They Raised Ten Times Less Than Their Competitor. Then They Won.

Good afternoon, distinguished guests.

各位来宾，下午好。

Welcome to the Astana Hub venue.

欢迎来到 Astana Hub。

The digital heart of our country.

这里是我们国家的数字心脏。

A unique place where new ideas take shape, and where the future of our digital ecosystem takes shape.

这是一个孕育新想法的地方，也是塑造我们数字生态未来的地方。

Here anyone can find their own path of growth in the world of IT.

在 IT 的世界里，每个人都能在这里找到自己的成长路径。

Both educational programs, and incubation and acceleration programs.

既有教育项目，也有孵化和加速项目。

This is a place that draws in bright minds, inspiration and networking — just like today.

这里聚集人才、灵感和人脉，就像今天这样。

And today we're glad to welcome you all to a fireside chat dedicated to our first unicorn in Kazakhstan's history.

今天我们很高兴欢迎各位来到这场 fireside chat，主题是我们哈萨克斯坦历史上第一家独角兽。

Now we'd like to invite to the stage the Deputy Prime Minister, Minister of Digital Development and Artificial Intelligence, Zhaslan Khasenovich Madiyev; also the CEO of Astana Hub, Magzhan Madiyev; and our guests today — the co-founder of the first unicorn in Kazakhstan's history, Higgsfield AI, Yerzat Dulat, and the largest venture investor in Central Asia, a serial tech entrepreneur and founder, Murat Abdrakhmanov.

现在有请副总理、数字发展与人工智能部部长 Zhaslan Khasenovich Madiyev 上台，还有 Astana Hub 首席执行官 Magzhan Madiyev，以及今天的两位嘉宾——哈萨克斯坦历史上第一家独角兽 Higgsfield AI 的联合创始人 Yerzat Dulat，还有中亚最大的风险投资人、连续技术创业者、创始人 Murat Abdrakhmanov。

Let's give a warm welcome to the legends of our IT industry.

让我们用掌声欢迎我们 IT 领域的传奇人物。

I'd like to welcome each of you and thank you for the interest you've shown.

我想欢迎在座的每一位，也感谢大家对这件事的关注。

Interest in the industry — in the venture financing industry, in startups, in the startup ecosystem — and most of all, of course, in our first startup, our first little star in artificial intelligence, the one that became the first unicorn in our region and proved that technology like this can be born in Kazakhstan.

大家关注这个行业——关注风险投资，关注创业公司，关注创业生态——而最重要的，当然是关注我们第一家创业公司、我们在人工智能领域的第一颗新星，它成了我们这片土地上的第一家独角兽，证明了这样的技术也能在哈萨克斯坦诞生。

Today, I'd say, we actually have two heroes here as our guests.

今天，我想说，我们其实请来了两位英雄。

Yerzat Dulat and Murat Abdrakhmanov, who at some point found this symbiosis, found this partnership, and laid the first stone of the technology, of the product that we have today.

Yerzat Dulat 和 Murat Abdrakhmanov，他们当年找到了这种共生关系，找到了这段合作，为我们今天看到的这项技术、这款产品奠下了第一块基石。

I'm sure this is only the beginning.

我相信这只是个开始。

There are many more funding rounds ahead, and many opportunities for Kazakhstan to be on the world map of the venture industry, of the startup industry.

前面还有很多轮融资，也会有很多机会，让哈萨克斯坦出现在全球风险投资和创业的版图上。

Today it's no secret to anyone that the kind of technology Higgsfield grew out of exists in only three countries.

今天大家都清楚，Higgsfield 所诞生的这类技术，全世界只有三个国家有。

The United States, China, and Kazakhstan.

美国、中国和哈萨克斯坦。

So we, as representatives of the public sector, of the government, of the country's leadership, are genuinely proud of this young generation that is laying the groundwork for Kazakhstan's technological future.

所以我们作为公共部门、作为政府、作为国家领导层的代表，真心为这一代年轻人感到骄傲，他们正在为哈萨克斯坦的技术未来打下底子。

The head of state pays a great deal of attention to digitalization and artificial intelligence.

国家元首非常重视数字化和人工智能。

This year was deliberately named the year of digitalization and artificial intelligence, so that the state could give every kind of major support — institutional support, through legislative initiatives, through allocating extra funding, through investment in human capital — to build a strong technological base, a foundation.

今年被特意定为数字化与人工智能之年，就是要让国家能在各个方面给出有力支持——制度上的支持、立法上的举措、追加拨款、对人力资本的投入——好打下一个扎实的技术基础，一个地基。

for the development of this area, because we have to give rise to a new economy.

支撑这个方向的发展，因为我们必须催生一种新的经济。

And with that in mind, I want to wish that more unicorns appear in our country — unicorns, decacorns.

借此机会我想说，希望我们国家能再出现别的独角兽，出现 unicorn，出现 decacorn。

And on our side, the main thing is to support that growth and not get in its way.

而我们这边，最重要的就是支持这种发展，别去妨碍它。

Thank you very much, Zhaslan Khasenovich, for those words of welcome.

非常感谢 Zhaslan Khasenovich 的欢迎致辞。

And now we can move straight on to our chat session.

现在我们可以直接进入今天的对谈环节了。

With your permission, I'll take on the role of moderator for our session today.

承蒙允许，今天我来担任这场对谈的主持人。

And thank you all so much for your interest and attention, for gathering here on such a cold winter day.

也非常感谢大家的关注，感谢各位在这样一个寒冷的冬日聚到这里。

And you can see, Make, Yerzat, how much interest there is in IT here in the capital and across the country as a whole.

Make、Yerzat，你们也看到了，首都乃至全国对 IT 的热情有多高。

Well, of course, I'd like to move on now to our guests' backgrounds.

当然，我想先聊聊两位嘉宾的背景。

First question, Make: how did you end up on the path of a venture investor in the first place?

第一个问题，Make，您当初是怎么走上风险投资这条路的？

How did the idea come to you?

这个想法是怎么冒出来的？

How long ago was it?

是多久以前的事？

And walk us through it a little, please.

也请您稍微讲讲这个过程。

Well, I'm a serial technology entrepreneur myself — I've been in business for about 35 years, and practically all of that business has been tied to the tech sector, and the core business is telecom.

我自己是个连续技术创业者，做生意大概有 35 年了，几乎所有的生意都跟技术行业有关，主业是电信。

Besides telecom, there were about twenty other projects in the technology space, all kinds of them.

除了电信，还有大约二十个各式各样的技术类项目。

So for me the venture industry feels very close, very familiar, very natural.

所以对我来说，风险投资这个行业感觉很近、很亲、也很自然。

The fact that I ended up there was a very evolutionary thing for me, very logical.

我走到这一步，是很自然的演进，逻辑上很顺。

It all started when, back in 2012, I decided to diversify my country risk — I mean, I asked myself what I could do abroad.

起点是 2012 年，我决定做国别风险的分散，就是问自己，在国外还能做点什么。

I looked at different options — hotels and so on.

我看了各种选项，酒店之类的。

In the end I realized you have to do what you actually understand, and what I understand is the technology space.

最后我明白，还是得做自己真正懂的事，而我懂的是技术这块。

So I started investing in projects aimed mainly at big markets — the US and East Asia — and mostly run by founders from the CIS.

于是我开始投项目，主要面向大市场——美国和东亚——创始人大多来自独联体。

Ukrainian projects, Belarusian ones, Russian ones.

有乌克兰的项目、白俄罗斯的、俄罗斯的。

And after that, of course, the first Kazakh projects came along too.

再往后，当然也出现了第一批哈萨克斯坦的项目。

That was Chocofamily in 2016.

那是 2016 年的 Chocofamily。

The first project where I was a major investor.

那是第一个我作为主要投资人参与的项目。

And starting somewhere around the 2000s, the venture industry here in Kazakhstan really began to develop, and then about three years ago there was, let's say, a qualitative leap, a qualitative change — because before that, honestly, we saw quite a big gap between our projects and, say, ones out of Kyiv or Moscow.

大概从两千年代起，哈萨克斯坦的风投行业就开始活跃起来，而三年前左右发生了一次可以说是质的飞跃、质的变化——因为在那之前说实话，我们的项目和基辅、莫斯科那边的项目，差距还挺大。

I mean, the founders there were better prepared, more professional.

就是说，那边的创始人准备得更充分，也更专业。

But about three years ago there was a qualitative leap.

但大概三年前出现了一次质的飞跃。

Apparently quantity turned into quality, or maybe it was just time — I don't even understand what happened — but we saw that a new generation of founders had simply arrived, completely prepared, sharp, competent.

大概是量变到了质变，也可能只是时候到了，我自己都说不清究竟发生了什么，但我们看到，一批全新的创始人出现了，准备充分、脑子清楚、能力过硬。

And today they're not behind anyone in anything.

到今天，他们在任何方面都不比别人差。

And today's event is about Higgsfield — and I don't want to call it a phenomenon, because it isn't a phenomenon; it's actually the natural result of everything we've been through.

今天这场活动讲的是 Higgsfield，我不太想把它称作某种现象，因为它不是现象，它其实是我们一路走来的必然结果。

It's probably the result of our very high level of education — the schools that prepare kids in STEM: NIS, KTL, RFMSh, schools that pick out talent from all over the country.

这大概是因为我们的教育水平很高，那些做 STEM 培养的学校，NIS、KTL、RFMSh，从全国范围里挑选人才。

And it's the level of digitalization, and the whole attitude toward it from the state and from society — these are our new heroes.

还有数字化的程度，以及国家和社会对这件事整体的态度——这些人就是我们新的英雄。

And that's really, really good.

这一点真的非常好。

And I'd like to say that Higgsfield itself was the trigger that put Kazakhstan on the world venture map.

我想说，Higgsfield 本身就是那个触发点，把哈萨克斯坦放上了全球风险投资的版图。

I mean, honestly, when you come to the Valley — Kazakhstan? Who's that?

说实话，你去一趟硅谷，人家会问：哈萨克斯坦？那是谁？

Kazakhstan — where is that, what is that?

哈萨克斯坦，那在哪儿，那是什么？

The first thing that comes into people's heads is Borat.

人们脑子里冒出来的第一个词是 Borat。

Now the first thing that comes into people's heads is Kaspi and Higgsfield.

现在人们最先想到的是 Kaspi 和 Higgsfield。

That's real, and that's something the country should be proud of.

这是实打实的，也是这个国家值得骄傲的地方。

There you go.

就是这样。

We'll definitely go into Higgsfield in a lot of detail — our guys' experience, Yerzat's, and so on.

我们后面一定会非常详细地聊 Higgsfield，聊我们这几位年轻人的经历、Yerzat 的经历等等。

But still, I'd like to unpack your background a bit more.

不过我还是想先把您的背景多展开一点。

I remember when I used to go to international trade shows and walk up to the different booths.

我记得当年我去参加国际展会，一个个展台走过去。

I remember there were startups from Ukraine, Belarus, Europe.

记得有来自乌克兰、白俄罗斯、欧洲的创业公司。

And they'd go, "Where are you from?" And I'd say, "I'm from Kazakhstan."

他们就问：“你从哪儿来？”我说：“我从哈萨克斯坦来。”

And they'd say, "We have investors from Kazakhstan."

他们说：“我们有来自哈萨克斯坦的投资人。”

And then the next one: "We have an investor from Kazakhstan."

然后下一家又说：“我们有一位哈萨克斯坦的投资人。”

And it turned out it was Murat Almazbekovich who had invested in those projects — but that was a long time ago, five to seven years back.

后来才知道，投这些项目的就是 Murat Almazbekovich，不过那是很久以前了，五到七年前的事。

So, tell us please — how many startups do you have in your portfolio right now, and roughly which sectors are they in?

那，可能先请您讲讲，您现在被投组合里有多少家创业公司，大概分布在哪些行业？

Well, I'm probably a bit past 70 by now.

我这边应该已经稍微超过 70 家了。

I haven't counted lately.

最近一次我都没再数过。

So, roughly around 70 startups.

反正大概 70 家左右。

Right.

就是这样。

As for sectors — basically, we don't have that much choice here, not the way you do in the US, so focusing on a single sector is hard.

至于行业，基本上是这样：我们这边本来就没那么多东西可选，不像美国选择那么大，所以只聚焦一个行业很难。

It's easier to say where we don't invest.

其实说我们不投什么反而更容易。

We don't invest in things I personally don't understand.

我自己看不懂的东西，我们不投。

Because for me it's always important to understand it, to have my own assessment and my own read on the risk for every project.

因为对我来说，最重要的永远是自己能看懂，对每个项目都有自己的判断、自己对风险的认知。

We don't invest in agriculture, and we don't invest in deep tech or biotech.

农业我们不投，深科技和生物科技也不投。

Not because there's no potential there, or because it isn't interesting, but because it's very complicated.

不是因为它没潜力、不够有意思，而是因为它太难了。

And honestly, neither I nor my team understand anything about it.

说实话，这块不管是我还是我的团队，都完全不懂。

And really, if we don't understand something, we don't go in.

其实就是，看不懂我们就不进。

That's our principle.

这是我们的原则。

We also don't invest in marijuana projects.

大麻类的项目我们也不投。

That one too — even though we do know a bit about it.

这个也不投，虽然我们对它多少还懂一点。

Right.

就是这样。

And since 2014, of course, out of those seventy, about 15% of my portfolio is Ukrainian.

另外从 2014 年开始，这 70 家里大概有 15% 是乌克兰的项目。

Back around 2015 I was actually known much better in Ukraine than in Kazakhstan.

也就是说，2015 年前后，我在乌克兰的名气比在哈萨克斯坦大得多。

I mean, I'm one of the well-known investors in Ukraine.

算是在乌克兰比较有名的投资人之一。

Yeah.

对。

Because the teams there were really competent.

因为那边的团队真的很专业。

And from 2012 on, Ukrainians went global right away.

而且从 2012 年起，乌克兰人一上来做的就是全球市场。

In all the eleven years I've been doing this, I have never once received a pitch in Russian from a Ukrainian team.

我做这行 11 年，从乌克兰人那里一次都没收到过俄语的 pitch。

And that doesn't say anything about Russophobia — it just means they simply don't have one in Russian.

这不说明有什么反俄情绪，只说明他们压根就没有俄语版。

They build for the global market from day one.

他们一开始就是冲着全球市场去的。

Right.

就是这样。

And I think that's just as relevant for us today.

我觉得这一点对我们今天同样成立。

We worked with Belarusian projects, and with Russian projects before the war, of course.

我们做过白俄罗斯的项目，也做过俄罗斯的项目——当然是战争之前。

Yeah.

对。

And basically, yes, there is this Russian-speaking VC network that spans the whole world.

总的来说，确实存在这么一个遍布全球的俄语 VC 网络。

I'm part of it.

我就是其中一员。

And that, essentially, is where the deal flow comes from.

说白了，我们的项目来源也正是从这里生出来的。

Founders know us well, and they pass us along hand to hand — not like a flag, more like a torch, I guess.

创始人都很了解我们，会把我们一个传一个地推荐出去——不像传旗帜，更像是传火炬吧。

And that's how the industry works.

这个行业就是这么运转的。

So, roughly that.

大概就是这样。

Mm-hm.

嗯。

I see we have a fairly young audience today, but still, while I have the chance, I'd like to send a signal to our society that venture investing is very profitable.

我看今天到场的观众比较年轻，不过还是想借这个机会，向社会传递一个信号：风险投资是非常划算的。

Obviously it's very risky, but it's very profitable.

当然，风险很大，但回报也很高。

We're doing a lot of work to convert our big entrepreneurs from traditional industries into venture investors.

我们正在做大量工作，想把传统行业里的大企业家转化成风险投资人。

And you said yourself that you were in the traditional sector as well.

您刚才也说到，您同时也在传统行业里做过。

You have a whole diversified set of businesses, including in the traditional sector.

您的生意很多元，其中也包括传统行业。

Could you tell us a bit about that?

能不能稍微讲讲这一块？

And then, how exactly did the idea of investing in IT come about?

还有，后来到底是怎么冒出投 IT 这个想法的？

Well, as I said, my goal was really geographic diversification — I've been in business a long time, I have a lot of experience and a lot of assets, and one fine day I realized that all of those assets were in Kazakhstan.

我前面已经说过，我的目标其实就是做地域上的分散，因为我做生意时间长、经验也多，手上资产不少，结果有一天我突然意识到，这些资产全都在哈萨克斯坦。

And geopolitically, Kazakhstan is a fairly risky place.

而哈萨克斯坦在地缘政治上是个风险不小的地方。

I've said before that black swans — entrepreneurs will get this — fly off to the West once every ten years, and the rest of the time they live here with us.

我以前讲过，黑天鹅每十年才往西方飞一次——做生意的人都懂——剩下的时间它们都住在我们这儿。

They may even be bred here.

说不定还是在我们这儿养出来的。

Every year here it's either a devaluation, or if not here, then something happens at the neighbors'.

我们这儿每年不是货币贬值，就是自己这边没事、邻居那边出事。

And by now we're practically genetically prepared for it.

到现在我们几乎在基因层面上就做好了准备。

Right.

就是这样。

And I realized that for stability I needed to move at least 50% of my assets outside Kazakhstan.

我意识到，要想稳住，至少得把 50% 的资产挪到哈萨克斯坦以外。

So I looked at different options — hotels, various businesses — and realized that none of it actually works.

我看过各种方案，酒店啊、别的生意啊，最后发现这些其实都行不通。

What does work is the thing I understand and know my way around: the technology business.

真正行得通的，是我看得懂、也在行的东西，也就是科技生意。

So I started looking at venture projects.

于是我开始看风险投资项目。

Yeah.

对。

And it started going well for us.

结果我们还真做得挺顺。

Partly, I think, because this is basically what I've done all my life, and I find it genuinely interesting.

我想，一部分原因是这本来就是我做了一辈子的事，而且我真的很感兴趣。

That's how I started putting money in.

我就是这么开始投钱的。

And it was quite a surprise to me that I actually managed to make money at it.

让我特别意外的是，我居然还真在这上面赚到了钱。

Right.

就是这样。

And not badly, either.

而且赚得还不少。

Yeah.

对。

And as of today, we do have traction.

到今天为止，我们的增长势头是有的。

The other funds in Kazakhstan still don't have enough traction, simply because not enough time has passed.

哈萨克斯坦其他基金的增长势头还不够，单纯是因为时间还不够长。

Right.

就是这样。

And that traction is very encouraging — it lets you say that yes, in venture, despite its high level of risk, you can make money too.

而这个增长势头非常乐观，足以让人说一句：风险投资虽然风险高，但照样能赚钱。

There are 18 asset classes in the world in total.

全世界一共有 18 种资产类别。

And the riskiest of them — you guessed it — is venture.

其中风险最高的，你们也猜到了，就是风险投资。

As of today I've already invested around 30 million dollars.

到今天为止，我已经投了差不多 3000 万美元。

And the portfolio's TVPI is running at about 4x across the whole portfolio, so it's pretty good.

被投组合的 TVPI 差不多是 4 倍，整个组合算下来，成绩还算不错。

We've already had 14-15 exits.

已经走完了 14 到 15 笔退出。

And just last night an offer came in for the sixteenth exit.

就在昨天夜里，第十六笔退出的报价也来了。

There you go.

就是这样。

So — congratulations.

那要恭喜您了。

Yes, in cash that number will probably climb to 24 million from the exits.

是啊，这些退出落到手里的现金，大概能涨到 2400 万。

So that's pretty good traction.

所以增长势头还是挺好的。

Yeah.

嗯。

So you can make money in venture, and probably you should.

也就是说，做风险投资是能赚钱的，大概也应该去做。

Right.

好。

Thank you so much for those points, for that message.

非常感谢您讲的这几点，感谢这段话。

We'll cut this into Reels now and run it on repeat.

我们这就把它剪成 Reels，反复播。

So the message here is that you were looking for diversification.

所以传出来的信息是：您当时在找分散配置。

You were looking for diversification, both geographic and sectoral.

