也认真听了他们的建议。

You know, I'm not an expert in security, but we hired a lot of experts in security and I listened to them.

这其实就是最主要的变化。

And that's that's the main change.

就是加大投入，多向我们请来的专家学习。

It's just larger investment and learning from the experts that we've brought in house.

我们是不是正进入网络安全的黄金时代？

Are we entering a golden age for cyber?

我的意思是，现在AI生成的代码越来越多，而这些代码很多时候都有漏洞。

What And I mean by that is we're seeing a huge amount of AI generated code, which in a lot of cases has holes.

我们看到各种各样的工具，不管叫什么名字，都能产生海量的输出。

We're seeing a lovable and a replet and a, you name it, produce a huge amount of output.

威胁会不会比我们预想的要显著增加很多？

Are the threats going to increase much more significantly than we're anticipating?

大概率会是的。

Most likely, yes.

我们最明显的感受是，这是一种很有意思的数据类型，因为它具有竞争性。在网络攻防领域，会出现类似AlphaGo那种局面，收益和表现没有上限，而且整个领域一直在快速变化。

Where we see it the most is it's an interesting data type because it's competitive, and you can have these alpha go type situations for for cyber offense and defense where you can have uncapped rewards and performance and the the field's constantly moving.

所以我们特别喜欢这类东西，因为从数据的角度来看，这就像是在玩游戏一样。

So we love this kind of stuff because it's like a game from like a data perspective.

而且我们看到，通过数据来增强网络防御能力的需求正在迅速增长，涉及的数据类型也非常有趣。

And we see very rapidly increasing demand for cyber defensive capabilities via data and very interesting data types.

这是一个例子，说明只要有足够的数据……

And this is an example of something where sufficient

等等，你能帮我理解一下吗？

Wait, can you help me understand?

在需求爆发之前，大家对安全领域有哪些数据类型的需求是以前没有的？

What data types do people want around security that they maybe didn't want before there was this explosion in demand?

得小心点，别把客户工作的细节透露太多。

Have to be careful not to reveal too much about customer work.

这个领域的增长非常快。

The category is growing very quickly.

而且很多安全工作的性质就是对抗性的，对吧？

And the nature of a lot of security work is that it's adversarial, right?

所以这不是那种只要达到某个标准就能完事的工作，比如更新一下客户关系管理系统，然后就没问题了。

So it's not this sufficiency style work like update a CRM and and then you're good.

进攻方和防守方之间一直在进行着猫鼠游戏。就像我们之前提到的，虽然90%的企业工作流程已经可以自动化完成，但在安全领域，永远不可能达到这个比例，因为目标一直在变。

There's a constant cat and mouse game between like the offensive capabilities and the defensive capabilities, which, you know, to our point earlier about like the 90% of enterprise workflows that can already be completed, like There's never going to be that 90% for security because the goalposts are always going to move.

因此，网络安全作为一个整体类别正在增长，而且这类数据的性质更加复杂，没有上限，不断演变，并且具有对抗性，因为目标一直在移动。

So most cyber as a category is growing and the nature of the data types is much more kind of like uncapped, evolving, adversarial in terms of the where the goalposts are.

你能帮我解释一下吗？

Can you help me out here?

你祖上是爱沙尼亚人吧。

I you're Estonian by kind of heritage.

我跟欧洲的创业者说，旧金山根本不是创业的好地方。

I say to European founders, SF is the worst place to start a company.

根本招不到合适的人才。

It is impossible to acquire talent.

也根本负担不起人力成本。

It is impossible to afford it.

而且很难留住人才。

And then it's impossible to retain it.

旧金山的人才争夺战真的像看起来那么残酷吗？

Is the talent war in SF as brutal as it seems?

是的。

Yeah.

确实非常残酷。

It's pretty brutal.

招人真的很难。

It is very difficult to hire.

留住人也难。

It is difficult to retain.

确实不容易。

It's difficult.