您找的分散，既是地域上的，也是行业上的。

And that's how you discovered the world of venture, the world of IT.

就这样，您发现了风险投资的世界，发现了 IT 的世界。

And yes, today IT — including IT people from Kazakhstan — is exactly the area with the fewest barriers to scaling globally, to getting out into foreign markets as much as possible.

对，恰恰是今天的 IT，包括哈萨克斯坦的 IT 人，在往全球做大、尽可能打进海外市场这件事上，碰到的壁垒最少。

And that's why we call on our businesspeople from traditional sectors to take an active part in venture investing.

所以我们呼吁传统行业的企业家，积极参与到风险投资里来。

All right.

好。

I remember the first time I heard about Higgsfield was when I posted on Facebook asking what Kazakh AI startups people knew about.

我记得我第一次听说 Higgsfield 这个项目，是我在 Facebook 上发帖问：大家知道哈萨克斯坦有哪些 AI 创业公司。

And Murat Abdrakhmanov wrote in the comments: "Higgsfield."

Murat Abdrakhmanov 当时在评论里回了一句：「Higgsfield」。

And I thought, what is this Higgsfield?

我心想，Higgsfield 是什么来头？

I should get to know them.

得去认识一下。

It's probably an interesting project.

多半是个有意思的项目。

That was about two or three years ago.

那大概是两三年前。

Make, and how did you meet Yerzat?

Make，您当时是怎么认识 Yerzat 的？

Well, again, like I say, this is really important for the whole venture industry — the thing that always amazed me: you show up somewhere, say abroad, you start something with someone, you start talking to someone, and the first thing they say to you is, how can I help you?

还是我刚才说的那句，这一点对整个风投行业都特别重要——一直让我震撼的是：你到了某个地方，比如说到国外，跟人一起起个头，跟人聊起来，对方开口第一句就是，我能帮你做点什么？

And that culture, that ethic — I personally picked it up, and I always try to offer help and support somehow.

这种文化、这种行事规矩，我个人是接了过来，一直尽量给别人搭把手、给点支持。

And all of it comes back to you, I want to say.

而这一切都会回到你身上，我想说的就是这个。

So I was introduced to Yerzat by Sanzhar, from the company Salay?.

我跟 Yerzat 是 Sanzhar 介绍认识的，他是 Salay? 那家公司的。

Right.

嗯。

And even though Salay? shut down and I lost money there, I didn't let it ruin the relationship with him, and even so, Sanzhar introduced me to Yerzat — he said, there's this incredibly talented guru, and he said: "I'll introduce you to him, only I'll be translating for you."

尽管 Salay? 后来关掉了，我在那儿也亏了钱，但我没有因此把关系搞僵；后来正是 Sanzhar 把我介绍给 Yerzat，他说，有这么一位特别有天赋的高手，我把你们两个引荐一下，不过我得给你当翻译。

I said: "Translating what?

我说：「有什么好翻译的？

What is he..." Well, he says, you'll see. I met Yerzat for the first time, and Yerzat is someone deeply immersed in technology.

他难道……」他说，你到时候就知道了。我第一次见到 Yerzat，Yerzat 这个人是真的一头扎在技术里。

He says, for him all of this is ordinary, normal — a pile of technical terms.

他说，对 Yerzat 来说这些都很平常、很正常，一堆术语张口就来。

That was about two years ago — almost, almost three.

那还是两年前的事，差不多，差不多三年了。

Yeah.

嗯。

And I realize I don't understand half of it.

我发现，这里头有一半我都没听懂。

And Sanzhar is translating for me from Russian into Russian, spelling it out and explaining.

Sanzhar 就在旁边给我做俄语到俄语的翻译，一句句掰开讲给我听。

Right.

嗯。

But what really struck me, of course, was Yerzat's level of expertise.

但真正让我震撼的，当然是 Yerzat 的专业水平。

Afterwards I did a little — how do you say — reference check, I collected references, and all the references were very, very positive.

后来我做了个小小的——怎么说呢——背景摸底，找人打听了一圈，反馈全都非常非常正面。

He has his own library on GitHub, he'd gotten an offer from OpenAI, and all the big projects know him — he consults for them.

他在 GitHub 上有自己的开源库，OpenAI 给他发过 offer，那些大项目都认识他，他给他们做顾问。

Right.

嗯。

And I simply believed in him.

我就是单纯地信了他。

Honestly, I'm not deeply enough versed in AI to figure it all out myself.

说实话，我在 AI 上没有深到那个份上，没法把所有事都搞明白。

Yeah.

嗯。

But just as a person — he simply trusted me.

但单从人这一层讲——他也就这么信了我。

I wrote him his first check.

我给他开出了第一张支票。

That's how it went.

事情就是这么来的。

Awesome.

太好了。

Here I'd also like to highlight a point for our founders and entrepreneurs: how important networking is, how important reputation and word of mouth are, because a lot of venture deals happen exactly this way.

这里我也想给我们的创始人、创业者点出一条：人脉网络有多重要，名声和口碑有多重要——因为风投圈里很多交易，恰恰就是这么谈成的。

All right.

好。

Now I'd like to hand the floor to Yerzat.

现在我想把话筒交给 Yerzat。

Yerzat, please tell us a bit about yourself.

Yerzat，请你简单讲讲你自己。

We'd like to hear more from your personal story — where you're from, where you studied, and how you got to what you're doing now.

我们更想听你个人的经历——你是哪里人，在哪里念的书，又是怎么一步步走到今天做的这件事。

Check, check.

喂，喂。

Okay.

好。

Yes, I'm really glad to see so many people here.

很高兴看到这么多人来。

Thank you so much for the interest you've shown.

非常感谢大家的关注。

It's great that so many young people have gathered here and are interested in venture, in startups.

这么多年轻人聚到这里，关心风险投资、关心创业公司，这很棒。

So, I'm from Almaty, I went to RFMSh, the physics and math school, and I did physics olympiads.

我是阿拉木图人，中学念的是 RFMSh 物理数学学校，那时候打物理奥赛。

Yeah.

嗯。

And after I finished school, I was lucky enough to travel to America for a short stretch of time.

中学毕业之后，我很幸运，有机会去了一趟美国，待的时间不长。

And that was right when the movie The Social Network came out.

那阵子正好上映了电影《社交网络》。

And I was using the first internet services.

我也用上了最早的那批互联网服务。

At that point it was 2013 or 2012, I don't remember exactly.

当时是 2013 年还是 2012 年，我记不太准了。

And coming from Kazakhstan, obviously, in America, when you see how technology makes a person's life easier — I remember using Amazon, delivery right to your door, using Google Maps — all of it impressed me enormously.

从哈萨克斯坦过去，到了美国，看到技术怎么让人的生活变轻松——我记得当时用 Amazon，东西直接送到家门口，还用 Google Maps——这些给了我极大的冲击。

And so I decided to learn to program.

于是我决定去学编程。

Around 2012-2013 I started programming, and I decided to put off going further with physics and take up programming instead.

大概从 2012、2013 年开始写代码，我决定把继续念物理这条路先放一放，转去搞编程。

And programming quickly got boring for me.

但编程很快就让我觉得没意思了。

It's pretty boring, honestly, just giving a computer precise instructions and having it carry them out.

单纯给计算机下精确指令、让它照着执行，说实话挺无聊的。

So I started looking into what might be even more interesting.

于是我开始找，有没有更有意思的东西。

And that's when I found out about artificial intelligence, about that whole science.

就这样知道了人工智能，知道了这门学问。

And from around 2014 I started working on artificial intelligence.

大概从 2014 年起，我就开始做人工智能了。

At that point the community was very small.

那时候这个圈子非常小。

The OpenAI founders were all hanging out on Reddit back then.

OpenAI 那几位创始人当时全都泡在 Reddit 上。

It was pretty easy to talk to everyone there, make connections, share code and projects.

在那儿跟谁都能聊得上，能建立联系，还能互相分享代码和项目。

So, yeah.

就是这样。

And after that I worked on all sorts of projects, worked remotely with US projects, did projects in Kazakhstan.

后来我做过各种项目，远程参与美国的项目，也在哈萨克斯坦做项目。

And the moment GPT came out, it was clear that the time for artificial intelligence had finally arrived and that I needed to build a company in this space.

GPT 出来的那一刻就很清楚了：人工智能的时代终于到了，必须在这个方向上做一家公司。

And so I started this project, Higgsfield.

于是我开始做 Higgsfield 这个项目。

I remember watching Arman Suleimenov's podcast with Murat Abdrakhmanov, and I watched the whole podcast straight through, kind of devoured it.

我记得看了 Arman Suleimenov 和 Murat Abdrakhmanov 的那期播客，一口气看完，看得入迷。

It was really interesting.

特别有意思。

And right around then I was talking with Sanzhar, and Sanzhar offered to make an intro to Murat.

正好那阵子我在跟 Sanzhar 聊，他提出可以把我引荐给 Murat。

So, yeah.

就是这样。

And that's how the Higgsfield story began.

Higgsfield 的故事就是这么开始的。

Murat agreed to invest when the company didn't even have a legal entity yet.

Murat 答应投资的时候，公司连法人主体都还没有。

And Murat's lawyer walked us through how to set up a legal entity in the first place, how to receive the money into a bank account.

Murat 的律师还指点我们，法人主体到底该怎么注册，投资款怎么进银行账户。

So right there at the very beginning there was already a big contribution from Murat Almazbekovich.

所以在最开始的阶段，Murat Almazbekovich 就已经出了很大的力。

Great.

太好了。

What's especially interesting is that today we have you all here on stage together, so we can look at this story from different angles.

特别有意思的一点是，今天你们几位同时坐在这个台上，我们可以从不同角度来看这段故事。

Yerzat, I have a question for you.

Yerzat，我有个问题想问你。

What does Higgsfield mean?

Higgsfield 是什么意思？

Higgsfield — that comes from the Higgs field, a term from physics.

Higgsfield 来自 Higgs field，也就是物理学里的希格斯场。

Back when I was in America, that was exactly when they officially found the Higgs boson at the Large Hadron Collider.

我在美国的那段时间，正好赶上大型强子对撞机上正式发现 Higgs 玻色子。

And that same year Instagram came out, and I reserved the nickname Higgs on Instagram.

也是那一年 Instagram 出来了，我在 Instagram 上把 Higgs 这个用户名占了下来。

And I've had it on GitHub ever since.

从那以后，GitHub 上我一直用这个名字。

And on social media after that.

后来社交媒体上也都用它。

So, yeah.

就是这样。

And right around then, actually even before that, I was talking with John Schulman — he's an OpenAI co-founder, the creator of ChatGPT, ChatGPT itself was his idea — and I remember when we talked, he really liked my GitHub nickname.

而且正好在那前后，其实还要更早一点，我跟 John Schulman 聊过——他是 OpenAI 的联合创始人、ChatGPT 的缔造者，ChatGPT 本身就是他的点子——我记得聊天的时候，他特别喜欢我 GitHub 上这个用户名。

And when I registered the company, Higgsfield was originally supposed to be a purely deep tech infrastructure play.

注册公司的时候，Higgsfield 本来打算做纯粹的深科技基础设施方向。

And the name Higgsfield fit that well.

Higgsfield 这个名字跟它很搭。

Then a couple of years later we ended up becoming a big B2C entertainment business, but the name, yeah, stayed Higgsfield.

结果两年之后，我们做成了一家很大的 B2C 娱乐公司，名字倒是留了下来，还叫 Higgsfield。

Great.

太好了。

A very natural name for an RFMSh graduate.

对一个 RFMSh 毕业生来说，这名字再自然不过了。

Yes, in fact just the other day we were discussing exactly that with Make — with Zhaslan Khasenovich — and with Bagdat Atyrbekovich?, when we held a meeting: what a big role that layer, the secondary education layer in our country, has really played — schools like RFMSh, KTL, so many people came out of them.

对，前几天开会的时候，我正好也跟 Make、也就是 Zhaslan Khasenovich，还有 Bagdat Atyrbekovich? 聊到这个：我们国家中等教育这一层到底起了多大作用——像 RFMSh、KTL 这样的学校，走出来的人特别多。

That's where they came from, and this is one more confirmation of it: Higgsfield's co-founders came out of our, let's say, mathematical, physics-and-math environment.

他们正是从那里出来的，这也再次印证了这一点：Higgsfield 的联合创始人都出自我们这种，怎么说呢，数学、物理数学的环境。

Great.

太好了。

Well, then I'd like to hear how the project developed from there.

那接下来我想了解一下，这个项目后来是怎么发展的。

So we understand how it all started.

开头是怎么回事，我们已经清楚了。

And thank you, Murat Almazbekovich, for believing in these guys and giving them that push.

所以要谢谢您，Murat Almazbekovich，谢谢您相信这几个年轻人，给了他们这一把推力。

What were the next steps?

接下来又做了什么？

I remember you telling me how you raised the first American investment.

我记得你讲过，第一笔美国投资是怎么拿到的。

Please tell us about that — Yes, after Murat sort of got the company started, I began meeting American investors and also looking for a co-founder.

麻烦你讲一讲这段——好，Murat 算是给公司开了个头之后，我就开始接触美国的投资人，同时也在找联合创始人。

And that's very important.

这一点非常重要。

In general, if you look at the statistics, very few companies are built by solo founders.

总体上，如果看统计数据，单人创始做起来的公司很少。

Statistically, successful companies usually have two or three founders.

统计上看，成功的公司通常都有两到三位创始人。

So I was actively looking for a co-founder, I met a lot of people, and there were several really great people I got to know.

所以我很积极地在找联合创始人，认识了很多人，其中有几位真的非常优秀。

And one of them was Alex Mashrabov.

其中一位就是 Alex Mashrabov。

Nikolay Davydov introduced us — he's also an active venture investor in the States.

给我们牵线的是 Nikolay Davydov，他也是在美国很活跃的风险投资人。

So, yeah.

就是这样。

And Alex and I really clicked.

我跟 Alex 一下子就聊到一块去了。

And at one point he just decided to up and fly out to Almaty.

然后有天他干脆决定，说飞就飞来阿拉木图。

And the funny thing is that all the other potential co-founders I was talking to kept dragging their feet.

好玩的是，我接触过的其他几位潜在联合创始人，全都在拖时间。

But Alex just flew straight out to Almaty.

而 Alex 直接就飞到了阿拉木图。

We started working together right away.

我们马上就开始一起干。

I remember we worked in the lobby of the hotel where Alex was staying.

我记得我们就在 Alex 住的那家酒店大堂里办公。

Then we took a spot at SmartPoint and worked there, and somehow it all came together fast — we started working together, and while we were there in Almaty we decided to try building different products and pitching them to investors.

后来我们在 SmartPoint 租了个工位，在那儿办公。事情一下就转起来了，我们开始搭档，而且人就在阿拉木图，我们决定试着做各种产品，再拿去 pitch 给投资人。

We'd build about one prototype a day, pitch it, and watch how American venture investors reacted.

我们差不多一天做一个原型，拿去 pitch，再看美国风险投资人的反应。

It was a fun experience, because we had a pitch deck with something like 10 slides, and each slide had nothing to do with the next one, because each slide was its own mini-prototype.

那段经历挺有意思，因为我们的 pitch deck 大概十页，每一页跟下一页毫无关系，因为每页就是一个小原型。

One slide was about a multi-agent system in Minecraft, another slide was about distributed computing for neural networks, for training.

有一页讲 Minecraft 里的多智能体系统，另一页讲神经网络训练用的分布式计算。

And every slide gave us a signal.

每一页都给了我们一个信号。

We were basically trying to feel out what venture investors wanted to hear, and to align that with how we saw the market, so we could maximize the expected value of landing a successful round.

我们大致是在摸：风险投资人想听什么，再把这个跟我们对市场的判断对齐，好让顺利拿下一轮融资的期望值最大。

And in the end we pulled it off, but it was a pretty hard road.

最后我们做成了，但这条路相当难走。

The market was extremely overheated at that time, and most investors didn't understand what was going on.

当时市场过热得厉害，大部分投资人根本没搞明白在发生什么。

They were very skittish about every idea, because back then — now, for example, OpenAI has given up a bit of ground, Gemini has come out, Anthropic has become a great company, there are tons of projects, DeepSeek showed up.

他们对任何想法都缩手缩脚。因为那会儿——你看现在，OpenAI 的位置已经松动了一点，Gemini 出来了，Anthropic 成了一家很厉害的公司，项目一大堆，DeepSeek 也冒出来了。

But at that time every investor thought AGI, strong artificial intelligence, was right around the corner and that OpenAI would just swallow everyone.

但在当时，所有投资人都觉得 AGI、强人工智能马上就要来了，OpenAI 会把所有人一口吞掉。

So raising money at that moment was a pretty wild feeling.

所以在那个时间点融资，感觉挺魔幻的。

We got something like 100 no's at the start, I'd say.

一开始我们大概吃了 100 个「不」。

The first 100 pitches were all rejections, because the market was overheated, and American venture investors believed in teams from Stanford, all those Stanford professors, people from MIT, some brilliant students.

最早那 100 次 pitch 全被拒了，因为市场过热，美国的风险投资人只信斯坦福出来的团队、那些斯坦福教授、MIT 的人、某些天才学生。

Well, it's understandable - they invest in what they really understand.

这可以理解，他们投的是自己真正看得懂的领域。

And that part was pretty hard for us.

而这一块对我们来说相当难。

In the end we raised ten times less than our direct competitor, who was doing video generation.

结果我们融到的钱，只有那个做视频生成的直接竞品的十分之一。

At that point that was a company called Pika Labs.

当时那家公司是 Pika Labs。

They raised 80 million dollars, and we raised 8 million dollars.

他们融了 8000 万美元，我们融了 800 万美元。

And then two years went by.

然后两年过去了。

Pika Labs still exists as a company, but it's close to bankruptcy.

Pika Labs 现在还在，但已经接近破产了。

And we became the market leaders.

而我们成了市场的领跑者。

And that's exactly what showed that back then American investors were let down by the fact that they put more money into the stories that felt familiar and recognizable to them.

这恰恰说明，当时的美国投资人就栽在这一点上——他们更愿意把钱投给自己熟悉、认得出的那类故事。

They pinned a stereotype on you, right?

等于给你们扣了顶刻板印象的帽子，对吧？

Great.

太棒了。

I think this deserves a round of applause.

我觉得这里可以鼓个掌。

Yes, and I'd like to point out another important layer here.

对，我想在这里再补一个很重要的层面。

Folks, Yerzat was just telling us from his own experience how they tested different hypotheses, kept pitching, got a hundred no's.

各位，Yerzat 刚才用自己的经历讲了，他们怎么去验证各种假设，一遍遍地 pitch，收了一百个 no。

That's also a very sound, effective approach - they worked by a methodology, by the lean methodology, and that's something worth learning too.

这也是一套非常正确、非常有效的打法——他们是按方法论在做，按精益方法论，这一点同样值得学。

And that's exactly what our accelerators teach startups today.

而这恰恰就是我们的加速器今天在教创业公司的东西。

So - commercial break here - if you want to launch startups, take an active part in the acceleration programs.

所以，插播一段广告：如果你想做创业公司，一定要积极参加加速器项目。

We have a lot of programs.

我们的项目非常多。

And we often don't get enough applications, so all of these opportunities are open to you, friends.

但我们常常收不到足够多的申请，所以这些机会对各位都是敞开的，朋友们。

All right.

好。

Actually, I feel like I glossed over something a little.

不过我感觉刚才有一段我讲得太快、漏过去了。

Those 8 million - how did you manage to raise that?

那这 800 万美元，你们当时是怎么融到的？

When we were pitching, we were feeling out what investors actually respond to.

我们在 pitch 的过程中，其实是在摸索投资人到底吃哪一套。

And after a hundred pitches we had a clear story that had emerged, one that actually worked.

一百场 pitch 之后，我们手里就有了一个说得清楚的故事，而且这个故事是奏效的。

The only problem was that all the top funds - Andreessen, Accel - had already told us no.

唯一的问题是，Andreessen、Accel 这些顶级基金，当时已经全都对我们说了不。

And only one of the top-10 funds was left: Menlo Ventures.

前十的基金里就只剩下一家——Menlo Ventures。

And by the time we pitched them, we already had a story that hung together.

等我们去 pitch 他们的时候，我们的故事已经很完整了。

We had already gotten term sheets from smaller investors, European ones.

我们已经从一些规模小一点的投资人那里拿到了 term sheet，欧洲的投资人。

We already had a term sheet from a European venture fund, out of London.

我们手上已经拿到一家欧洲风险基金给的 term sheet，来自伦敦。

So by the time we pitched Menlo Ventures, we had a clear story that had been calibrated.

所以等我们 pitch Menlo Ventures 的时候，我们的故事已经很清楚，也已经被打磨校准过。

And they said yes - there was a partner there who had invested in mobile apps before, in B2C projects, in entertainment, and she believed in our pitch, in our idea.

他们就说了「可以」——他们那边有一位合伙人，之前投过移动应用、投过这类 B2C 项目、投过娱乐方向，她认可我们的 pitch，也认可我们的想法。

And we built a mobile app very fast, launched it very fast, and at the time it even went relatively viral.

而我们很快做出了移动应用，也很快上线，当时它甚至还有了一定的病毒式传播。

And that's how we raised our first 8 million, from Menlo.

我们就这样从 Menlo 那里融到了第一笔 800 万美元。

But as I said earlier, our closest competitors raised 80, so we were serious underdogs.

但就像我前面说的，最近的竞争对手融了 8000 万，我们当时完全是被看衰的一方。

The big American investors thought we had no chance, because our direct competitors had far more funding, and their team was seen as much stronger, so it was a pretty hard time.

也就是说，那些大的美国投资人认为我们没机会——直接竞品的融资额比我们多得多，大家也觉得他们团队更强，所以那段时间确实挺难熬的。

You've touched on a very interesting point here - that Kazakhstan today isn't really a trusted destination for venture capital.

您刚才提到了一个很有意思的点：哈萨克斯坦今天在风险资本眼里还算不上一个值得信任的目的地。

Nor are people coming out of Kazakhstan, out of Central Asia in general.

从哈萨克斯坦、从整个中亚走出来的人也一样。

And right now there's this prejudice that all the very top founders have to be based in Silicon Valley.

现在还存在这样一种成见：最顶尖的创始人就该待在硅谷。

And your case is probably exactly what shows that isn't always so.

而你们这个案例，大概恰好说明了事情并不总是这样。

Right.

是这样。

Murat, in your opinion, how should we work on this stereotype, this barrier, so that more and more of our founders can raise American venture capital?

Murat，在您看来，我们该怎么去处理这种刻板印象、这道门槛，让越来越多我们的创始人能拿到美国的风险资本？

I think, first of all, the situation has already changed.

我觉得首先，情况其实已经变了。

I'll tell you - last year, the year before, I was testing international PR, I spent on it, set aside a budget for it, and I was writing articles for Forbes.com and various tech outlets.

我讲个事，去年、前年我试过做国际 PR，实打实花了钱、划了预算，在 Forbes.com 还有各种科技媒体上写文章。

Right.

是这样。

And we tested different things.

我们试了不同的说法。

At first we introduced ourselves as one of the leading venture investors in Central Asia, and it landed absolutely nowhere.

一开始我们把自己介绍成「中亚领先的风险投资人之一」，结果完全没人理。

Nobody had heard a thing about Central Asia and technology, right?

关于中亚和科技，根本没人听说过任何事情，对吧？

They even refused to publish it.

他们甚至拒绝刊登。

Then we changed it and said that we're actually one of the leading venture investors in Kazakhstan.

后来我们换了个说法，说我们其实是「哈萨克斯坦领先的风险投资人之一」。

And that worked.

这就走通了。

Kazakhstan was already on the map, professionals had heard of it - thanks to Kaspi, by the way.

也就是说，哈萨克斯坦已经在地图上了，圈内人是听说过的——顺便说一句，这要谢谢 Kaspi。

Right.

是这样。

Because Kaspi is the kind of case that, by and large, the whole world knows.

因为 Kaspi 这个案例，说到底全世界都知道。

Right.

是这样。

And now that Higgsfield has come along and, as I say, put Kazakhstan on the technology and venture map of the world, when you say Kazakhstan, everyone is immediately interested.

而现在有了 Higgsfield，我刚才说了，它把哈萨克斯坦放上了全球科技和风投的地图，现在你只要说哈萨克斯坦，所有人立刻就来兴趣。

I can even see it myself - on LinkedIn I used to get, I don't know, maybe 10 connection requests a week, and now I get 150.

我自己都看得出来：我 LinkedIn 上以前一周大概来，我不知道，也许十个人加我，现在是一百五十个。

That's because of Higgsfield.

这就是因为 Higgsfield。

It's mostly foreigners reaching out, they all want to meet, they all ask, "So what's going on over there?" GITEX? is coming here, GITEX? is coming.

来的主要是外国人，都想认识一下，都在问「你们那边到底在发生什么？」GITEX? 要来这里办，GITEX? 也在往这儿来。

Right.

是这样。

And of course thanks to Astana Hub, but also thanks to Higgsfield, they see that something is happening here, and they want to be a part of it.

当然，这要归功于 Astana Hub，但同样也归功于 Higgsfield——他们看到这里正在发生一些事情，自己也想参与进来。

Yes, that's a very good point you've made.

对，您提的这个点非常好。

Friends, take advantage of it.

朋友们，好好利用这一点。

From now on, founders from Kazakhstan are going to get a lot more attention.

从现在开始，哈萨克斯坦的创始人会得到多得多的关注。

Another commercial break.

又到广告时间。

We have a lot of programs.

我们有非常多的项目。

We send people to Silicon Valley.

我们会把人送到硅谷去。

Those are the Hero Training and Alchemist programs.

比如 Hero Training、Alchemist 这些项目。

So send in your applications, definitely take part.

所以一定要投申请，一定要参加。

All right.

好。

Yerzat, tell us how things developed from there.

Yerzat，接着讲讲后面事情是怎么发展的。

From my own experience I know - and a lot of the founders here know it too - that a first investment round, even one as big as 8 million dollars, is far from a guarantee of success.

我自己亲身经历过，在座很多创始人也都清楚：第一轮融资，哪怕是 800 万美元这个量级，也远远不等于成功。

You yourself mentioned that you were sort of underdogs.

你自己刚才也说了，你们当时算是不被看好的那一方。

So how did you manage to get a foothold in the market after that?

那后来是怎么在市场上站住脚的？

Where, let's say, did you find your first market fit?

你们最早的产品市场契合点（PMF）是在哪儿找到的？

Yes, when we raised that first eight million, we went straight out and signed a contract with Google for something like 5 million, purely for GPUs, for compute.

是的，我们拿到第一笔 800 万之后，马上就去跟 Google 签了一份大概 500 万的合同，纯粹买显卡、买算力。

And that was a risky move too.

这一步其实也挺冒险的。

That was a big chunk of the funding, and we signed it right away with Google, with Google Cloud.

这是融来的钱里的一大块，我们当场就跟 Google、跟 Google Cloud 签了。

And our first strength was technical expertise - the fact that we had put together a pretty strong ML team.

我们的第一个强项是技术能力——我们攒起了一支相当强的 ML 团队。

I remember how Alex and I went around together looking for people on the market, and we assembled the core of Higgsfield - they were all engineers plus a few strong product engineers.

我记得当时我和 Alex 两个人到处跑，在市场上找人，凑出了 Higgsfield 的核心班底，全是工程师，外加几个很强的产品工程师。

We managed to put together a strong team.

确实把一支强队搭起来了。

And we had this confidence that technologically we were no worse than the American teams at understanding how the technology works.

我们当时有一种底气：论技术，论对这套技术怎么跑通的理解，我们不比美国团队差。

And so we started building the technology.

于是我们就开始写这套技术了。

And jumping ahead a bit, on the subject of how recognizable Kazakhstan is now, I'd also like to point out that in the end Higgsfield became a fairly complex product, on top of the infrastructure that we started building first and thanks to which we managed to build the product efficiently.

顺便往前说一点，讲到现在哈萨克斯坦的知名度——我还想说，除了我们最先动手搭的那套基础设施（正是靠它，我们才把产品做得这么高效）之外，Higgsfield 最终也长成了一个相当复杂的产品。

And after that we managed to make a great product and build great distribution.

再往后，我们既做出了很棒的产品，也搭出了很棒的分发。

And right now Higgsfield has a pretty good reputation in the Valley.

现在 Higgsfield 在硅谷的口碑相当不错。

Practically all the top funds know us, all the strong teams know us.

几乎所有头部基金都知道我们，那些很强的团队也都知道我们。

And recently some people I know who went through the Andreessen Horowitz Speedrun accelerator - it's a new accelerator that's considered even cooler than Y Combinator - they told me that at Speedrun they study Higgsfield's playbook for distribution.

最近有几个朋友——他们进过 Andreessen Horowitz 的 Speedrun 加速器，这是个新加速器，被认为比 Y Combinator 还要牛——他们跟我说，Speedrun 里在研究 Higgsfield 的分发打法（playbook）。

And personally it was a huge honor for me to learn that what we built here in Almaty - Higgsfield's distribution pipeline - is now being taught at the best accelerator in the world, to the best teams, to those very graduates from Stanford and MIT, as the right playbook for how to launch a GenAI startup.

对我个人来说，知道这件事是莫大的荣幸：我们在阿拉木图做出来的东西，也就是 Higgsfield 的分发流程，如今在全世界最好的加速器里，被当作启动 GenAI 创业公司的正确打法，讲给最好的团队、讲给那些斯坦福和 MIT 的毕业生听。

Very cool.

太酷了。

And they study their own mistake too - how they failed to invest in Higgsfield.

顺带也在复盘他们自己的失误——当初怎么就没投 Higgsfield。

Yes, the situation with Andreessen was funny, because they had invested in something like five of our direct competitors.

是啊，Andreessen 那边的情况挺逗的，因为他们投了差不多五家我们的直接竞争对手。

So they tried five times, from a different angle, to capture the market.

也就是说，他们换了个角度，前后试了五次，想把这个市场拿下。

In the end all of those companies, our direct competitors, are companies like Captions and Krea AI.

结果这些公司，我们的直接对手，都是像 Captions、Krea AI 这样的公司。

If anyone here works in AI, they know these companies well.

在座如果有人做 AI，对这几家公司肯定不陌生。

And we beat all of those companies and took market share away from them.

而这些公司我们全都打赢了，把它们的市场份额抢了过来。

And now Andreessen is still trying to invest in us, but of course at a different valuation now, on different terms.

到现在 Andreessen 还在试着投我们，只不过当然是换了估值、换了条件。

Great.

厉害。

Well, continuing on the topic of product-market fit, I'd like to dig into that story a bit as well.

那接着 PMF 这个话题往下聊，我也想把这段经历稍微挖一挖。

I remember you had different products - Hera?, I think, a mobile app with endless TV series.

我记得你们做过好几个产品——好像有个叫 Hera? 的，是个能刷无限连续剧的手机 App。

And then the HXF mobile app, where you could also generate clips of some kind.

后来又有个叫 HXF 的手机 App，也能生成一些短视频。

I remember we had a Zoom call, and you even ran customer development on me: "Magzhan, we're curious, what business models do you think could take off in the market?"

我记得我们开过一次 Zoom，你们甚至还拿我做了一轮用户访谈：「Magzhan，我们挺好奇，你觉得市场上哪些商业模式能跑起来？」

For me that was something new too.

对我来说这也是个新鲜事。

Back then I was analyzing it together with you, so to speak - just generating ideas.

当时我算是跟你们一起分析吧——说白了就是一起出主意。

But again, by moving along the methodology, testing different hypotheses, we can see that you did manage to find your model.

但话说回来，你们按方法论推进，不断验证不同假设，我们现在也看到了：你们确实找到了自己的模式。

So tell us in more detail, please, how you arrived at the model that works in the market today.

所以请你详细讲讲，你们是怎么走到今天这个在市场上跑得通的模式的。

Yes, this is a really important question.

对，这其实是个非常重要的问题。

It's about the fact that successful startups are often just survivorship bias, because the startup simply got lucky - it was in the right place at the right time and took off.

它说的是：成功的创业公司往往只是幸存者偏差——因为它就是运气好，恰好在对的时间出现在对的地方，然后就起飞了。

And it really does often work that way, that a lot of startups just got lucky.

而且这确实经常成立，很多创业公司说到底就是运气好。

But we're on the other side of that - we didn't want to play that survivorship-bias game and count on luck.

但我们站在另一边——我们不想去赌这种幸存者偏差，不想指望运气。

So at Higgsfield we developed an approach where we always try to maximize success.

所以在 Higgsfield，我们形成了一套做法：永远尽全力把成功放到最大。

That's why at the beginning, when we were raising, we did more than 100 pitches.

所以最开始融资的时候，我们做了 100 多场 pitch。

Then, when we were building the product, we did an enormous number of product iterations.

再往后做产品的时候，我们做了海量的产品迭代。

Even now we ship at least two major new products every week, and we do small releases every day.

到今天我们每周至少还要发两个大的新产品，每天还有小版本上线。

Because we understand that we found product-market fit precisely thanks to that enormous number of iterations.

因为我们明白：现在能找到 PMF，靠的正是这么大量的迭代。

It would probably have been impossible not to find product-market fit with that many iterations.

迭代到这个量级，大概不可能找不到 PMF。

As long as you take feedback from the market properly - and even now we're aware that some new young startup could disrupt us, because everything changes, the market changes, and we don't want that.

前提是能实实在在地接住市场的反馈——而且到今天我们也清楚，随时可能冒出个年轻的新公司把我们颠覆掉，因为一切都在变，市场在变，而我们不希望那样。

So we still operate with that same approach.

所以我们到现在还是这套打法。

We still ship two new products a week.

到现在还是每周发两个新产品。

That's our Higgsfield approach, if you like.

这算是我们 Higgsfield 的方式吧。

And a lot of people in the Valley already know this approach.

这套做法，硅谷现在也有很多人知道了。

They know it - a lot of investors and founders in the Valley have heard about the company's work ethic, about how we ship products and how we build distribution.

他们是知道的——硅谷相当多的投资人、创始人都听说过这家公司的工作作风，听说过我们怎么发产品、怎么做分发。

And obviously that reflects really well on Kazakhstan, having that kind of reputation.

这显然对哈萨克斯坦是件大好事——能有这样的口碑。

I remember some very serious Chinese businessmen, from a very big Chinese company, coming up to Alex in the Valley and saying they'd heard that some insanely strong team had come together in Almaty - and that even they, since they're our direct competitors, were a little bit scared to compete with us.

我记得有几位非常有分量的中国商人，来自一家很大的中国公司，在硅谷特地找到 Alex，说他们听说阿拉木图冒出了一支强得离谱的团队，连他们——因为他们是我们的直接对手——都有点怵跟我们竞争。

So there you go.

就这样。

And all of this comes from the approach we developed: don't count on luck, pull yourself up by your own bootstraps.

这一切都来自我们摸索出来的这套做法：不指望运气，而是自己拽着自己的头发把自己拔出泥潭。

That's why every single department of ours works this way.

所以我们每个部门都是这么干的。

So if it's product, then instead of one prototype the developers make 100 prototypes a month.

比如产品这块，开发不是做一个原型，而是一个月做 100 个原型。

Our marketers, instead of one marketing campaign, one narrative, make 100 narratives, 100 marketing campaigns.

我们的市场同事也一样，不是做一个营销活动、一套叙事，而是做 100 套叙事、100 个营销活动。

Right now we're actively working with big media agencies in the US, with major celebrities.

我们现在正跟美国的大型媒介代理公司、跟顶级明星密切合作。

And we use the same approach there.

用的还是同一套做法。

If we're trying to sign some big Hollywood star right now and the first answer we get is no - a lot of people in Hollywood are afraid of AI - we don't give up, we keep going.

比如我们现在想签某位好莱坞大牌，第一次被拒了——好莱坞很多人怕 AI——我们不会就此罢手，而是继续往下推。

We go and try to sign a different Hollywood star.

我们就去签另一位好莱坞明星。

And as a result we now have some very serious Hollywood stars in the pipeline who - you'll see in a while - have partnerships with Higgsfield.

结果现在我们的 pipeline 里已经有几位非常有分量的好莱坞明星，过段时间你们就会看到，他们现在跟 Higgsfield 有合作。

Yerzat, Keanu Reeves, please.

Yerzat，麻烦把 Keanu Reeves 签下来。

Murat, from the height of your experience, so to speak, do you think our Kazakh startups know how to work like this - by the methodology, testing hypotheses, moving lean, looking for their own market fit - or do we still need to pull up these business skills?

Murat，以您的经验来看，您觉得咱们哈萨克斯坦的创业公司懂不懂这么干——按方法论走、验证假设、用精益的方式推进、去找自己的 PMF——还是说这些商业能力我们还得再补一补？

I think that by and large they do know how to work, but Higgsfield has set a completely different bar.

我觉得基本的做法大家是会的，但 Higgsfield 把标准整个拉到了另一个高度。

How fast - and from my conversations with Yerzat, from what I understand and what I see today, beyond the product itself the two key things are distribution and speed, those are the key factors.

速度有多快——从我跟 Yerzat 的几次交流，从我今天看到和理解的东西来看，除了产品本身，两件关键的事就是分发和速度，这才是决定性的因素。

Meaning it has to be done differently from how it goes here - we come up with one idea, then we spend 3 months shipping it, and after 3 months we collect feedback and think about what to do with it next, right?

也就是说，不能再像我们这边那样：想出一个点子，然后花三个月把它做上线，三个月后再收反馈，再想下一步怎么办，对吧？

These days all of this is done in literally days.

这些事现在真的是几天之内就做完了。

So you have to do it as fast as possible.

所以要尽可能快。

Whoever moves faster is the one who wins.

谁做得更快，谁就赢。

That's how it works right now. Yes, I agree with you.

现在的情况就是这样。是的，我同意你的说法。

This is probably a time that demands completely different speeds.

现在大概就是这么个时代，需要的是完全不同量级的速度。

Everything has to be done very fast.

也就是说，什么事都得做得非常快。

And moving on, if we go by the startup business-model playbook — Yerzat, I'm personally very curious how your unit economics came together.

那接着往下说，如果按创业公司商业模式那套教科书来走——Yerzat，我个人特别想知道，你们的单位经济模型是怎么算平的。

Your business is B2C, customer acquisition is very expensive, and on top of that you have huge costs — salaries, and a lot of money goes into GPUs.

你们做的是 B2C，获客成本非常高，同时开销也很大，工资，还有大量的钱花在 GPU 上。

I remember that early on it even got to the point where you were burning a million dollars a month on GPUs.

我记得早期甚至到过每个月在 GPU 上烧掉一百万美元的地步。

That did happen, right?

有过这回事吧？

Yes, that happened.

是，有过。

Back when we weren't profitable yet, we did get to that number.

那会儿我们还没盈利，确实到过这个数。

That was for training the model, the thing that— Yes.

这是花在训练模型上的，就是那个——是的。

Before you can sell a model, you have to train it.

在把模型卖出去之前，得先把它训出来。

And when we launched — March 31st was the launch of our flagship product, Higgsfield AI.

我们上线的时候——3 月 31 日是我们旗舰产品 Higgsfield AI 的发布日。

And that was our own model, which we trained on our own GPUs.

那是我们自己的模型，在我们自己的 GPU 上训出来的。

And because it was our own model running on our own infrastructure, we had a very high gross margin — thanks to our efficiency, our lean approach, and thanks to that high gross margin.

正因为是我们自己的模型跑在我们自己的基础设施上，我们的毛利率非常高——靠的是效率、精益的做法，也正是靠这个高毛利。

First of all, in the GenAI market, gross margin is a painful topic right now, because in classic SaaS a startup's gross margin was supposed to be 90-plus percent.