我觉得比之前更难了，但要是公司像火箭一样飞速增长，那就另当别论了，对吧？

I think it's harder than before, but it's easy when you're on a rocket ship, right?

其实一直都不容易。

It's always easy.

公司像火箭一样飞速发展的时候，反而很难做出正确的招聘决定，不知道到底该选谁。

When you're on a rocket ship, to get someone, it's hard to make the right decisions about who you want to hire.

如果你招错了人，事后回想，你当时没看出来、但希望自己能看出的地方是什么？

When you've made a bad hire, what did you not see that you wish you'd seen?

在面试过程中，真的很难判断一个人的主动性和主人翁意识。

It's really hard to assess agency and ownership in the interview process.

我特别有才华。

I am super freaking talented.

我非常有能力。

I'm super talented.

我有点混蛋。

I'm a bit of an asshole.

我不是那种彻底的混蛋，但确实有点讨人厌。

I'm not like a total asshole, but I'm a bit of a douche.

你能接受这一点吗？

Are you okay with that?

如果你特别有才华，那是可以的。

If you're super talented, yeah.

我们公司的文化强调高自主性、高绩效和高责任感。

The company culture here is of high agency, high performance, high ownership.

人的性格是可以改变的。

Personalities can change.

你可以学习如何更好地与人相处。

You can learn how to work with people better.

我们看重成长，也看重那些真正在乎工作的人。

We care about growth and we care about we want to hire people who give a shit.

比起搞搞团建、确保大家关系融洽、一起参加Happy Hour这些表面功夫，培养一个人的内在驱动力要难得多。

That's a lot harder to coach into someone than smoothing it out with your colleagues, getting some you know, making sure that we have happy hours, people all get along.

像搞搞人际关系这种，其实很容易做到。

Like that that's easy to work.

你知道的，偶尔有几个刺头也是可以的。

You know, like, you got you can have a couple assholes.

大家出去喝几次酒，慢慢就把关系理顺了。

You they get drinks together a few times and then you smooth it out.

真的很难让人真正上心。

It's really hard to make someone give a shit.

对。

Yeah.

而且，你要是招了一堆刺头，他们就能凑一块儿玩了。

Also, you hire multiple assholes, they can just hang out together.

那也没事。

It's fine.

这就成了个固定小团体了。

That's that's a group of fact.

我们可不会招那么多刺头。

We don't we don't hire a lot of assholes.

什么？

What?

我的……不是。

My my No.

不是。

No.

我也……比如，像搞搞团建聚会之类的？

I I I can be also, like, like, happy hours?

真的吗？

Like, really?

我们最近刚跟标注团队搞了一次很棒的团建。

We had a great off-site just recently, actually, with our annotation team.

我们去了加拿大的托菲诺。

We went to Tofino in in Canada.

那地方在加拿大西海岸。

It's on the West Coast Of Canada.

那是唯一能冲浪的地方。

It's the only place you can surf.

大家都去上了冲浪课。

And everyone did surfing lessons.

我们去了个漂浮桑拿，玩得特别开心。

We went to a floating sauna, and it was a great time.

我觉得这对团队特别有帮助，钱花得也值，大家都挺喜欢的。

I thought it was actually great for the team and is a great use of of money, and everybody loved it.

我觉得做这些户外活动挺好的，让大家动起来。

And I think that doing these, like, outdoor activities where people are being active is is good.

准备好来一轮快问快答了吗，兄弟？

Are you ready for a quick fire round, dude?

没问题，来吧。

Sure, yeah.

过去这一年里，你改变看法最多的事情是什么？

What have you changed your mind on most in the last twelve months?

说实话，我觉得可能是大环境，或者是咱们自己的市场环境。因为去年刚开始做的时候，这些交付流程太复杂了。

Honestly, I think it's probably the environment, our own environment market, because when we were starting it off last year, it was so complicated to do these deliveries.