首先，在 GenAI 这个市场，毛利率现在是个挺棘手的话题，因为过去在经典的 SaaS 里，一家创业公司的毛利率被认为应该在 90% 以上。

But in GenAI that's not the case, because you need GPUs, you need compute, and it's very expensive.

但在 GenAI 不是这样，因为你需要 GPU，需要算力，而算力非常贵。

So gross margin is a real problem.

所以毛利率是个实打实的问题。

For example, just recently — literally the day before yesterday, I think — Dario Amodei, the CEO of Anthropic, said they're going to aggressively take market share from Google and OpenAI and lower their gross margin, meaning they'll dump prices.

比如说，最近，我记得就在前天，Anthropic 的 CEO Dario Amodei 说，他们要从 Google 和 OpenAI 手里激进地抢市场，会主动压低自己的毛利率，也就是打价格战。

And because of that, it's genuinely very hard to make the unit economics work in GenAI.

正因为这个，在 GenAI 里想把单位经济模型算平真的非常难。

We got very lucky, because we're extremely efficient and we have talented engineers.

我们很幸运，因为我们效率极高，工程师也都很有才华。

From the very start we shipped a product with a high gross margin.

我们一开始发布的产品就带着高毛利。

And since we already had a lot of experience working with data centers, we understood how to set up a company's economic planning properly.

而且我们本来就有大量跟数据中心打交道的经验，我们明白一家公司的经济规划该怎么搭才对。

And we're probably also one of the first on the market to start practicing high mobility.

我们大概也是市场上最早开始实践高机动性的公司之一。

Meaning we can move from one data center to another in less than 24 hours.

意思是我们可以在 24 小时之内从一个数据中心搬到另一个。

Our infrastructure is mature enough that we live on short-term reservations.

也就是说我们的基础设施成熟到可以只靠短期的算力预留过日子。

We don't sign long contracts with data centers, the way most of our competitors do.

我们不像大多数竞争对手那样跟数据中心签长约。

If you go the classic route and want to get compute, the data center will ask you to sign a long contract — on average they ask for three years, you can push it down to one year, but that's a big risk.

按常规路子走，你想拿算力，数据中心会要求你签长约，平均要签 3 年，能压到 1 年就不错了，但这个风险很大。

Signing contracts that long in GenAI, when everything is changing so fast.

在 GenAI 这种什么都变得飞快的领域，签这么长的合同。

And it's great that we learned to work fast, we learned to negotiate, and we sign contracts for two or three months at a pretty low price.

很棒的是，我们学会了快速行动，学会了谈判，我们签的是两三个月的合同，价格还相当低。

And thanks to our efficient infrastructure, if there are problems with one data center, we move to another in less than 24 hours.

而靠着这套高效的基础设施，某个数据中心一旦出问题，我们 24 小时内就搬到另一个。

And thanks to that lean approach, we've been able to make the unit economics work very efficiently.

正是靠这种精益的做法，我们把单位经济模型算得非常漂亮。

And then that negotiating skill worked really well for us when we started plugging in other models.

再往后，这套谈判的本事在我们开始接入其他模型时也发挥得很好。

And what does "plugging in" mean?

接入是什么意思呢？

For example, Google has Nano Banana, a powerful new model.

比如说，Google 有 Nano Banana，一个很强的新模型。

And to plug it in — we understand what's going on on Google's side.

要接入它——我们清楚 Google 那一头在发生什么。

They have GPUs too, and you don't pay per API call like with a classic GPT API; at our volumes we reserve GPUs from Google.

他们同样有 GPU，而你付的不是像经典的 GPT API 那样按 API 调用付费；以我们的用量，我们是直接向 Google 预留 GPU。

So we go to Google and start negotiating with them not over the model, but — we understand that behind it they have TPUs and GPUs, so we're already talking about reserving their compute.

也就是说，我们去找 Google，跟他们谈的不是模型本身——我们明白背后是他们的 TPU 和 GPU，所以我们直接谈的是预留他们的算力。

And because we understand the market and how it's structured so well, we manage to squeeze better terms out of the other model providers.

正因为我们对市场和市场结构理解得足够透，我们才能从其他模型供应商那里争取到更好的条件。

And because of that, we're one of the highest-margin businesses in our category.

也正因为这个，我们是同类里毛利最高的生意之一。

And that's why our unit economics are great.

所以我们的单位经济模型才这么漂亮。

By the way, we also have a very strong partnerships team.

顺便说一句，我们还有一支非常强的 partnership 团队。

They're our fellow Kazakhstanis too.

他们同样是我们哈萨克斯坦本地人。

We have a very strong head of partnerships.

我们有一位非常强的 head of partnerships。

A young guy, only 25 years old, he graduated from Lyceum No. 165 in Almaty.

很年轻的小伙子，才 25 岁，在阿拉木图读的第 165 中学。

And he's gotten so good at negotiating that he does it at the level of a VP at one of the biggest corporations.

他把谈判这件事练到了什么程度呢——他能拿出顶级大公司副总裁的水准去谈。

He runs negotiations at the level of a Google Cloud vice president — at the very highest level — and he's only 25.

也就是说，他谈判的对位是 Google Cloud 的副总裁，也就是最高层级，而他才 25 岁。

He actually joined Higgsfield to work on a completely different track.

他当初进 Higgsfield 做的完全是另一条线的事。

But it really shows how talented the specialists we have in Kazakhstan are — that guys sitting in Almaty on calls can close deals at that level, deals worth tens of millions of dollars.

但这确实说明哈萨克斯坦的专业人才有多强——小伙子们坐在阿拉木图打电话，就能在这个层级上谈成几千万美元的单子。

I'd like to add a small comment about speed here.

关于速度，我还想再补一句。

If we look at how it used to be — and by "used to be" I don't mean last century, just two years ago — the average age of a unicorn, the time it took to become one, was 10 years.

以前是什么样呢——我说的以前不是上个世纪，就是两年前——一家公司长成独角兽的平均年限是 10 年。

And you just remembered the date when they launched the product — March 31st of last year.

而现在，你刚才还记得他们上线的日期，去年 3 月 31 日发布的产品。

It hasn't even been a year, and they're at 200 million.

一年都还没到，他们已经做到 2 亿。

Their valuation today is already approaching 4 billion.

估值今天已经逼近 40 亿了。

I wanted to save those numbers for dessert.

这些数字我本来想留到最后当甜点的。

Sorry.

抱歉。

Super, great.

太好了，很棒。

So, yes, Yerzat has been talking a lot about how many of your achievements, your business results, were built thanks to your team.

是的，刚才 Yerzat 讲了很多，说你们的很多成绩、很多业务上的成果，都是靠你们的团队做出来的。

And a lot of your people are still our local guys.

而你们的很多人其实都是我们本地的人。

It's not like you're hiring expats from Silicon Valley or anything.

并不是说你们从硅谷之类的地方招外派专家。

Tell us a bit more about the team, please.

麻烦你再多讲讲团队。

How many people do you have now?

你们现在有多少人？

Who are you hiring?

你们在招哪些人？

What kind of talent?

招什么样的人才？

Yes, we now have more than 100 people on the team, around 130.

是的，我们团队现在有 100 多人，大概 130 个。

Of those, only the back office is in the US.

其中只有后台职能在美国。

That's our head of finance, our lawyer, and Alex, the co-founder — they're in the US, and our Chief Strategy Officer is in the US too; so everyone who works directly on sales, on the product, on R&D, they're all here in Kazakhstan.

就是我们的财务负责人、我们的律师，还有联合创始人 Alex，他们在美国，我们的 Chief Strategy Officer 也在美国；而直接做销售、做产品、做 R&D 的人，全都在这儿，在哈萨克斯坦。

And this past December our Americans came to Almaty.

就在 12 月，我们的美国同事来过阿拉木图。

They're really great specialists too.

他们也都是非常出色的专家。

Our general counsel graduated from Harvard, and our CFO graduated from Stanford.

我们的总法律顾问是哈佛毕业的，CFO 是斯坦福毕业的。

There they all are.

他们全都在这儿。

Our Chief Strategy Officer — he's honestly a legendary entrepreneur out in the Valley.

我们的 Chief Strategy Officer，他在硅谷本身就是个传奇创业者。

He actually bought Elon Musk's very first company.

他当年还收购过 Elon Musk 的第一家公司。

He was CEO of several decacorns, he was CEO of Opera.

他当过好几家十角兽公司的 CEO，也当过 Opera 的 CEO。

They all came out to Almaty.

他们全都来过阿拉木图。

We all worked together, and they were shocked by our pace.

我们一起干活，他们被我们的节奏惊到了。

They'd just come up to me and say they were stunned by what we'd managed to build, because they had never seen that kind of pace, that kind of speed.

他们就直接走过来跟我说，我们做出来的东西让他们特别震惊，因为他们从没见过这种节奏、这种速度。

Yeah.

嗯。

It was a great experience, all of us working together like that.

大家一起工作的那段经历特别棒。

And we really did manage to put together a very strong team.

而且确实，我们真的攒出了一支非常强的团队。

Practically all of our people — all the key employees, the engineers, the marketers, the B2B sales people — they all hold options in the company.

我们几乎所有员工——所有核心员工、工程师、市场、B2B 销售——都拿着公司期权。

That was our first approach too, back when Alex and I were just starting to work together.

这也是我们最早的做法，那会儿我刚跟 Alex 开始合作。

The first thing we started doing was studying how to properly grant equity to Kazakhstani employees — American-style options — so that everything would be transparent and clean.

我们做的第一件事，就是研究怎么合规地给哈萨克斯坦员工发股权、发美式期权，让整件事透明、干净。

And it really paid off, because our people are now top of the market, and all our competitors and the big corporations that see us as a potential future rival or want to buy us are constantly trying to poach our people — the engineers, the software engineers, our marketers, our B2B guys.

这一步确实走得非常对，因为现在我们的员工在市场上是顶尖的；所有竞争对手，还有那些把我们看成未来潜在对手、或者想收购我们的大公司，都在不停挖我们的人——工程师、软件工程师、市场的人、B2B 的人。

And what protects us here is exactly that: all of our employees are shareholders.

而保护我们的恰恰就是这一点：所有员工都是股东。

A lot of them have already vested millions of dollars in company stock.

不少人手里已经归属了数百万美元的公司股票。

So the whole team is super loyal.

所以整个团队超级忠诚。

And as we can see, companies in the Valley don't care whether the guys are in Kazakhstan or not.

而且我们看到，硅谷的公司根本不在乎人是在哈萨克斯坦还是不在。

They understand that these people have unique experience.

他们明白，这些人有独特的经验。

That experience is very expensive on the market, and in the Valley people like that are very expensive to find.

这种经验在市场上非常贵，在硅谷要找到这样的人成本极高。

So all the top startups in the Valley, our direct competitors, are constantly trying to poach our employees.

所以硅谷所有顶尖创业公司，也就是我们的直接竞争对手，都在不停地挖我们的人。

And that's another one of our strengths — the team has a real sense of cohesion, a sense of patriotism, that we want to build something really big.

这也是我们的一大强项——团队里有一种很强的凝聚感、一种爱国心，我们想做一件真正大的事。

And that's a competitive advantage for us too, because in the Valley there are a huge number of startups, and there's this trend where if a startup starts doing even slightly badly, the founders can walk into the office the next day and there are no employees there, because they've simply moved to another startup.

这同样是我们的竞争优势，因为硅谷创业公司太多了，那边有种风气：一家创业公司只要稍微不顺，第二天创始人走进办公室，人可能就没了，因为大家直接跳去了别的创业公司。

And turnover is a huge problem at every startup in the Valley.

硅谷每一家创业公司的人员流动问题都非常严重。

Maybe those of you who follow this closely saw that Anthropic's lead engineer moved to Cursor, worked at Cursor for a few days, didn't like it, and went straight back to Anthropic.

可能有人一直密切关注这个圈子，看到过 Anthropic 的首席工程师跳去 Cursor，在 Cursor 干了几天，觉得不合适，又回了 Anthropic。

Same thing at OpenAI — OpenAI employees left for Thinking Machines, Mira Murati's new company.

OpenAI 也一样——OpenAI 的员工去了 Thinking Machines，Mira Murati 的新公司。

Some of them were even cofounders of that company.

其中有几个人还是那家公司的联合创始人。

They didn't like it, they worked there a few months, and now they're back at OpenAI.

他们干了几个月，觉得不合适，现在又回了 OpenAI。

And that kind of rotation happens at every level in the Valley.

这种人员轮转，在硅谷是各个层级都在发生的。

So the fact that in Kazakhstan we managed to assemble such a tight-knit team and show how operationally effective it is — venture investors know all these problems well, which is why our region is now becoming interesting to them.

而我们在哈萨克斯坦能攒出这么有凝聚力的一支团队，还把这支团队的运营效率做出来了——风险投资人对这些问题门儿清，所以现在开始对我们这个区域感兴趣。

They understand that access to talent is worth an enormous amount.

他们明白，能拿到人才这件事本身就值很多钱。

For example, in our last round Accel invested in us.

比如上一轮，Accel 投了我们。

And I was the one pitching the Accel partner, telling him about Kazakhstan; he found it very interesting, and what really hooked him was precisely that a new access to talent had opened up.

当时正是我给 Accel 的合伙人做的 pitch，跟他讲哈萨克斯坦，他非常感兴趣，真正打动他的就是：多出了一条拿到人才的新通道。

And that last Series A, the $130 million, we raised specifically to scale the team.

最近这轮 1.3 亿美元的 A 轮，我们融来就是专门用于扩团队的。

We're hiring very aggressively right now.

我们现在招人非常猛。

And we don't hire on the principle that we have some urgent open role and we want to close it by finding the right person.

而且我们招人的逻辑，不是手上有什么急着要填的坑、然后去找个合适的人补上。

We hire on the principle that if we know there's a strong specialist out there — it doesn't matter whether they're strong in business, strong in tech, strong at communication or strong in sales — we want to hire them, and the tasks will show up for them later, we'll find a use for them later.

我们现在的逻辑是：只要知道哪里有个很强的人，不管他强在业务、强在技术、强在表达还是强在销售，我们就想把他招进来，事情之后自然会有，之后自然会给他找到用武之地。

Right now our goal is simply to hire every A-player in Kazakhstan and Central Asia.

我们现在的目标，就是把哈萨克斯坦和中亚的 A-player 全招过来。

And beyond that, just all of the strongest specialists, because, as I said earlier, we have a 25-year-old who's now closing deals worth tens of millions of dollars.

再往下就是把最强的人都招进来，因为像我刚才说的，我们有个 25 岁的员工，现在在谈几千万美元的单子。

We have students who haven't even finished university managing millions of dollars of budget a month on media, on our distribution.

我们还有大学都没毕业的学生，每个月管着几百万美元的预算，投在媒体上、投在我们的分发上。

And our ML engineers, who only recently graduated from Nazarbayev University, are now working on clusters that also cost millions of dollars.

我们的 ML 工程师，刚从纳扎尔巴耶夫大学毕业没多久，现在跑的集群也是几百万美元级别的。

We can see that our specialists really are excellent, they have high potential, and this talent is still underrated.

我们看到，我们的人真的很强，潜力很高，而且这批人才现在还被市场低估。

And because of that, I think Higgsfield will be able to scale a lot further.

所以我觉得，靠这一点 Higgsfield 还能再放大很多。

My dream is to scale to 500 people in Kazakhstan, to show that you can build a 500-person company of super-highly-qualified specialists and still not slow down, still ship great, innovative products.

我的梦想是在哈萨克斯坦扩到 500 人，证明你可以做出一家 500 人、全是超高水平专家的公司，而且速度不会慢下来，照样能持续发出很棒的创新产品。

Excellent.

太好了。

This is a very good example of the hypothesis we've been discussing for years — that you can work from here, hiring world-class specialists while carrying far lower costs, because launching that kind of business here is much cheaper, and yet still work for the whole world.

这正好印证了我们讨论了很多年的那个假设：你可以在这里做事，招世界级的人才，同时成本低得多——在这里启动这样一门生意的开销要低很多——但服务的却是全世界。

A great example.

非常好的例子。

Thank you.

谢谢。

Make?, you said the average age on the Higgsfield team is, what, 26?

Make?，你说过 Higgsfield 团队的平均年龄好像是 26 岁？

Yes — 25.

对——25。

25.

25。

So I take it they wouldn't hire you and me — unfortunately.

那我理解，咱们俩是进不去了——可惜。

No, Magzhan, we'll have to come up with something else.

没戏，Magzhan，咱俩得另想辙。

Ah, all right.

啊，好。

And I want to keep going on the go-to-market topic.

我还想接着聊 GTM 这个话题。

Honestly, I'd really like to know — and I think a lot of people here would too — how you attracted your first users.

说实话，我特别想知道——我想在座不少人也想知道——你们最早的用户是怎么来的。

Say, the first 10,000 users — how did you get them?

比如说最早的 1 万用户，你们是怎么拉来的？

Yeah, that's a great question.

对，这是个特别好的问题。

Because as the company started to grow — from the release onward, basically — whereas before that I was only working on the technology, since there wasn't a product yet.

因为公司开始长起来之后——基本上就是从发布那一刻起——而在那之前我只做技术，因为还没有产品。

Now I spend 90% of my time on GTM, and we've come up with and iterated through a lot of different frameworks.

现在我 90% 的时间都花在 GTM 上，我们想了很多套框架，一轮轮迭代过来。

But the very first one — what really decided it was the product, because we iterated through a huge number of products.

但最开始那一条，说到底还是产品定胜负，因为我们迭代过特别多的产品。

And by the time we'd iterated our way to the Higgsfield we launched on March 31st, there had been several iterations.

等我们一路迭代到 3 月 31 号发布的那版 Higgsfield，中间已经过了好几轮迭代。

We built the first version a year before that, and it wasn't as good, because the technology hadn't matured yet.

第一版其实早一年就做出来了，但质量没那么好，因为技术还没成熟。

And because we had a very strong team, we were able to iterate all the way to this final product.

正因为团队非常强，我们才有能力一路迭代到最后这个产品。

And when it came together, when I personally saw the generations our models were capable of producing, it was obvious to me that it would be an instant hit.

等它真的做出来，等我亲眼看到我们的模型能生成什么效果，我立刻就明白它一定会火。

I remember the feeling I had then — I was 100% certain that we'd release it and it would take off right away.

我记得当时那种感觉——我百分之百确定，我们一发出去，它马上就会起飞。

It would go viral, it would fly across the internet, because we already had the expertise, we understood how the market works, and we had complete confidence in the product.

它会传开，会在互联网上疯传，因为我们已经有相应的经验积累，懂这个市场怎么转，对产品也有百分之百的信心。

But to get to that point, we built an enormous number of other products and took an enormous number of knocks and made a lot of mistakes.

但为了走到那一步，我们做过巨量的其他产品，撞过巨量的墙，踩过无数的坑。

And when we'd built this product, we knew 100% that it would go viral and get huge usage.

等这个产品做出来的时候，我们百分之百确定它会病毒式传播，会有很大的使用量。

But then came a new challenge: holding on to that initial momentum.

但接下来就是新的挑战：怎么守住最开始这股势头。

And at that point we already needed to build GTM frameworks that would let us do more than just stay afloat, because these days a startup's product often goes viral and then after a while everyone forgets about it — it just can't sustain that momentum.

那时候我们就得搭出一套 GTM 框架，让我们不只是勉强活着——因为现在创业公司常常是产品一夜爆火，过一阵子所有人就把它忘了，那股势头根本接不住。

And we understood that very well internally.

这一点我们内部看得很清楚。

So we started developing our own GTM framework.

于是我们开始做自己的 GTM 框架。

And here as well — sorry, could you explain what a GTM framework is, please?

这里也一样——抱歉打断一下，能先讲讲什么是 GTM 框架吗？

Well, it's a go-to-market framework — in simple terms, right, it's a list of rules, laws, approaches that you have to follow so that in the end your product gets big reach, so that a large number of people hear about it, get interested in it, start using it, and then it has retention, so people keep using the product for a long time after that.

go-to-market 框架，说白了就是一份规则、法则、打法的清单，你得照着它走，最后产品才能拿到高曝光，让大量的人知道它、对它感兴趣、开始用；再往后还要有留存，让大家能长期用下去。

Right.

就是这样。

And there are a whole lot of these go-to-market frameworks.

go-to-market 框架其实特别多。

For example, there's this accelerator, Y Combinator.

比如有个加速器叫 Y Combinator。

A lot of people here probably know it, anyone who follows startups.

在座关注创业的人，应该不少都知道。

They popularize a lot of different GTM frameworks.

他们推广过很多不同的 GTM 框架。

And I was very skeptical about all of them, because the way I saw it, there are something like 2,000 of these Y Combinator startups — well, Y Combinator turns out many hundreds of startups a year.

但我对这些都非常怀疑，因为我当时觉得，Y Combinator 出来的创业公司有两千家上下——一年就要批量产出好几百家。

And it's unlikely they can teach you anything useful when they're producing such a huge number of startups.

一年量产这么多公司，他们大概教不出什么真正有用的东西。

And obviously most of those startups aren't particularly high-quality ones — they don't develop any further after that.

而且很明显，这里面大多数公司质量一般，后面也没能再往前跑。

And at the time I spent a long while thinking about where to find a new approach, one that would set us up against that classic Y Combinator approach.

所以那阵子我一直在想，去哪儿找一套新打法，能跟 Y Combinator 那套经典路子对着来。

Because that Y Combinator approach — right now every startup out there believes in it blindly, because YC is such a popular label.

因为 YC 那一套，现在所有创业公司都在盲信，YC 这个标签太火了。

And what probably helped here is that as a kid I read a huge number of books on marketing.

这时候帮到我的，大概是小时候读过特别多营销方面的书。

One of my favorite books is Marketing Warfare.

我最喜欢的书之一是《Marketing Warfare》（营销战）。

It's a very old book, from the eighties, about how marketing was done back in those days, how they took products to market back then.

这是本很老的书，八十年代的，讲那个年代的人怎么做营销、怎么把产品推向市场。

And back then all those corporations like Coca-Cola, Pepsi weren't toothless the way they are now — they were very aggressive.

那时候 Coca-Cola、Pepsi 这些大公司可不像现在这么没牙，一个个都极其凶悍。

They were very aggressively taking market share away from each other.

他们非常凶地互相抢市场份额。

And there were a lot of approaches to marketing in there, how to release products the right way.

书里讲了大量营销打法，讲产品到底该怎么发。

There are concepts like guerrilla marketing, there's defensive marketing, there's the one where you attack from the flanks — a whole lot of frameworks.

里面有游击式营销，有防御战，还有从侧翼包抄那一路——框架特别多。

It's this forgotten knowledge that — well, if you tell someone in the Valley that you read books like that, they'll give you a very odd look.

这些算是被遗忘的知识，你在硅谷跟人说你读这种书，对方会用很奇怪的眼神看你。

Like, what is this ancient stuff?

「这什么老古董？」

And meanwhile I realized that all startups do exactly the same thing.

而我同时也看明白了，所有创业公司做的都是同一套。

Literally every startup is identical.

毫不夸张，每一家都长得一模一样。

That one playbook that everybody in Silicon Valley just copies from each other.

就是硅谷里大家互相照抄的那本 playbook。

And our framework, compared to that classic one, is very aggressive.

而我们的框架，跟那套经典打法比，攻击性非常强。

And it was very new when we did our first release.

我们第一次发布的时候，它非常新。

It turned out to be like a breath of fresh air among startup releases in the Valley.

在硅谷那一堆创业公司的发布里，它像一股新鲜空气。

And we also got lucky here, because Kazakhstan has a really strong creative scene.

这里我们还有一点运气：Kazakhstan 有非常强的创意圈子。

There are a lot of strong directors there, and editors, cinematographers.

有很多很强的导演，还有剪辑师、摄影师。

We have a strong school of that kind of creative work, music videos and so on.

我们在创意、音乐 MV 这块底子很扎实。

And we managed to adapt that strong Kazakh creative school to startup product launches.

我们把 Kazakhstan 这套很强的创意功底，搬到了创业公司的产品发布上。

We started hiring top Kazakh directors.

我们开始招 Kazakhstan 顶尖的导演。

We started hiring top Kazakh producers.

也开始招 Kazakhstan 顶尖的制片人。

And these are people with a completely different, fresh way of thinking.

这些人的思维完全不一样，很新鲜。

And they all started working together with our product engineers, with our engineers.

他们就跟我们的产品工程师、跟我们的工程师一起干活。

And what came out was a new kind of framework — super stylish, super youthful, super loud.

做出来的是一套新框架——特别时髦、特别年轻、特别吵。

And that's what our marketing is like — our first release video, the one we posted on Twitter, looked like a stylish music video, not like the classic video from a Silicon Valley startup.

我们的营销就是这个调性：第一条发布视频发在 Twitter 上，看着像一支很潮的音乐 MV，而不是硅谷创业公司那种标准发布片。

And it immediately started getting big reach organically.

它一上来就靠自然流量拿到了很高的曝光。

We didn't buy bloggers, we didn't buy traffic, we didn't have a high customer acquisition cost.

我们没买博主，没买流量，获客成本也不高。

We just made a cool video with cool, creative people, and everyone started loving it.

我们只是找了一群很棒的创意人，做了个很棒的片子，然后大家就都喜欢上了。

I mean, it was one hundred percent organic.

也就是说，百分之百是自然流量。

And after that we started scaling that process.

接下来我们把这个流程规模化。

We started hiring more creatives in Almaty.

我们在 Almaty 招了更多创意人。

We started hiring more advertising people in Almaty, and all together we simply started disrupting marketing in America.

又在 Almaty 招了更多做广告的人，然后大家一起，直接把美国的营销给颠覆了。

What startups in San Francisco do is very boring, it's very — I don't know, in my view it's primitive.

San Francisco 那些创业公司做的东西很无聊，说实话，在我看来挺原始的。

Right now, for example, we're buying up billboards in San Francisco, and we have a great creative director, Zee?.

比如我们现在在 San Francisco 包下了一批广告牌，我们有位很棒的创意总监 Zee?。

He came up with really cool graphic design that's going up on the billboards in San Francisco.

他做了一套很酷的平面设计，会上到 San Francisco 的广告牌上。

It's being posted right now.

现在正在投。

And it's a different approach — if you've been to San Francisco, you'll see very similar billboards there, all of them techy, all alike, and we do it differently.

这是另一种打法——去过 San Francisco 就知道，那儿的广告牌长得都差不多，全是技术味，彼此雷同，而我们反着来。

And it really worked.

这招确实奏效了。

And this is the thing Peter Thiel teaches everybody — the great venture investor, right — that you always have to be contrarian, you always have to understand how everyone else is doing it.

这也正是 Peter Thiel 教所有人的——那位了不起的风险投资人——永远要做逆向的那一个，永远要先搞清楚别人都是怎么做的。

And to stand out, you have to come up with something orthogonal.

想跳出来，就得想出一个正交的东西。

So you and I clearly read different books as kids, huh?

看来我俩小时候读的书不一样啊？

I thought Yerzat was a tech guy, but it turns out he was reading Marketing Warfare as a kid.

我一直以为 Yerzat 是个技术派，结果他小时候读的是《Marketing Warfare》。

Awesome, thank you so much.

太棒了，非常感谢。

Yeah, it was genuinely valuable to hear about your GTM frameworks.

听你们讲 GTM 框架，真的很有价值。

So, as I understand it, the main thing was virality.

那么按我的理解，核心就是病毒式传播。

Got it.

明白了。

You said the first thing, the main thing, is the product, right, so that you have high retention, good feedback and so on.

你说第一位、最关键的是产品，这样才有高留存、有好的用户反馈等等。

And the second thing, still, is virality.

第二位的，还是病毒式传播。

That's the growth hack you found to keep your unit economics at an optimal level.

这就是你们找到的那个增长黑客打法，用它把单位经济模型维持在最优水平。

Great.

很棒。

Okay, so you mentioned Elon Musk.

好，你刚提到了 Elon Musk。

Tell us, please — I know you were in talks with him.

请讲讲——我知道你们跟他谈过。

He was one of your, let's say, potential investors, potential backers.

他算是你们潜在的投资人、潜在的出资方之一。

You had direct meetings — how did you get on his radar?

你们直接见过面——你们是怎么进到他视野里的？

And all of us here, let's say the young folks and the not-so-young ones, would just love to hear that story.

在座的各位，年轻的和不那么年轻的，都特别想听听这个故事。

Yeah, he just saw our clip on Twitter, the one that went viral.

对，他就是在 Twitter 上刷到了我们那条爆火的视频。

We had a lot of clips there that started going viral, pulling in 10 to 20 million views each.

我们有不少视频在那上面开始疯传，一条能拿到一千万到两千万播放。

Twitter is a small social network, pretty compact, so if you go viral there and rack up several tens of millions of views, that most likely means practically every Twitter user saw it, including Elon Musk.

Twitter 是个不大的社交网络，挺紧凑的，所以你在上面火了、拿到好几千万播放，基本上就意味着 Twitter 上几乎所有人都刷到了，包括 Elon Musk。

And we saw right away that he was liking things, leaving comments, and then his people reached out to us directly, and obviously the interest was about M&A.

我们当时就看到他在点赞、留言，后来他的人直接找上了我们，来意很明显，是冲着收购（M&A）来的。

That surprised us a lot, of course, because we had only released the product on March 31st, and Elon reached out somewhere in early May, I think, yeah, it was early May, that's when he got in touch with us.

这当然让我们非常意外，因为产品是 3 月 31 日才发布的，而 Elon 大概在 5 月初——对，就是 5 月初——就联系我们了。

The product had only just started to pick up steam.

也就是说，产品那时才刚开始起量。

On the one hand it was fun, it was cool that we'd only just released the product and we were already talking with Elon Musk.

一方面这挺有意思、挺酷的——产品才刚上线，我们就已经在跟 Elon Musk 打交道了。

But on the other hand it's a huge distraction, of course, because the whole team, everyone, obviously, if you're a startup, you have to think about growing the company every day, every hour, every minute.

但另一方面，这当然非常分散精力，因为整个团队、所有人——你要是一家创业公司，就得每天、每小时、每分钟都想着公司的增长。

When something is going on with Elon Musk, everyone gets distracted, everyone starts thinking about it, the excitement level is high across the whole team.

一旦跟 Elon Musk 扯上关系，大家的注意力就都被带跑了，都开始琢磨这件事，整个团队都特别兴奋。

So yeah, there were both pluses and minuses in it.

所以这件事既有好处，也有坏处。

The plus, of course, was that if Elon believed in us, then we must be moving in the right direction.

好处当然是，既然 Elon 都相信我们，那说明我们的方向没走错。

The minus was that it pulled our attention badly away from growing the company.

坏处是，它把我们的注意力从公司增长上拽走了太多。

I remember I asked Murat for advice, and Murat was categorically against even talking to Elon at all.

我记得我找 Murat 商量过，Murat 态度非常坚决，连跟 Elon 接触都反对。

I remember exactly how Murat put it: "Wait another 10 months and you'll be 10x."

我到现在还记得 Murat 的原话：“再等 10 个月，你们就是 10 倍。”

And that's exactly what happened.

后来真就这样了。

We actually grew even more than that.

我们涨得甚至比这还多。

And if you remember, I also advised you not to agree to it. I said, make sure you meet him, and invite him to Digital Bridge on top of that.

而且你要是还记得的话，我当时也劝你别答应。我说，人一定要见，顺便把他请到 Digital Bridge 来。

Yeah.

是的。

And we essentially made the right decision, because obviously, as I said earlier, products often take off and then can't sustain that momentum.

我们本质上是做了个正确的决定，因为像我前面说的，很多产品起飞之后就守不住那股势头。

And it's important to exit fast and smartly while you're still riding the hype.

所以要趁着还在风口上，快速、漂亮地把退出做掉。

And a lot of people took that view, though not Murat and not Magzhan.

很多人就是这么看的——Murat 和 Magzhan 不是。

But a lot of people were like, "You need to sell, you grew so fast, you're at the peak, sell, do an exit, it'll be great."

但很多人都在说：“你们得卖掉，涨得这么快，正在高点上，卖了吧，做个退出，多好。”

But back then we understood clearly, we could see what a strong team we'd managed to put together.

但我们当时看得很清楚，我们知道自己攒出了一支多强的团队。

And personally, I felt it would be a real shame — we'd managed to put together such a great team.

而且我个人觉得特别可惜，好不容易攒起这么棒的一支团队。

I understood that we could scale it, that we could hire a lot more people and spread this framework of ours, try out our GTM hypotheses, the ones that later turned out to be successful.

我知道这支团队还能再扩，可以再招很多人，把我们这套框架铺开，去验证那些 GTM 假设——后来这些假设确实跑通了。

So we made the decision, and in retrospect it really was the right one, and that product of ours didn't turn out to be a one-off story.

于是我们做了这个决定，回头看确实是对的，我们那个产品并不是昙花一现。

We figured out how distribution works in GenAI.

我们搞明白了 GenAI 里的分发是怎么运转的。

And now, once a week, the way we went viral back then, now we go viral like that on social media once a week.

而现在，当年那种爆火，我们每周都能在社交媒体上来一次。

Back then we went viral on Twitter, now we go viral on Instagram, we go viral on YouTube, on LinkedIn.

那次是在 Twitter 上火的，现在是在 Instagram 上火、在 YouTube 上火、在 LinkedIn 上火。

And we've built out a B2B pipeline.

我们还搭起了一条 B2B 的销售管线。

Our B2B sales are growing very fast.

我们的 B2B 销售增长非常快。

So we've figured all of this out, and now we've become a solid business.

也就是说，这些东西我们都摸透了，现在已经是一家站得住的公司了。

Well, now we roughly understand, let's say, what's behind today's success: it's the team, it's the founders, first and foremost leadership, and the entrepreneurship skills you've shown.

那我们现在大致明白了，今天这些成绩背后是什么——是团队，是创始人，首先是领导力，还有你们身上展现出来的创业素质。

And now I'd probably like to talk about your current metrics.

接下来，我想聊聊你们目前的这些指标。

Murat Almazbekovich, as the first investor who believed in Higgsfield, could you tell us in more detail about the current metrics, specifically from the point of view of investment attractiveness?

Murat Almazbekovich，作为第一个相信 Higgsfield 的投资人，请您从投资吸引力的角度，详细讲讲现在的这些指标。

Well, I don't know, I've never had a startup like this in my life, with metrics like these.

怎么说呢，我这辈子还没碰到过这样的创业公司，没见过这样的数据。

There's this idea, you know, that one project can carry the whole portfolio, right?

有那么一种说法，一个项目就能把整个被投组合拉起来，对吧？

And now I've run into it myself.

现在我自己碰上了。

It's a real thing.

这事是真的。

Before, we had exits that were hard-earned, 5x, 3x, and a 10x was already a jackpot, right?

以前我们的退出都是辛苦钱，5 倍、3 倍，能有 10 倍就已经是中大奖了，对吧？

That's how it was.

就是这么回事。

And now you count these multiples and it's already 100x, 200x, 300x, right?

现在再去算这个倍数，已经是 100 倍、200 倍、300 倍了，对吧？

It's a completely different story.

完全是另一回事了。

For any VC, stumbling onto a startup like this is a dream, of course, but what amazes me most is this.

对任何一家风投来说，能撞上这样的项目当然是梦寐以求，但最让我震撼的是下面这点。

These days a lot of people tell me too: "Murat, you should lock something in, just in case, this is GenAI, everything changes very fast, you should sell something."

现在也有很多人跟我说：“Murat，你多少得落袋为安，万一呢，这可是 GenAI，什么都变得飞快，你该卖掉一点。”

And I say: "Listen, how can you sell when their whole September was 80% growth month over month?"

我说：“你听我说，人家整个九月都是 80% 的月环比增长，这怎么卖？”

I mean, that was awesome, right?

这多带劲，对吧？

A billion in September, and by November it turns out to be 2 billion already, right?

九月还是 10 亿，到十一月一看，已经 20 亿了，对吧？

And now it's even more than that.

现在还要更高。

It's just something incredible.

这简直不可思议。

As for the metrics, let Yerzat tell you.

至于具体指标，还是让 Yerzat 来讲。

He's the one who achieved them, so he's the one to talk about them.

指标是他做出来的，该由他来说。

Yerzat, tell us about the metrics.

Yerzat，讲讲指标吧。

Metrics, right.

指标，好。

It's what we call a growth mindset.

这就是我们说的增长思维（growth mindset）。

It means everyone in the company, and we try to stick to this, we think about growth all the time.

意思是公司里所有人——我们一直努力守住这一点——时时刻刻都在想增长。

And you have to realize, I think I actually saw this on TikTok, that one week is 2% of the whole year.

而且要意识到一点，这个我好像还是在 TikTok 上看到的：一周就是一整年的 2%。

So if you give it your maximum for even one week, well, for every week of the year, that's 2%.

也就是说，哪怕你只在一周里全力以赴——或者说一年里的每一周都全力以赴——那也就是 2%。

And even doing 2% a week, thanks to compounding, that's gigantic, you can get several dozen times the growth.

而就算每周只做到 2%，靠复利也是巨量的，最后能做出几十倍的增长。

That's why in our company we have that growth mindset, where every day matters, every release matters.

所以我们公司就是这种增长思维：每一天都重要，每一次发版都重要。

And everyone in our company simply knows what today's revenue is, what yesterday's revenue was, what retention was, how much we've grown.

公司里每个人都知道今天的收入是多少、昨天的收入是多少、留存是多少、我们涨了多少。

And every employee in the company understands their impact, understands which metric they move.

每个员工都清楚自己的价值在哪，清楚自己在推动哪一个指标。

A product engineer understands how the product they ship lands and how many percent of growth it will produce for the company.

产品工程师清楚自己发出去的功能落地之后，能给公司带来百分之几的增长。

A marketer understands that the more outreach they do, the more percent the company will grow.

市场同学清楚自己多做多少触达，公司就能多涨百分之几。

And an ML engineer understands how much research they do and how much the company will grow as a result.

ML 工程师清楚自己多做多少研究，公司就能因此多涨多少。

And that mindset, it was really important for us to build a company culture where everyone thinks about growth.

这种心态很关键，我们特别看重把公司文化做成人人都在想增长。

And that's why we managed to grow so fast.

所以我们才能涨得这么快。

At the end of 2025, in the fall and winter, we were already growing 50%, 80% month over month.

2025 年年底，秋天和冬天那阵，我们的月环比增长已经到了 50%、80%。

And we haven't dropped below 30% month over month even once since the release.

而且从产品发布到现在，月环比就没掉到 30% 以下过。

And for example, I recently saw Paul Graham giving an interview, and he said he has a company that's growing something like 28% a month, and he advised them to grow 20% a month, because holding 28% is just super unrealistic.

举个例子，我最近还看到 Paul Graham 的一个采访，他说他手上有家公司月增长大概 28%，他反而建议对方按每月 20% 来，因为一直保持 28% 实在太不现实了。

And on the contrary, he advised them to slow their growth down, because that growth scared him — he was worried the company might start burning resources inefficiently or something like that.

结果他反过来劝他们把增长慢下来，因为这个增速把他吓着了，他担心公司会开始低效地烧资源之类的。

And it was fun for me to read that, because for us that's just our baseline.

我读到这段觉得挺有意思，因为这个数字对我们来说就是基准线。

We can't even imagine growing at less than 30%.

我们没法想象自己的增长会低于 30%。

And what makes that possible?

那这是靠什么做到的？

Back in the day, at classic SaaS companies, there was a growth department — what did it actually do?