当时实现起来特别难，我甚至以为这事儿肯定成不了。

It was so hard to get it to work that I just thought it wasn't going to work out.

我还以为没法规模化呢，结果居然做到了。

I thought it wasn't going to scale, but then it did.

所以我当时还挺意外的。

So I was like, I was pretty surprised.

发生了什么变化？

What changed?

需求特别高，我们就是硬着头皮把它搞定了，对吧？

The demand was very high and we just, we got it to work, right?

我们尝试了好多不同的方法，才让环境真正提升了模型的表现。

We just had to try like a lot of different things to get environments to actually improve model performance.

所以我们就一直坚持做，最后还真成了。

So we just kept going at it and it ended up working.

我是你弟弟，今天我在大学里学计算机专业。

I'm your little brother and I'm studying computer science at universities today.

你把我叫过来跟我说，弟弟，你得知道这些。

You sit me down and say, little brother, you should know this.

我该知道些什么呢？

What what should I know?

赶紧去找个正经的实习，因为你在学校学到的东西很快就会过时。

Get a real internship as soon as possible because whatever you learn in school is probably gonna be outdated quickly.

我该去哪儿找真正的实习机会呢？

Where should I get a real internship?

我知道这听起来有点傻，但我是该自己创业吗？

I know that sounds stupid, but, like, should I start my own company?

我是该加入一家发展迅速的公司吗？

Should I join a fast growing company?

还是该加入一家非常成熟的大公司，毕竟那里有真正懂行的前辈坐镇嘛？

Should I join a super established company where there's, you know, adults in the room, so to speak?

也许我有点偏颇，但加入一家在旧金山快速成长的公司，不需要非得有资深人士在场，只要处于行业前沿，就能反映出这个领域的发展方向。

Maybe I'm biased, but join a fast growing company in San Francisco doesn't need to have adults in the room, but somewhere on the frontier that's indicative of where the field is going.

团队规模稍微大一点，比如超过十个人，别太早期，主要是为了筛掉那些可能没什么前途的公司。

A bit larger than 10 people, not super early, just to kind of like filter out the companies that might not go anywhere.

你觉得我加入Mercor是不是太晚了啊？

Would you say that you're too late for me?

Mercor？

Mercor?

不是。

No.

不是。

No.

它就像初创公司一样，门槛卡得很死。

It's locked like a startup.

你们现在有多少人？

How many people do you have?

大概五百人左右。

Maybe 500.

文化上，我们还是初创公司的风格。

Culturally, we're a startup.

我们总是保持着一种危机感。

We we're paranoid.

大家几乎一直都在办公室。

We're in office all the time.

我们行动很快。

We're fast moving.

我们想尽可能长久地保持这种状态。

We wanna hold on to that as long as possible.

我很喜欢。

I love it.

太棒了。

That's amazing.

完全同意。

Totally.

确实如此。

Absolutely.

对。

Yes.

你最佩服哪个竞争对手？为什么？

Which competitor do you most respect and why them?

我其实没怎么想过竞争对手的事。

I don't think about competitors too much.

他们跟Mercor比起来都还差得远呢，水平都差不多。

They're all kind of even in that they're all behind Mercor.

我们尽量不去关注他们，而是把心思都放在客户身上。

We really try not to think about them as much as we try to think about our customers.

所以我非常尊重我们的客户。

So I respect our customers a lot.

我很欣赏他们所做的工作。

I love the work that they're doing.

我们会密切关注竞争对手的动态，但每次我看他们的网站，发现他们做的东西，其实我们一两周前甚至一个月前就做过了。

We stay on top of what competitors are doing, but every time I look at one of their websites, they're just doing something we did like a week or a month ago.

你看，他们写篇博客，我们也写篇博客，结果内容几乎一模一样，只是晚了一周而已。

You know, write a blog, someone else writes a blog like a week later, that's the exact same thing.