以前那些经典的 SaaS 公司都有个增长部门，它到底在干什么？

It worked on user activation.

它做的是用户激活。

You build some SaaS product, and then all the growth frameworks are about user activation.

就是说你做出一个 SaaS 产品，接下来所有增长框架都围绕用户激活转。

You have to get the user to the aha moment, activate them as fast as possible — the moment the user understands the product's value — and you build this whole funnel, starting from where he saw the ad or where he landed on the landing page.

你得让用户尽快到达 aha 时刻，也就是他真正明白产品价值的那一刻，然后把整条漏斗搭起来——从他看到广告、或者落到 landing page 那一刻开始。

And you try to activate the user and optimize that moment — that's the classic SaaS approach.

然后你就想办法激活用户、优化这个环节，这就是经典的 SaaS 打法。

But in GenAI everything has changed, because products ship much faster, everything changes much faster.

但到了 GenAI 这里全变了，因为产品上线快得多，一切都变得快得多。

And all those old optimization things can go to the scrap heap, in my view — you can throw out the old growth frameworks.

在我看来，那些老的优化玩意儿可以扔进垃圾堆了——老的增长框架可以扔了。

The main framework now is that everyone has to understand that the cost of building a product is already close to zero, thanks to Claude Code, Cursor, the things Anthropic is doing.

现在最主要的框架是，所有人都得明白：做一个产品的成本已经接近于零，靠的是 Claude Code、Cursor，靠的是 Anthropic 在做的那些东西。

Products get built very fast now.

现在产品做得非常快。

The cost of a product comes down to the price of the tokens it takes to build it, especially if it's a prototype.

产品的成本基本就等于把它做出来所消耗的 token 钱，尤其是做原型的时候。

So your job is to ship a large number of products as fast as possible, stay afloat, and not miss the market's next iteration of whatever is happening.

所以你的任务就是尽可能快地上线大量产品，先在水面上浮住，别错过市场的下一轮迭代。

And that's exactly what we live by.

我们信奉的就是这一套。

We ship a lot of products.

我们上线的产品非常多。

We have some really cool products inside Higgsfield.

Higgsfield 内部有一些非常棒的产品。

And originally they weren't products at all.

而它们一开始根本不是产品。

For us, our first product is our founding engineer, Almaz.

对我们来说，第一个产品其实是我们的 founding engineer，Almaz。

And our second product is Aksultan — he's a former founder, he came to us from England, he lived in England for 12 years, a super smart guy.

第二个产品是 Aksultan，他以前自己创过业，从英国过来加入我们，在英国住了 12 年，非常聪明。

And they're very opinionated people, they don't build optimizations, that user-activation stuff.

他们都特别有主见，不会去搞什么优化、用户激活那一套。

They have a ton of ideas, super opinionated, and they're the ones building the products.

他们脑子里一堆想法，主见极强，产品就是他们做出来的。

And obviously most of the products we make don't go anywhere, they don't get traction, but because we keep velocity high and ship a large number of products, one product out of ten really takes off.

当然，我们做的大部分产品后面都没什么下文，起不来增长势头，但因为我们把迭代速度压得很高、上线的产品够多，十个里总有一个真的能爆。

And that's how we manage, month after month, to put out a new hit, some new cool product, a new release that lets us push our growth higher.

靠这个，我们才能每个月、月复一月地做出一个新爆款、一个新的好产品、一次新发布，把增长往上再推一截。

And we understand this structure really well.

我们把这套结构理解得非常透。

We understand the structure of distribution well, the structure of how to ship products fast, what a minimal prototype of a product should look like.

我们很清楚分发的结构、怎么快速上线产品、一个产品的最小原型该长什么样。

And because of that we're scaling it horizontally even harder.

所以我们现在还在横向把它扩得更狠。

Right now we're hiring for new products, we're hiring more product engineers, we're growing the team, because we understand we can scale even more: at first we only built the product for professional filmmakers, then we started building for people who do ads, marketing and video on social media, and now we see there's a market of motion designers and a B2B market for educational videos, and we cover all of those markets.

现在我们在为新产品招人，招更多产品工程师，把团队做得更大，因为我们知道还能继续放大：一开始我们只给专业电影人做产品，后来开始做给那些在社交媒体上投广告、做营销、做视频的人，而现在我们看到还有 motion designer 这个市场，还有 B2B 的教学类视频市场，这些市场我们全都覆盖。

That's what lets us grow very fast.

靠这个，我们才能长得这么快。

Let me add something so you can picture it clearly — guys, if Higgsfield sticks to that baseline of theirs, the minimum one, then by the end of the year they'll become a decacorn and catch up with Kaspi in market cap.

我再补一句，让大家有个直观的概念——各位，如果 Higgsfield 就按他们说的那个最低基准线走，到年底他们就是十角兽了，市值追平 Kaspi。

A decacorn is a company that has already crossed the 10 billion mark.

所谓十角兽，就是估值已经突破 100 亿美元大关的公司。

Great.

太棒了。

All right.

好。

So Yerzat, what's next?

那 Yerzat，接下来呢？

What goals are you setting for yourselves?

你们给自己定了什么目标？

What milestones do you see coming next?

接下来的里程碑你们怎么看？

Globally, as a company we don't plan very far into the future.

整体来讲，公司层面我们不做特别远的规划。

We're usually always focused on the current week.

我们通常永远只盯着当前这一周。

We don't make any super long-range plans.

我们不做那种特别长远的计划。

That's part of our mindset too — because everything in GenAI changes so fast, you can simply get it wrong.

这也是我们的一种心态：因为 GenAI 里一切变得太快，你很可能就是会判断错。

You start mapping out the future, and in the end everything changes differently, everything turns out differently.

你开始设想未来，结果一切都朝另一个方向变了，事情完全按另一种方式发生。

So our mindset is, first of all, that it's very easy for us to kill projects — we have projects we did a lot of R&D on, spent several weeks building, and then one day something comes out and within a minute we all make the decision that, okay, we need to throw this away, forget about it and rebuild everything.

所以我们的心态首先是：放弃项目对我们来说非常容易——有些项目我们做了大量 R&D，花好几周开发，然后某一天外面出来个什么东西，我们一分钟之内就集体决定：行吧，这个扔掉，忘了它，全部重来。

That's our mindset, that's how we always think, we're super adaptive, because everything in the market changes fast.

这就是我们的心态，我们一直都是这么想的，我们适应能力极强，因为市场上什么都在快速变。

Who knew a year ago that OpenAI would fall off its pedestal?

一年前谁能想到 OpenAI 会从神坛上跌下来？

It was completely impossible to imagine that OpenAI's model — probably a lot of people here have already stopped using ChatGPT, because the model, relative to Gemini, relative to Anthropic, has gotten dumb, and OpenAI lost enterprise, and everything in the market changed.

当时完全无法想象 OpenAI 的模型会这样——估计在座很多人已经不用 ChatGPT 了，因为跟 Gemini 比、跟 Anthropic 比，这个模型已经变笨了，而且 OpenAI 把企业市场也丢了，市场整个变了。

And there was a moment when Meta released Llama and everyone was burying Google.

还有一阵子，Meta 发布了 Llama，所有人都在给 Google 送葬。

Google's stock was falling, everyone was saying it's over — goodbye, Google.

Google 股价往下掉，所有人都在说完了，Google 再见吧。

And now Gemini is the top model on the market.

而现在 Gemini 是市场上最顶的模型。

You have to understand that if players that big get it that wrong, then we're small — we need to be even more adaptive.

你们得明白：这么大的玩家都能错得这么离谱，那我们这么小，就更得有适应能力。

And that's the approach we apply.

我们用的就是这个思路。

We know for sure that if you stick to these growth numbers, we know for sure that if you understand the structure and understand these frameworks, they let you grow fast and grow hard, if you strengthen the team, because a team that does 30% month over month for a whole year has a very high probability of continuing to do 30% month over month the next year, especially when it's not survivorship bias, not just getting lucky, but when we literally pulled ourselves up by our own bootstraps. We ran a huge number of hypotheses, and usually only one out of ten hits.

我们非常确定，只要守住这些增长数字——只要理解这套结构、理解这些框架，它们就能让你快速、大幅地长，前提是不断加强团队；因为一支连续一年做到每月环比 30% 的团队，第二年大概率还能继续做到 30% 环比，尤其当这不是幸存者偏差、不是运气好，而是我们硬把自己拽起来的时候。我们跑了大量假设，十个里通常只有一个能成。

And we know how to iterate on them, we know how to launch them.

我们知道怎么迭代它们，也知道怎么把它们推出去。

So our plan is to keep up this growth and, as Murat and Magzhan said earlier, to reach those numbers — becoming a decacorn is of course our goal.

所以我们的计划就是把这个增长延续下去，像 Murat 和 Magzhan 刚才说的，达到那些数字——成为十角兽当然是我们的目标。

We'll have a short Q&A session at the end, but as the moderator I'd like to ask a couple more questions.

最后我们还有一个简短的 Q&A 环节，不过作为主持人，我想再多问几个问题。

You understand the AI industry really well, really deeply.

你们对 AI 行业的理解非常好、非常深。

So in your view, Yerzat, what's happening in this space right now?

Yerzat，在你看来，这个赛道现在正在发生什么？

Where is it all heading, how soon will we get AGI, and what other business models do you think could take off in the near future?

一切要往哪儿走，AGI 还有多久，还有哪些商业模式你觉得近期可能会爆？

Great question.

这个问题太好了。

Right now the biggest revolution has probably been the one Anthropic made.

现在掀起最大一场革命的大概是 Anthropic。

A lot of people here have probably used the latest Claude Code, the latest model, Opus from Anthropic.

在座很多人估计都用过最新的 Claude Code，用过 Anthropic 最新的 Opus 模型。

And the model really can carry out a very large number of complex sequential actions.

这个模型确实能完成非常多、非常复杂的连续动作。

The model hardly does anything dumb anymore, and it makes the maximum number of correct decisions.

模型现在几乎不犯傻了，做对决策的次数拉到了最大。

And because of that, these agents everyone has been talking about for the last two years — they were more of a hype thing — now these agents are really becoming a reality, they're really becoming useful.

正因为这样，过去两年大家一直在讲的那些 agent——以前更多是炒作——现在真的在变成现实，真的开始有用了。

Maybe you've heard of Clawdbot — it's been hyped hard in America over the last 10 days.

你们可能听说过 Clawdbot，最近 10 天在美国被炒得很厉害。

That's an open source product too.

它现在也是个开源产品。

And the most promising thing you can do right at this moment is to start building a product on Claude, on Anthropic's model.

所以现在最有前景的事情，就是开始在 Claude、在 Anthropic 的模型上做产品。

You can cover a huge number of verticals there — legal, marketing, adtech?, anything.

在上面能覆盖的垂直领域非常多——legal、marketing、adtech?，什么都行。

Anywhere that requires coordinating people.

凡是需要协调人的地方。

Where a large number of actions have to be performed.

凡是需要执行大量动作的地方。

Now you can automate that with Claude, and it unlocks a huge number of markets.

现在这些事情用 Claude 就能自动化，这一下解锁了大量市场。

For example, there was Manus, which Meta bought, and Manus was considered a really cool agent.

比如之前有个 Manus，被 Meta 收购了，当时 Manus 被认为是很厉害的智能体。

Now this Clawdbot has come out, and you can't even compare Clawdbot's traffic to Manus's.

现在这个 Clawdbot 出来了，Clawdbot 的流量跟 Manus 根本不在一个量级。

Manus Bot — if you just look at the traffic on Google Trends, Manus basically doesn't exist and never did.

Manus Bot——你到 Google Trends 上看一眼流量就知道，Manus 基本不存在，过去也没存在过。

How much has Clawdbot grown by now?

Clawdbot 现在长到什么程度了？

Clawdbot.

Clawdbot。

It's built right on top of Anthropic's model.

它正好就是基于 Anthropic 的模型做的。

And it's only 10 days old.

而且它才上线 10 天。

And it has already overtaken Cursor, it has overtaken Anthropic in traffic.

流量上它已经超过了 Cursor，也超过了 Anthropic。

You can literally go into Google Trends right now and see it.

这个你现在就可以打开 Google Trends 看。

Things like that happen in an instant, and you have to keep your finger on the pulse.

这类事情就是一瞬间发生的，你得随时把握动向。

And it really is a huge business opportunity to build something on top of Claude.

在 Claude 之上做点东西，这确实是个非常大的商业机会。

Murat, from your perspective, from your vantage point as a venture investor, where would you advise investing?

Murat，从您的角度、从您作为风险投资人的视角看，您建议往哪儿投？

What trends are you seeing right now?

您现在看到了哪些趋势？

Yes, like I say, we've actually worked out a kind of gradation for ourselves: there are startups doing something like copy-paste, meaning already known models that have been tried and tested in various markets.

对，我说过，我们其实给自己做了一个分级：一类创业公司做的是所谓的复制粘贴，也就是照搬那些已经在各个市场验证过的成熟模式。

And over the last five years those were the models, they were the majority.

而过去五年，这类模式占了绝大多数。

All the well-known projects — Choco?, Aleksey Li?, the well-known Ticketon, Aviata — they're all really copy-pastes that were simply adapted from an outside market to ours.

那些知名项目——Choco?、Aleksey Li?、大家熟悉的 Ticketon、Aviata——本质上都是复制粘贴，只是把外面的模式搬过来适配本地市场。

Then come the transitional models, where you take a copy-paste model and then it gets adapted to the market and pivots very heavily.

再往上是过渡型模式：先拿一个复制粘贴的模式，然后按本地市场去调整，中间会有很大幅度的转型。

And what comes out is something new and interesting.

最后长出来的是全新的、有意思的东西。

That's the case with ProsperPay, for instance, which started out advancing people their salary before payday, and today it's turning into a neobank for the gig economy, because these new pains and problems that surfaced have only appeared now, and now it's the one closing them.

比如 ProsperPay 就是这样，一开始做的是在发薪日之前先把工资预支给你，今天它正在变成面向零工经济的新型数字银行——因为这些新暴露出来的痛点和问题是现在才出现的，而现在正由它来解决。

Right.

就是这样。

And the last category, of course, the most active one right now and the one we've been putting money into exclusively all of last year, is AI-native projects.

最后一类，当然也是现在最活跃的一类，说白了也是我们去年一整年唯一在投的方向，就是 AI Native 项目。

And practically 100% of ours — last year we closed, I think, about 12 deals together with the club, for somewhere around 12 or 13 million dollars.

我们这边几乎是 100%——去年我们和俱乐部一起大概做了 12 笔交易，总额差不多 1200 万到 1300 万美元。

These are all AI-native projects that have that kind of potential.

这些全是有这种潜力的 AI Native 项目。

Among them there's a project called Chart?.

其中有一个项目叫 Chart?。

It's roughly the same thing as the Clawdbot that Yerzat just described.

它和 Yerzat 刚才讲的 Clawdbot 差不多是一回事。

And it'll probably be released in the first half of February.

它大概会在 2 月上半月发布。

Take a look and give it a try.

大家可以去看看、试一试。

It's our Kazakh project.

这是我们哈萨克斯坦自己的项目。

Excellent.

很好。

All right.

好的。

Yerzat, I think we've all noticed that you — Murat and Yerzat — are in suits today.

Yerzat，大家可能都注意到了，您和 Murat 今天都穿了西装。

It's unusual to see you looking like this.

看到二位这个样子还挺不习惯的。

Well, we know that you went to an official meeting with the head of state.

我们知道，你们去参加了与国家元首的正式会见。

And you know that this year has been declared the year of digitalization.

另外您也知道，今年被定为数字化之年。

A lot of time and attention is being given to developing artificial intelligence right now, and a Ministry of Artificial Intelligence has been created.

现在我们在人工智能发展上投入了非常多的时间和精力，还成立了人工智能部。

Very soon we'll be launching the Alem.AI artificial intelligence center.

我们很快就要启动 Alem.AI 人工智能中心。

We've been actively getting it ready for launch for the past year and a half.

过去一年半我们一直在积极准备它的启动。

And as far as I know, you discussed setting up a lab there.

据我所知，你们讨论了在那里设一个实验室。

Please say a couple of words about that.

请您就这件事讲几句。

Yes, we told the head of state that together with the ministry and together with Astana Hub we're planning to open an artificial intelligence lab based at Alem.AI.

对，我们向国家元首汇报了，我们计划和部里、和 Astana Hub 一起，在 Alem.AI 的基础上建一个人工智能实验室。

And obviously the talent question is the sharpest one, because if we want to keep growing like Higgsfield, and if we want new startups, AI startups, to appear in Kazakhstan, it all comes down to AI expertise.

很明显，人才是最尖锐的问题，因为如果我们想继续像 Higgsfield 那样成长，如果我们希望哈萨克斯坦出现新的 AI 创业公司，归根到底都是 AI 专业能力的问题。

So we want to build a lab where we'll train AI specialists, because AI specialists are the core of any AI startup, any AI product — they're the company's intellectual property, they're what gives the company its valuation, what produces those big multiples.

所以我们想做一个实验室来培养 AI 人才——AI 人才是任何一家 AI 创业公司、任何一个 AI 产品的内核，是公司的知识产权，是撑起公司估值、撑起那些高倍数的东西。

Because at an early stage those AI multiples are literally just the number of AI engineers and AI researchers you have.

因为在早期阶段，所谓的 AI 倍数其实就是你手上有多少 AI 工程师、AI 研究员。

And a supercomputer has appeared in Kazakhstan.

哈萨克斯坦已经有了超级计算机。

Kazakhstan already has several powerful supercomputers and data centers with several hundred of NVIDIA's latest high-end GPUs.

哈萨克斯坦已经有好几台强大的超级计算机和数据中心，里面装着几百块 NVIDIA 最新的高端 GPU。

And AI specialists will be trained there as well.

AI 专业人才也会在那里培养。

We obviously have a lot of experience with this, and we're ready to share that experience.

我们在这方面显然积累了很多经验，也愿意把这些经验分享出来。

Because it matters enormously for the whole ecosystem.

因为这对整个生态都非常重要。

We want as many AI startups as possible to appear in Kazakhstan.

我们希望哈萨克斯坦出现尽可能多的 AI 创业公司。

And in the future, obviously, artificial intelligence will account for a substantial share of global GDP — even now.

未来人工智能显然会占到全球 GDP 相当大的一块，其实现在就已经是这样了。

These are big numbers, driven by those data centers, by building new power plants.

这些都是很大的数字，背后是那些数据中心，是新建的发电站。

A pretty large supply chain is already being built around AI, already at the level of geopolitics.

围绕 AI 已经在建起一条相当长的供应链，而且已经上升到地缘政治层面。

Trump is out there discussing data centers in Greenland.

Trump 那边都在讨论要在格陵兰建数据中心。

And with this AI that's going to take up a large share of global GDP, Kazakhstan now has a chance to plug into that niche, the way South Korea or Japan managed to claim their niches in past technological revolutions.

而这个将会占据全球 GDP 很大一部分的 AI，给了哈萨克斯坦一个嵌进去的机会，就像当年韩国、日本在过去几次技术革命里抢下自己的位置一样。

At the time South Korea wasn't a super developed country, but in the twentieth century it managed to capture a lot of great technological niches.

当时的韩国也不是什么特别发达的国家，但在二十世纪它拿下了不少很好的技术赛道。

And today it's a rich, stable country.

今天它已经是一个富裕、稳定的国家。

Now that same chance has fallen to Kazakhstan.

现在这样的机会轮到哈萨克斯坦了。

And it's really great that the head of state supports all of this.

特别好的一点是，国家元首对这一切都很支持。

And we have great initiatives like Astana Hub, and the ministry is doing a lot of work to support the ecosystem.

我们也有 Astana Hub 这样很好的机制，部里也在为支持生态做大量工作。

We want to contribute to that too, and I hope we'll manage to build a great AI lab.

我们也想为此出一份力，希望能把这个 AI 实验室建好。

Yes, that really is very valuable.