我们更新一下网站，他们跟着更新，内容也是完全一样的。

We make an update to our website, someone else makes an update to their website, that's the exact same thing.

我花了很多

I spend a

时间

lot

你刚才说这话是指Surge吗？

of Did you say that about Surge?

以前也发生过类似的事。

It's happened before.

是的。

Yeah.

以前也发生过类似的情况。

It's happened before.

他们确实有点不按常理出牌。

They're they're a bit out there.

说实话，我花在他们身上的心思不多，因为我更多精力都放在客户身上。

We honestly, like, I don't spend that much time thinking about them because I spend more time thinking about customers.

我们注意到他们确实有点特立独行，不怎么模仿我们，跟行业里的其他人也不太一样。

We've seen it they're they're a bit out there in in that they they don't copy us as much and they do seem a bit different from others in the field.

很难说具体原因是什么。

Hard to say why.

他们非常低调保密。

They're very secretive.

确实如此。

Yes, absolutely.

你能不能详细说说，Mercor 是怎么做到两千亿美元市值的？

Can you please paint the bull case for how Mercor is a $200,000,000,000 company?

当然可以。

Yeah, for sure.

看起来我们是在卖服务。

So it looks like we sell services.

我们基本上是一家技术驱动的服务型公司。

We're basically like a tech enabled services company.

我们的服务对客户提升收入非常关键，主要靠的是更强大的模型能力。

Our services are incredibly valuable in driving revenue gains for our customers, primarily through better model capabilities.

目前，评估数据和训练数据是制约模型表现的主要瓶颈。

Evals and training data are the primary bottlenecks to model performance right now.

每家企业都需要专属的私有模型。即便模型能力逐渐饱和，评估集依然能像产品需求文档那样，明确你到底想要什么，同时也是优化模型性能的目标。

If every enterprise needs to have specialized proprietary models, even if the capabilities start to saturate, the eval serve as the PRD for kind of like exactly what you want, but also the optimization objective for better performance.

只要更优秀的模型对经济有价值，评估集和训练集的需求就会一直存在。

As long as more and more as long as better models are valuable to the economy, there will be demand for eval sets and training sets.

如果我们能不断加快这个流程，就能满足日益增长的人类数据需求，用于评估和训练，同时我们也在拓展企业级的智能体部署业务。

If we can make that process faster and faster, we can serve a growing demand for human data for eval training, and then we also have a growing agent deployment enterprise arm as well.

目前你们还没有哪条收入线，是你认为三年后会变得非常重要的？

What line of revenue do you not have today that you think will be very significant in three years' time?

我觉得现实世界中的物理数据在未来三年会有显著增长。

I think that real world, like, physical data is going to grow significantly over the next three years.

机器人领域对我们来说是个很有意思的方向。

Robotics is an Robotics interesting area for us.

跟生成式AI还有自动驾驶比起来，机器人领域的数据市场还处在起步阶段。

The data market for robotics is nascent relative to Gen AI, relative to, you know, like autonomous vehicles as well.

我们觉得这块市场会有很大的增长空间。

And we think that's going to grow a lot.

你们会提前扩充供应，去应对未来的需求吗？

Do you scale supply ahead of demand?

有时候我们会留住一些顶尖人才，让他们做一些现在可能用不上、但对未来有价值的工作。

We, at times, retain exceptional talent to do work that might be valuable in the future.

我们可以直接利用现成的数据创建流程，在需求低的时候先囤积数据，等以后需求高了再转手卖掉。

And we can do, like, off the shelf data creation to basically, like, make use of supply when demand is low and then resell that data later.

如果是这种情况，我们就做；否则就不做。

In that case, we do, otherwise we don't.

你听过最好的建议是什么？

What was the best piece of advice you've ever been given?

很多人建议我加入小公司、初创企业，或者搬到旧金山去。

I got a lot of advice to join small companies, join startups, move to San Francisco.

我在加拿大长大，在多伦多上的学。

I grew up in Canada, I went to school in Toronto.