对，这确实非常有价值。

Thank you, Yerzat, for this initiative.

谢谢你，Yerzat，谢谢这个倡议。

Having worked on Astana Hub for so many years, we understand that we've managed to get many things on track, many things here work.

做 Astana Hub 这么多年，我们清楚很多事情已经走上了正轨，很多东西在这里是跑得通的。

Accelerators, venture investors, the centers we're building, taking startups abroad, and so on and so forth.

加速器、风险投资人、我们在建的各种中心，把创业公司带出国门，等等等等。

But it's very hard to bring big tech R&D here.

但要把 big tech 的 R&D 引进来非常难。

We don't pass on a lot of factors and so on — when you talk to some hypothetical big tech company about R&D, they just hang up a sign here, and in essence it turns into some kind of showroom.

很多条件我们都过不了关，跟某个假想的大厂谈 R&D 的时候，他们只是在这儿挂个牌子，实质上就变成了一个展示间。

That's why I've been waiting so long for Kazakh companies to appear that would build real R&D here, because we understand that it will go on to become an icebreaker for our whole industry.

所以我一直在等哈萨克斯坦本土出现这样的公司，能在这里做真正的 R&D，因为我们清楚，它会成为给整个行业开路的破冰船。

All right.

好。

Then before we move on to the Q&A, one last question from me, a small short one.

那么在进入 Q&A 之前，我这边还有最后一个问题，一个很小很短的问题。

How can you help other startups, beyond this lab?

除了这个实验室之外，你们还能怎么帮助其他创业公司？

Are you thinking about — if you're ready to answer, maybe it's still too early and your focus is somewhere else right now — but are you ready to invest, or to share, I don't know, technology, datasets, so that other Kazakhstani tech companies? can grow?

你们有没有想过——如果现在方便回答的话，也许还太早，你们眼下的重心也在别处——但你们愿不愿意投资，或者说分享技术、数据集，好让哈萨克斯坦其他的科技公司?也能发展起来？

Yeah, good question.

对，好问题。

We're doing that — I mean, we understand that if you take our B2B side, because the market is changing so fast, today's startups can become big clients in no time, so, in general, this is very common now, when companies, even mid-sized ones, put together various products, offers, some kinds of subsidies.

我们在做这件事——我是说，我们很清楚，就我们 B2B 这条线来看，因为市场变得太快，今天的创业公司过不了多久就可能成为大客户，所以现在这种做法其实很普遍：连中等规模的公司都会推出各种产品、方案，还有一些补贴。

for other startups that then quickly grow into large clients, because it's not just Higgsfield that's growing fast right now — there are a lot of fast-growing projects in the market in general.

面向其他创业公司，而这些公司之后会很快长成大客户，因为现在成长快的不只是 Higgsfield，市场上快速增长的项目其实很多。

So yes, that direction is interesting to us, and within our B2B track we're planning to do some kind of subsidies for earlier-stage startups, because, naturally, we hope that today's startups will grow into large clients.

所以是的，这个方向我们很感兴趣，我们也打算在 B2B 这条线上给更早期的创业公司做一些补贴，因为我们当然希望，今天的这些创业公司将来能长成大客户。

Great.

太好了。

Well then, we'll push all of these activities under the flag of the Higgsfield lab based at Alem.AI?.

那我们就把这些事情都放在设于Alem.AI?的 Higgsfield 实验室这面旗下来推进。

Excellent.

非常好。

Thank you.

谢谢。

So then, let's move on to the Q&A session, please.

那么接下来我们进入问答环节，有请。

I— uh, now now, hold on, hold on, wait, I'm the moderator here, we'll be passing the microphone around, please.

我——呃，别急别急，稍等，稍等，这里我是主持人，我们会把话筒传下去，请稍候。

I think we'll have time to hear a lot of questions, so please, be patient.

我想我们有时间听不少问题，所以请大家耐心一点。

Mhm.

嗯。

Please, introduce yourself and ask your question.

请先做个自我介绍，然后提问。

Hello, my name is Arman.

您好，我叫 Arman。

I'd like to thank the Astana Hub organizers for such a great event.

我想感谢 Astana Hub 的组织者办了这么棒的一场活动。

I'd also like to thank Magzhan for moderating — you really draw people out with such deep questions.

也想感谢 Magzhan 的主持，你的问题很深，特别能把人问开。

So, I have a question — and I also wanted to say a huge thank you to the Higgsfield team for representing Kazakhstan on the world stage.

我有一个问题——另外也想向 Higgsfield 团队表达由衷的感谢，感谢你们在世界舞台上代表哈萨克斯坦。

I'd like to wish the Higgsfield team, Yerzat and everyone on it, the best of luck.

祝 Higgsfield 团队、祝 Yerzat 和整个团队好运。

We're really proud, honestly, we're incredibly proud.

说实话，我们非常自豪，真的特别自豪。

A question for Murat Almazbekovich: at the early stage, what matters most?

想问 Murat Almazbekovich：在早期阶段，什么最关键？

And, well, what's the signal?

还有，那个信号是什么？

What tells you that a project is worth backing, even when the numbers aren't all that good at the early stage?

你凭什么判断一个项目值得支持，哪怕它早期的数字并不好看？

And a question for Yerzat: what key decisions did you make?

另外还有一个问题问 Yerzat：你们做过哪些关键决策？

What led to such strong growth for Higgsfield?

是什么让 Higgsfield 增长得这么强劲？

Thank you. Well, I want to say that our tactics have changed a bit.

谢谢。我想说的是，我们的打法有了一些变化。

Before, we demanded hard traction — we needed product-market fit, confirmation, all of that — but now we see that time and speed have really changed, and by the time we wait for product-market fit to be confirmed, the company's valuation has already flown past 100 million dollars.

以前我们要求有硬的增长势头，就是说必须先有产品市场契合、有验证，这些都得有；但现在我们看到，时间和速度真的变了，等我们等到产品市场契合被验证，公司估值已经飞过一亿美元了。

That's why we changed our tactics.

所以我们才改了打法。

We started coming in at pre-seed with smaller checks — we cut our check size by about three times, so we could support the company and keep a finger on the pulse, because otherwise you just don't make it in later.

我们开始用更小的支票在 pre-seed 阶段进场，把单笔金额砍到原来的三分之一左右，这样既能支持公司，又能一直把着脉，因为再往后你根本挤不进去。

Right.

就是这样。

And last year, that's how we went into about eight companies, with a small check, 100 to 150 thousand, when there isn't even any real traction yet — there's an MVP, and that's about it.

去年我们就是这样进了大概八家公司，单笔金额不大，十万到十五万，那时候连像样的增长势头都还没有——有个 MVP，也就这样了。

So what do we look at then, how do we decide whether a project is a fit for us or not?

那我们到底看什么？我们怎么判断一个项目适不适合我们？

Of course, we look at the product itself and at the vertical, at what potential it has.

当然，我们会看产品本身，也看它所在的垂直领域，看它有多大潜力。

For example, we look at the Mirai? project, and we see, okay, these are insoles for sports, but we also look at the product and see the resolution is 2 millimeters, and we realize this is artificial tactile skin for robotics.

比如我们看Mirai?这个项目，看到的是运动用的鞋垫，但我们再看产品本身，分辨率是 2 毫米，于是明白了：这其实是机器人用的人造触觉皮肤。

Robotics is the main trend, one of the main trends right now, for the next few years, and that's the kind of potential that can become global.

机器人是接下来几年最主要的趋势之一，这种潜力是能做到全球规模的。

And on top of that, of course, we look at how much experience and knowledge the founders have, because everything is changing so fast that you have to change things fast, act, and be very flexible and proactive.

在这之上，我们当然还要看创始人有多少经验和认知，因为一切变得太快了，你得非常快地调整、动手，得非常灵活、非常主动。

Those are the two factors that let us make a decision today.

这就是今天让我们做出决定的两个因素。

Thank you.

谢谢。

One second.

稍等一下。

There was a second question, for Yerzat.

刚才还有第二个问题，是问 Yerzat 的。

And let's agree, please, one question each.

另外我们说好，请每个人只问一个问题。

Mhm.

嗯。

Yes, if we're talking about key decisions, then probably, as I said earlier, the first one is thinking outside the bubble.

对，说到关键决策，大概就像我前面讲的，第一条是跳出泡泡去想问题。

Once we stopped copying the mainstream — what Y Combinator teaches, what's fashionable in Silicon Valley — we realized we needed to come up with our own frameworks, to move in our own way.

当我们不再跟着主流走——不再照搬 Y Combinator 教的那套、硅谷流行的那套——我们意识到得自己想框架，自己走一条不一样的路。

That worked really well.

这一步效果特别好。

And mindset is also very important.

另外，心态也非常重要。

And that's genuinely hard when you're sitting in Kazakhstan — if you've read Zero to One, Peter Thiel has this quadrant there where he lays out the mindsets of different countries.

而人在哈萨克斯坦，这一点确实很难——如果你读过 Zero to One，Peter Thiel 在书里画了一个四象限，把不同国家的心态摆在一起。

And how does he contrast America with China there?

那他在书里是用什么把美国和中国对照起来的呢？

By the fact that in America people have — well, at least at the time Zero to One was written — this super positive way of thinking, a belief in a bright future, that tomorrow will be better than today.

是这一点：在美国，人们——至少在 Zero to One 写作的那个年代——有一种极其正向的思维方式，相信未来是光明的，相信明天会比今天更好。

the market will grow, the S&P will grow, there will be new startups, everything will get disrupted.

市场会涨，S&P 会涨，会有新的创业公司出来，一切都会被颠覆。

That mindset is really a matter of faith.

这种心态说到底其实是个信不信的问题。

You believe the economy really will grow, that technology really will advance.

你相信经济真的会增长，相信技术真的会往前走。

It's that optimistic mindset.

就是这样一种乐观的心态。

And with China, he shows that it works differently there.

而对中国，他展示的是另一套运行方式。

And so at the time the book was written, China's mindset was that you don't need to invent, you need to copy.

所以在那本书写作的时候，中国的心态是：不用自己发明，照着抄就行。

and a more pessimistic view of the future.

以及对未来更悲观的看法。

In Kazakhstan, given our history — the Soviet Union, we've had a pretty hard history — that mindset is hard for people; the American mindset isn't going to just switch on here at the snap of a finger.

在哈萨克斯坦，考虑到我们的历史——苏联那段，我们这段历史确实挺沉重——这种心态对人们来说很难，美式的那种心态不可能在这里一按开关就运转起来。

And despite the fact that Kazakhs as a nation are generally pretty positive and energetic, you have to understand what Peter Thiel writes in Zero to One: that it's very important for a startup to work with that mindset, to build a startup with it, because it affects a huge number of decisions down the line.

尽管哈萨克人作为一个民族整体上挺乐观、挺有劲头，但这里得理解 Peter Thiel 在 Zero to One 里写的：带着那种心态去做事、去做一家创业公司非常重要，因为它往后会影响到大量的决策。

For example, when we raised 8 million and immediately spent the first 5 million on GPUs — that's a pretty risky decision.

举个例子，我们融到 800 万，转头就把头 500 万砸在 GPU 上，这是个相当冒险的决定。

And we did it with the mindset that we really would be able to train a great model on those GPUs.

我们当时的心态就是：我们真的能用这些 GPU 训出一个很强的模型。

Let's take the risk, we'll pull it off.

赌一把，我们能成。

That's a question of mindset.

这就是心态的问题。

And in America it's easier for them, because they have that mindset by default — it's American culture.

在美国，他们要轻松一些，因为那种心态是默认自带的——这就是美国文化。

And here, that's very important.

而在我们这里，这一点非常重要。

So it's great that Astana Hub, for example, has these programs Magzhan mentioned earlier, where you can fly out to the Valley and see how it works there.

所以特别好的一点是，比如 Astana Hub 有 Magzhan 前面说的那些项目，你可以飞一趟硅谷，看看那边到底是怎么运转的。

That mindset is very important for startups.

这种心态对创业公司非常重要。

Great.

很好。

Yes, thank you.

好的，谢谢。

Mhm, right.

嗯，好。

Yes, let's take the next question.

好，我们下一个问题。

Oh, hello.

啊，你好。

My name is Assel?, I head the business incubator at the Agrotechnopark.

我叫 Assel?，是 Agrotechnopark 商业孵化器的负责人。

And I have a question for you, Murat Abdrakhmanov.

我有个问题想问您，Murat Abdrakhmanov。

And by the way, a huge thank you to Astana Hub — it's a really great, advanced venue.

另外顺便非常感谢 Astana Hub，这里真是个特别棒、特别先进的场地。

And the first thing I always tell students is: "Come here, come to the events and start your journey here."

我跟学生说的第一句话永远是：「来这儿，来参加活动，从这里开始你们的路。」

And, well, we all know that you, Murat Abdrakhmanovich?, are a pretty well-known investor.

我们都知道，Murat Abdrakhmanovich?，您是位相当有名的投资人。

We watch your podcasts very often, read your articles, we study a lot of what you put out.

我们经常看您的播客、您的文章，学到不少东西。

But, you know, people always want to eat.

但您知道，人总归是要吃饭的。

So my question is about agriculture.

所以我的问题是关于农业的。

You said you don't look at it because you don't have the expertise.

您说过您不看这个方向，因为没有这方面的专业积累。

And I'd like to clarify — so what is it that stops you from building up that expertise, or do you still think it's too early for our market and just not interesting?

我想请教一下——到底是什么拦着您把这块的专业积累补起来？还是说您依然觉得这对我们的市场来说太早了、也没什么意思？

I mean, over three months I studied the market in detail, looked at what problems there are in each sector.

我用了三个月时间细致研究市场，看每个行业各自卡在什么问题上。

And so I'd really like to hear your expertise on this, so that we don't kill the students' motivation to build out the agritech direction.

所以我特别想听听您在这件事上的判断，免得打消学生做农业科技方向的劲头。

And sorry, can I still ask a second question, because I might not get a chance to come up to you, Yerzat — you'll be busy.

不好意思，我还是想再问第二个问题，因为待会儿可能没机会走过去找您了，Yerzat，您肯定很忙。

My question is this: you test a lot of hypotheses.

我的问题是这样：你们会测试非常多的假设。

Do you actually have any desire to test hypotheses in the direction of MedTech, or is that a closed topic for you, not interesting, and you don't see anything big there for Higgsfield to achieve major results?

你们是真的有意愿往医疗科技（MedTech）方向去测假设，还是说这对你们来说是个封闭话题、没兴趣，觉得那边没有足够大的空间让 Higgsfield 做出大成绩？

Thank you very much.

非常感谢。

Well, I'll say this — there are a lot of projects anyway, and of course we've worked out certain internal rules for ourselves.

我这么说吧——反正项目本来就很多，我们当然也给自己定了一些内部规矩。

One of the main rules I've set for myself is not to invest in what I don't understand — because it can be very promising, very profitable.

其中我给自己立的一条最重要的规矩，就是不投自己看不懂的东西——因为它看上去可能非常有前景、非常赚钱。

Right.

嗯。

But still, agriculture projects do come in from time to time, and we do look at them.

但农业类的项目还是会时不时飞过来，我们也会看。

We've even evaluated them.

我们甚至还评估过。

I even got in touch — when we don't have enough expertise of our own, we always try to find it on the outside.

我甚至专门去联系过人——我们自己的专业能力不够时，总会想办法去外面找。

And I reached out to guys who work in agriculture.

我联系了一些在做农业的朋友。

And here's what they told me.

他们是这么跟我说的。

They told me, "Murat, you know, all of this is great, but in agriculture there are still so many other problems that haven't been solved," and they said, "We haven't grown up to your level yet."

他们说：「Murat，你知道吗，这些都挺好，但农业里还有那么多别的问题没解决」，还说：「我们还没长到你们那个层次。」

Well, I don't know, it may sound offensive or maybe not, but there are a lot of other unresolved issues there — improving efficiency, the models and so on and so forth — before you even get to IT.

我不知道这话听着伤不伤人，但那边确实还有一堆没解决的问题——提高效率、模型这些等等等等，都得先过了这一关，才轮得到 IT。

So.

就这样。

And they told me, "Brother, a bit later — we still have to reach that level."

他们还跟我说：「兄弟，再往后一点吧，我们还得先长到那个层次。」

Yeah.

对。

And overall I want to say that one of my criteria for where I invest is that it has to be interesting to me.

总体上我想说，我投不投还有一条标准，就是这事得让我觉得有意思。

Right.

嗯。

And honestly, I'll say it straight — I'm a pure “asphalt Kazakh,” I don't understand anything about agriculture, and something just doesn't pull me that way.

老实说，我就是个典型的「柏油路上长大的哈萨克人」，农业我一窍不通，也提不起兴趣。

Well, let me add to that — at the same time, agritech is, yes, a promising thing for the country.

我补充一句——同时，农业科技对国家来说确实是个有前景的方向。

Murat Mazbekovich?, as a private venture investor, picks his own niches, and it's great that projects like these are showing up here.

Murat Mazbekovich? 作为私人风险投资人，挑的是他自己的赛道，这很好，而且我们这儿能冒出这样的项目是好事。

But there are a lot of other investors out there.

但外面还有很多别的投资人。

Let's get them involved more actively.

我们要更主动地把他们拉进来。

The ones who are tied to agriculture — there's plenty of interesting stuff there too — but I agree with many of the points: there are still, let's say, some basic layers that need to be built out as well.

那些跟农业沾边的投资人，那边其实也有很多有意思的东西——不过他很多观点我是认同的：那边还有些基础的底子需要先搭起来。

Okay, good.

好的。

And there was a second question about MedTech — when I said we launch a very large number of products, what I meant is that we're still inside one market.

还有第二个问题，是关于医疗科技的——我说我们会推出非常多的产品，意思是我们始终待在同一个市场里。

Our market is video generation, and within video generation we run a large number of hypotheses and iterations.

我们的市场是视频生成，我们是在视频生成内部跑大量的假设和迭代。

MedTech really is a completely different market.

医疗科技毕竟是完全不同的另一个市场。

We're unlikely to ever get into it within Higgsfield.

在 Higgsfield 的范畴里，我们大概不会去碰它。

Mm-hm.

嗯。

Excellent.

太好了。

Yes, go ahead.

好，请讲。

Hello, everyone.

大家好。

My name is Zhenis?.

我叫 Zhenis?。

I'm the founder of a project called Lastyle?.

我是 Lastyle? 这个项目的创始人。

And — please don't address one of them specifically, please address everyone.

另外——请不要只冲着某一位提问，请面向所有人。

Please keep your questions very short.

另外请把问题说得尽量简短。

Thank you.

谢谢。

Okay.

好的。

We have a platform called Network?, and we organize group games and events.

我们有个叫 Network? 的平台，做的是组织团体游戏和活动。

My question is for Murat.

我的问题想问 Murat。

Do you look at projects like that for investment, or only AI?

这类项目您会考虑投吗，还是只看 AI？

No, I don't.

不，不考虑。

Short question, short answer.

问题短，回答也短。

Thank you.

谢谢。

Let's move on.

我们继续。

But again, gamedev needs to be developed.

不过话说回来，游戏开发这块也得发展起来。

We'll be running a lot of programs here too and bringing in more investors.

我们这边也会做很多扶持项目，把更多投资人引进来。

Uh-huh.

嗯。

And over here — this guy has had his hand up from the very beginning.

这边这位小伙子从一开始就一直举着手。

No, it wasn't you.

不，不是你。

All right, go ahead.

行，你说吧。

Divi?, you can't hear me.

Divi?，听不见我说话。

All right, so.

好，那么。

No, no.

不，不是。

Turn on the microphone, please.

请把麦克风打开。

We've also got the livestream running.

我们这边还在直播。

Yeah, so, my name is Adis.

嗯，我叫 Adis。

I'm a digital creator.

我是做数字内容的创作者。

Thank you.

谢谢。

First of all, thanks to Yerzat for Higgsfield.

首先要谢谢 Yerzat 做出了 Higgsfield。

We're making 10x right now, uh, and some people are making 100x.