我听了那个建议，我觉得特别棒。

I followed that advice, I think it was great.

我很喜欢住在这里，我也喜欢小公司，还有发展特别快的公司。

I've loved living out here, and I like small companies, I like fast growing companies.

这段经历特别有意思，对我的职业生涯帮助也很大。

It's been super fun and great for my career.

最后一个问题。

Final one for you.

你最期待什么，而且觉得大家讨论得还不够的？

What are you most excited about that you don't think enough people are talking about?

答案可能跟之前差不多，就是看好机器人领域未来三年的机会。

Probably the same answers before in that, like, the three years out opportunity of robotics.

我觉得现在推特上关于机器人的讨论很多，还有在

I think there's a lot of discussion around robotics on Twitter and in

我确定，抱歉啊，兄弟。

I'm certain sorry, dude.

你能帮帮我吗？

Can you just help me out here?

这时候我就容易出错。

This is where I get in trouble.

现在是周五下午。

It's Friday afternoon.

已经六点了。

It's past six.

管他呢。

Fuck it.

我想说啥就说啥。

I can say what I want.

我不明白。

I don't get it.

好吗？

Okay?

每次看你们那个机器人演示，画面里机器人走得歪歪扭扭，在屋里转悠，然后看它从冰箱里拿瓶水，居然花了十五分钟。结果你发现，其实Brandon一直在另一个房间，全程都是他在搞鬼，你心里肯定在想：你这是在逗我吗？

Whenever you watch your robotics doc demo, they're like, you see this kind of terribly moving robot around a home, and then after watching it, like, take one water out of a fridge in fifteen minutes, it goes, and Brandon was in the other room all along, all gonna and you're like, are you fucking kidding me?

我厨房里明明有个机器人在那儿磨蹭了十五分钟拿水，结果其实是Brandon在隔壁房间干的？

I had this absolute spacko in my kitchen for fifteen minutes getting a water, and Brandon was in my backroom doing it?

我们就发展到这一步了？

That's where we're at?

我到底漏掉了什么？

What am I not seeing?

帮帮我，让我提起点兴趣。

Help me help get excited.

对。

Yeah.

我觉得吧，要是往回倒个十年左右，自动驾驶汽车那时候车里一直都有人。

Think I think if you go back like, you know, a decade or so, like self driving cars had, you know, they had the people in them all the time.

你会看到那些自动驾驶车队在旧金山街头跑，而且车里一直都有人坐着，这情况持续了好多年。

You would see crews driving around San Francisco and there was like a person in it for years.

持续了好多年。

For years.

对吧？

Right?

但现在我坐Waymo的次数比坐Uber还多。

But now I take Waymo more than I take Uber.

车里有人，我在……

There's someone in I'm

我真为你高兴。

I'm thrilled for you.

欢迎来到伦敦。

Welcome to London.

我们车里还是有人坐着。

We still have these people in cars.

我很喜欢，但它目前只在一个城市运营。

I love it, but it's in one city.

它处理不了那些特别模糊、不确定的数据。

It can't deal with like very ambiguous data.

它在其他方面还是不太行。

It's still like pretty irrelevant.

不过在过去这段时间，它在旧金山、奥斯汀和凤凰城这些地方确实取得了巨大的进步。

It's made leaps and bounds in the past, at least in San Francisco and Austin, Phoenix.

这确实有点难，因为，对，我觉得分布不太均匀，但旧金山这边的服务真的非常棒，当地人用得特别多。

It's tough because, yeah, I guess the distribution is unequal, but it's an incredible service here in San Francisco and people here use it a lot.

所以从技术层面来说，它是行得通的，可能只是，你知道的，在扩展规模时会遇到监管方面的挑战或者其他问题，但是……

So like technically it works and it might be, you know, there might be like regulatory challenges or other challenges with scaling, but

不过，你觉得机器人领域会出现类似 ChatGPT 那样的爆发时刻，从而引发使用和普及率的拐点吗？

But you think we'll hit a ChatGPT moment with robotics, which will cause an inflection in usage and adoption?

是的，我觉得会。

Yeah, think I think so.

不过我觉得情况可能跟无人驾驶汽车类似，扩展实体硬件确实比扩展软件难得多。

But I think it might play out similar to driverless cars where it's really hard to scale physical things as opposed to software.

所以这更像是一个像Waymo、RoboTaxi或Cruise那样的时刻，而不是ChatGPT那样的时刻。

So it might be more of like a Waymo, Robo Taxi Cruise type moment than ChatGPT moment.

但我相信进步会有的。

But I think the progress will be there.

哥们，你讲得太棒了。

Dude, you have been fantastic.

非常感谢你包容我这种天马行空、结构松散但又很精彩的对话，我真的特别感激。

Thank you so much for putting up with this, like, incredibly wayward, poorly structured conversation, which was brilliant, and I so appreciate you putting up with it.

谢谢你的邀请。

Thanks for having me.

是的。

Yeah.

真的特别有意思。

It was super fun.

不过在结束今天的对话之前，我想说大多数公司都尝试过AI。

But before we leave you today, most companies have tried AI.

但大多数公司都没看到什么成效。

Most aren't seeing results.

这并不是因为AI没用，而是因为它还没真正融入大家的工作流程。

Not because AI doesn't work, it's because AI hasn't reached the workflows yet.

这正是Asana致力于填补的空白。

That's the gap Asana is built to close.

Asana 是为人机协作团队打造的操作系统，能帮你轻松实现全团队的 AI 生产力提升。

Asana is the operating system for human agent teams, your easy button for AI productivity across every team.

开箱即用的 AI 队友，专为市场、运营和 IT 团队预先配置好。

Ready to go AI teammates, prebuilt for marketing, ops, and IT.

不需要你专门去学怎么写提示词。

No prompt engineering.

也不用费心做复杂的设置。

No set up.

它们直接出现在工作发生的地方，已经融入你的工作流程，随时准备交付成果。

They show up where the work is happening, already onboarded in your workflows, ready to deliver.

有了 Asana，无论你们是十人的小团队还是上万人的大组织，全公司都能在同一份计划下朝着同一个目标努力。

With Asana, your whole company can work on the same plan towards the same goal, whether you're a team of 10 or a team of 10,000.

Asana，让人类和智能体协同工作。

Asana, where humans and agents workflow together.

访问 asana.com 即可体验。

Try it at asana.com.

网址是 asana.com。

That's asana.com.

Asana 让工作顺畅推进，而 Zero Hash 则助力链上资金流转。

While Asana keeps the work moving, Zero Hash helps move money on chain.

任何优秀的软件公司最终都会面临资金问题。

Every great software company eventually runs into money.

比如 Uber 就必须解决资金流动的问题。

Uber had to move it.

Shopify 必须得持有它。

Shopify had to hold it.

Airbnb 必须得结算它。

Airbnb had to settle it.

资金流转不再只是金融科技的问题了。

Money movement stopped being a fintech problem.

它变成了软件工程的问题。

It became a software problem.

而且越来越明显，天哪，这甚至成了个AI问题。

And increasingly, god, it's an AI problem too.

Zero Hash就是让全球资金即时流动变得无缝衔接的基础设施。

Zero Hash is the infrastructure that makes global instant money movement seamless.

只需一次API集成，就能搞定稳定币、数字资产和现代支付通道，让开发者专心做开发。

One API integration for stablecoins, digital assets, and modern payment rails so builders can stay builders.

Zero Hash为全球一些最大的企业和金融机构提供动力，包括Kalshi、Stripe、摩根士丹利、Gusto和盈透证券。

Zero Hash powers some of the world's largest enterprises and financial institutions, including Kalshi, Stripe, Morgan Stanley, Gusto, and Interactive Brokers.