我们现在赚的是过去的 10 倍，还有人赚到了 100 倍。

We're really happy for them.

我们真心替他们高兴。

Thank you, Murat, for believing in them, and Magzhan for organizing this.

谢谢 Murat 当初信了他们，也谢谢 Magzhan 张罗了这场活动。

It's all been great.

一切都特别棒。

But what I have is probably not so much a question as a proposal, because we like to give back too.

不过我要说的可能算不上问题，更像一个提议，因为我们也喜欢回馈。

You give a huge amount back to Kazakhstan.

你们给哈萨克斯坦回馈了太多。

You've got that patriotic mindset.

你们身上有那种爱国的心气。

And you all know there's the New York Film Academy, which is based in New York.

大家都知道，纽约有一所 New York Film Academy。

Their headquarters is in LA.

他们的总部在 LA。

And they're looking into making full-fledged films through generation and turning that into a whole discipline.

他们正在考虑用生成技术做出完整的电影，并且把这件事做成一门正式的学科。

Right now our guy is there, an Oscar nominee, he's a producer, he's even shot documentaries for Netflix, and he has a direct line to the rector of that academy, and we'd like to propose Higgsfield as the core discipline, so that it doesn't end up being some other players.

我们的人现在就在那边，他是奥斯卡提名者，是制片人，还给 Netflix 拍过纪录片，跟这所学院的院长有直接联系；我们想把 Higgsfield 提议成那门课的核心，别让位置被别的玩家占了。

Obviously you've wiped them all out, but still - to offer him a concrete teaching plan within that academy, both on the LA campus and the New York campus, and teaching Higgsfield directly, your models, your methodology and so on.

当然你们已经把他们全都干掉了，但还是想给他一份具体的教学方案，覆盖 LA 校区和纽约校区，直接教 Higgsfield、教你们的模型、你们的方法论等等。

So, yeah.

就是这样。

And we also have a gift for you, but that's for later, as I understand it.

另外我们还给你们准备了礼物，不过据我了解那是之后的事。

So, yeah.

就是这样。

How do you feel about it?

你们觉得怎么样？

Yeah, great proposal, thank you very much.

好，很棒的提议，非常感谢。

What can I say?

我还能说什么呢？

Awesome.

太棒了。

Go ahead and use Higgsfield.

那就用起来，用 Higgsfield。

Yerzat has no objections.

Yerzat 没意见。

Uh-huh.

嗯。

Good evening.

晚上好。

I have questions for Yerzat.

我有几个问题想问 Yerzat。

I'm a mathematician, I specialize in mathematical intuition and the foundations of mathematics.

我是数学家，专攻数学直觉和数学基础。

I want to work with you.

我想和你们一起工作。

How do I get in touch with you in a way that actually gets read?

要怎么联系你们，才能真的被读到？

Ah, interesting, great background.

哦，有意思，背景很棒。

Before Higgsfield I worked at a Swiss lab.

我在做 Higgsfield 之前，在一家瑞士的实验室待过。

There I was the only non-mathematician in the company, and everyone there was trying to build AI from the standpoint of category theory.

我在那家公司是唯一一个不是数学家的人，其他人都在试着从范畴论的角度做出 AI。

Cool background, yeah.

背景真的很棒。

So, in general, we have an HR person, Dolida? - you could try writing to her, or we can exchange contacts here. And Dolida? is right over there, she's our HR, our VP for hiring, so there you go.

我们其实有位 HR，叫 Dolida?，你可以试着给她写信，或者我们现在就在这儿交换一下联系方式。Dolida? 就在那边，她是我们的 HR，也是负责招聘的 VP，就这样。

Yes, hello again.

好，再次向大家问好。

Aiora? Community, Tech People.

Aiora? Community，Tech People。

So, Yerzat, right now more than half of the people here -

Yerzat，现在在场的人里有一半以上——

- they're from our community.

——都来自我们社区。

We have a community of startup founders and IT people.

我们是一个创业者和 IT 从业者的社区。

And there are going to be a lot of questions like this - how to get into your company, how to meet you, how to ask you some personal questions.

接下来这类问题会非常多——怎么进你们公司、怎么见到你们、怎么问些私人问题。

So.

嗯。

And with the community we'd like to bring literally 10 to 15 people to your company in Almaty for a tour, to see how Higgsfield develops from the inside, what kind of problems you work on and so on.

我们社区想带十到十五个人去阿拉木图你们公司参观一次，看看 Higgsfield 内部是怎么长起来的、你们手上都在攻什么题，等等。

Would that be possible, so that not everyone has to ask their own question right now, so that people could meet you in person and - I know the address where they sit.

有没有这个可能——这样大家就不用现在一个个提问，而是能当面见到你们，然后……我知道他们办公室在哪。

Great, Magzhan, then I'll go through you.

太好了，Magzhan，那我就走你这条线。

Yeah, actually it worked out that Dolida? is based in Astana, so you can always get in touch with her.

对，其实正好 Dolida? 常驻阿斯塔纳，所以随时都能联系上她。

She's super active in the community, pretty much everyone there knows her, so it's easy enough to reach out, get the information, and arrange something.

她在社区里特别活跃，那边基本上人人都认识她，所以联系起来很容易，问信息、约时间都行。

As for an actual tour, we don't really do that, because if you want to come see the workday at its peak, when everything is happening, you'd have to show up around 4 or 5 in the morning, when our releases are going out and everyone's working.

至于参观本身，我们不太搞这个，因为你要是想看工作日最热闹、什么都在发生的那段，得凌晨四五点过来，那时候我们在发版，所有人都在干活。

By the way, I wanted to pick up on that - you mentioned working nights.

对了，你刚提到通宵工作，我正好想接着聊这个。

I remember you telling us that you met with Elon Musk at midnight at their office, and the whole team was working.

我记得你讲过，你们跟 Elon Musk 是半夜十二点在他们办公室见的面，整个团队都还在工作。

Yeah.

对。

Yeah.

对。

I actually tell that story to my colleagues all the time, when they complain that we keep them until seven.

我经常把这个故事讲给同事听，就在他们抱怨我们让他们待到七点的时候。

And another important point I took away from my conversations with Yerzat is that they all work offline, in person.

还有一个重点，也是我跟 Yerzat 聊下来记住的：他们全员都在线下办公。

The company basically has no remote work at all.

公司基本上没有远程这一说。

And a lot of companies in Silicon Valley work in that format now.

现在硅谷很多公司都是这种模式。

That's another one for my colleagues.

这句也是说给我的同事们听的。

Yerzat said that people from Almaty are laid-back and would hate Higgsfield with a work schedule like that.

Yerzat 说，阿拉木图的人比较松弛，这种作息会让他们受不了 Higgsfield。

Yeah, we have almost no employees from Almaty.

对，我们几乎没有从阿拉木图来的员工。

Almost none of them, yeah, are from Almaty, because in Almaty everyone's usually just chilling.

基本上都不是阿拉木图的，因为在阿拉木图大家平时都挺松弛的。

As-salamu alaykum, Yerzat, Magzhan - thank you very much for this event.

As-salamu alaykum，Yerzat、Magzhan，非常感谢你们办了这场活动。

Enormous food for thought, enormous motivation to act.

给了我巨大的思考素材，也给了巨大的行动动力。

Thank you.

谢谢。

I have two questions, they're short, the first one -

我有两个问题，都很短，第一个——

Maybe just one question.

可能只能问一个问题。

Okay.

好的。

Murat Almazbekovich, I have a question for you.

Murat Almazbekovich，我有个问题想问您。

As an investor and a serial entrepreneur — in your opinion, what's the biggest mistake and the biggest red flag with founders and entrepreneurs from our region?

您作为投资人，同时也是连续创业者，在您看来，我们这个地区的创始人和创业者身上最大的错误、最大的危险信号是什么？

Well, it's a classic one.

这个嘛，很经典。

First and foremost, they try to solve a problem that doesn't exist.

首先，他们想解决的是一个根本不存在的问题。

That's roughly 42%, right?

这个大概占 42%，对吧？

I mean, it's an illusion, right — it happens when people don't have enough industry expertise.

说到底这是一种错觉，出现在人缺乏行业专业积累的时候。

So there you go.

就是这样。

And like Yerzat says, you have to try different options, test a ton of hypotheses, to find that product-market fit.

就像 Yerzat 说的，得多试不同的路子，测一大堆假设，才能找到 product-market fit。

And the second one, right behind it, is a false understanding of product-market fit.

紧跟在后面的第二个问题，是对 product-market fit 的错误理解。

You don't even need a fortune teller for that one — it's every other founder.

这个真是不用找算命的——一半的创始人都这样。

Thank you very much.

非常感谢。

Good evening.

晚上好。

First of all, I'm really glad that we now have a unicorn.

首先，非常高兴我们终于有了一家独角兽。

And I strongly believe that next time you'll be sitting here as a decacorn.

我也坚信，下次你们坐在这儿的时候，身份已经是十角兽（decacorn）了。

A question for Yerzat: generative AI is often so unpredictable, and you never really know which prompt will work and which won't.

想问 Yerzat 一个问题：生成式 AI 常常很不可预测，你永远说不准哪个 prompt 会奏效、哪个不会。

Take classification or detection — there it's clear, it works within its metrics.

比如分类或者检测，那就很清楚，在指标范围内它是能工作的。

But when it comes to gen AI.

但换到生成式 AI 这边。

Do you ever get anxious about that, and how do you sleep at night?

你会因为这个焦虑吗？这事儿会不会影响你睡觉？

Yeah, good question.

嗯，好问题。

I started working on deep learning somewhere around 2014, 2015.

我大概是从 2014、2015 年开始做深度学习的。

And back then people believed in deep learning far less than they do now.

那时候相信深度学习的人比现在少得多。

And I basically became a specialist inside a community of people who didn't believe in neural networks, in generative models.

我等于是在一群不相信神经网络、不相信生成式模型的人里面，成了这个方向的专家。

All my friends from that time are classic ML engineers — gradient boosting, XGBoost, decision trees.

我那个年代的朋友全都是传统的 ML 工程师——梯度提升、XGBoost、决策树。

Those were the models that were popular back then.

这些就是当时流行的模型。

And back then I probably spent hundreds of hours arguing — over my whole life — proving that deep learning was the winning strategy.

光是为了论证深度学习才是赢的路线，我这辈子大概花了几百个小时去争论。

And now it has already become mainstream.

而现在，它已经成了主流。

And to me it's super obvious that deep learning — definitely in this century, in the 21st century — is going to be the main tool: right now in NLP, in computer science, and in the future, I think, in medicine, in every field of human activity.

对我来说这太明显了：在这个世纪、在 21 世纪，深度学习一定会是主要工具——现在是在 NLP、在计算机科学，未来我想还会到医学，到人类活动的各个领域。

I believe deep learning will win.

我认为深度学习会赢。

By the way, I think it was back in 2015, right here — when it had only just, well, maybe not just then, but I remember it was at Astana Hub, I think in 2015, I gave a lecture, and I called it "2030".

对了，好像是 2015 年，就在这里——当时才刚刚，嗯，也许不只是那一次，但我记得是在 Astana Hub，好像就是 2015 年，我做过一场讲座，题目叫《2030》。

Back then — and it's still on YouTube, by the way — I was predicting what would happen in 2030.

那场讲座现在 YouTube 上还能找到，里面我算是在预测 2030 年会是什么样。

And my whole talk was simply about how deep learning could be applied to every sphere of human activity.

整场演讲讲的就是深度学习可以怎样应用到人类生活的所有领域。

And I was talking about 2030 there, even though we're now in 2026.

我当时讲的是 2030 年，而我们现在才 2026 年。

And a lot of what I said back in 2015 has come true now.

我 2015 年说的很多东西，现在已经应验了。

So I wouldn't advise anyone to bet against deep learning, against neural networks.

所以我不建议任何人去赌深度学习会输、赌神经网络会输。

That's definitely a direction that, in my view, will win this century for sure.

在我看来，这绝对是一个在本世纪一定会赢的方向。

Oh, hello.

哦，您好。

My name is Mark.

我叫 Mark。

A question for Yerzat.

想问 Yerzat 一个问题。

Thanks, by the way, for Higgsfield.

顺便说一句，谢谢你们做了 Higgsfield。

It's a really handy tool.

非常好用的工具。

And here's a somewhat delicate question.

然后有个比较敏感的问题。

Your edge right now is that you're very fast — you ship something like 200 releases a year, I believe.

你们现在的杀手锏是快——我记得你们一年大概发 200 个版本。

And the problem with Big Tech is that their release roadmap is very long.

而 Big Tech 的问题是，他们的发布 roadmap 特别长。

I mean, just to ship a single model.

光是发一个模型就得那么久。

So my question is: what are you going to do when, say, Adobe — which as far as I know is pretty stagnant right now — or when Big Tech, when Google, gets its act together and adopts the same mindset, that they need to hire the same kind of people who chase speed? Are you afraid that Big Tech companies like that will disrupt you?

所以我的问题是：如果哪天，比如 Adobe——据我所知它现在挺停滞的——或者 Big Tech，比如 Google 缓过劲来，也换上同样的心态，也去招那种一味追速度的人，你们打算怎么办？你们怕不怕被这类 Big Tech 公司颠覆掉？

Yeah, good question.

嗯，好问题。

Actually, Google already works that way.

其实 Google 已经在这么干了。

Anyone who follows this probably knows that Sergey Brin has come back to day-to-day operations at Google.

关注这块的人大概都知道，Sergey Brin 已经回到 Google 一线做具体业务了。

He sits with the researchers every single day, and that's why Gemini has overtaken OpenAI.

他每天都跟研究员坐在一起，所以 Gemini 才反超了 OpenAI。

And there's a joke going around right now — Google is a big, clumsy company.

现在还有个流行的说法——Google 是家又大又笨重的公司。

And to change anything there, you first have to win a Nobel Prize.

在那儿想推动点什么改变，得先拿个诺贝尔奖。

And Demis Hassabis did win a Nobel Prize.

而 Demis Hassabis 真拿了诺贝尔奖。

And only after that did they give him the resources, the chance for DeepMind to move like a startup.

只有在那之后，公司才给了他资源，给了 DeepMind 像创业公司一样往前跑的空间。

Sergey Brin joined in.

Sergey Brin 也加入进来。

And Google really is moving very fast.

Google 确实跑得非常快。

Even so, Google is a giant corporation.

即便如此，Google 还是一家巨型公司。

And with a hegemon — this is a classic pattern not just in business but globally, with countries too, with a hegemon.

而霸主——这不只是商业里的经典规律，放到全球、放到国家层面也一样，霸主都有这个毛病。

A hegemon is usually blind, because, well, that's exactly what makes it a hegemon — like the United States of America.

霸主通常是盲的，因为它之所以成为霸主就在于此，就像美国那样。

And in any case there's the problem of size.

而且不管怎样，都存在体量的问题。

Google simply can't grab every single low-hanging thing out there.

Google 不可能把所有低垂的果实都一网打尽。

And that's what it shows — the institution of startups didn't appear in 2026.

这也正说明了——创业公司这套体制不是 2026 年才出现的。

The institution of startups has been around for many decades.

创业公司这套体制已经存在好几十年了。

And thanks to it, both Google and Facebook managed to stay afloat and become these giant companies, because they bought startups at the right moment, at exactly the right point in time, and that's how they stayed afloat.

正是靠着它，Google 和 Facebook 才能一路挺住、长成这样的巨头，因为它们在关键时刻买下了创业公司，踩在正确的时点上，从而站稳了脚跟。

Google bought YouTube at the right time, and Facebook bought Instagram and WhatsApp at the right time.

Google 在正确的时间买了 YouTube，Facebook 在正确的时间买了 Instagram 和 WhatsApp。

And there are tons of examples like that.

这类例子一大堆。

What you need to understand is that startups and big corporations work in symbiosis.

要理解的是，创业公司和大公司之间是共生关系。

That's why professional venture capitalists rarely ask us whether we're afraid of Google.

所以真正专业的风险投资人很少问我们怕不怕 Google。

They usually think in categories — they know who our competitors in the category are, and they compare us against them.

他们通常按赛道来看——他们清楚这个赛道里跟我们竞争的是谁，然后拿我们跟这些人比。

Because further down the line, the strategy is that companies get bought out.

因为再往后，按这条路径走，公司本来就是要被收购的。

Take Adobe — it's a giant company, it's a great business, Adobe.

就说 Adobe，那是家巨型公司，Adobe 是一门了不起的生意。

If you look at their P&L — it's a public company — if you look at their balance sheet, you'll see they have 20 billion in ARR.

你去看他们的 P&L，它是上市公司，你去看他们的资产负债表，会发现他们有 200 亿的 ARR。

Adobe's gross margin is somewhere around 88-89%.

Adobe 的毛利率大概在 88% 到 89%。

Which means it's a giant cash machine.

也就是说，这是一台巨型现金机器。

Out of their $20 billion, they have nowhere to spend that money, nowhere to deploy it.

手里躺着 200 亿美元，他们没地方花，没地方投出去。

The biggest chunk of where they put their cash is buying back their own shares.

他们的现金大头花在哪儿？回购自家股票。

So buying a startup is not a problem for them at all - buying Higgsfield for, say, a few tens of billions of dollars is a decision that wouldn't even be hard for them to make, because they spend tens of billions on buybacks of their own stock, meaning they literally don't know where else to put that cash.

所以买家创业公司对他们根本不算事——花几百亿美元把 Higgsfield 买下来，这种决定他们做起来一点都不难，因为他们本来就拿几百亿去回购自家股票，说白了就是不知道这笔现金还能往哪儿放。

That's how enormous Adobe is, and what a great company it is.

Adobe 就是这么庞大的一家公司，这么伟大的一家公司。

That's also why startups like Higgsfield raise such big rounds - because venture capitalists understand that there's a huge M&A market, that a huge number of these giant businesses don't know where to spend their cash.

所以像 Higgsfield 这样的创业公司才融得到这么大的轮次——因为风投们清楚，M&A 市场极其庞大，一大批这样的巨头根本不知道现金该往哪儿花。

And at the same time the venture capitalists have big AUMs, big pools of capital, so they can grow really large companies.

与此同时，风投手里的 AUM 也大，资本池也大，足以把公司养到非常大的体量。

That's exactly the beauty of the institution of venture financing, which has proven how effective it is.

这正是风险投资这套制度的妙处，它已经证明了自己的效率。

If you look at the top 10 companies in the S&P, most of them will be venture-backed companies.

你去看标普前十大公司，大部分都是风投培育出来的。

It really is a great institution.

这确实是个很棒的制度。

China, for example, has already imported it.

比如中国，早就把它引进去了。

And thanks to Magzhan, thanks to Astana Hub, we're importing it into Kazakhstan.

而靠着 Magzhan，靠着 Astana Hub，我们也在把它引进哈萨克斯坦。

Right now it's the most effective institution of a market economy there is.

到今天为止，它是市场经济里最有效率的一套制度。

Thank you.

谢谢。

Colleagues, let's do maybe three more questions at most.

各位，我们最多再提三个问题吧。

The session has been running for two hours already, so we'll start wrapping up.

这场对谈已经进行两个小时了，我们慢慢收尾。

Hello, my name is Asanali?, and I have a question for Yerzat.

您好，我叫 Asanali?，我有个问题想问 Yerzat。

I want to ask about vibe coding.

我想问的是 vibe coding。

So on your technical side, vibe coding is used as one of the main tools.

在你们的技术这块，vibe coding 是主要工具之一。

How do you cover security and the weak, well, the weak spots?

那安全性、还有那些薄弱环节，你们是怎么堵上的？

And one more question.

还有一个问题。

Are those same tools used in marketing or the legal side inside Higgsfield itself?

在 Higgsfield 内部，市场和法务这些部门也用同样的工具吗？

As I understand it, you're a dogfooding? company?

我理解你们是一家 dogfooding? 公司？

Sorry, I didn't catch the second question.

不好意思，第二个问题我没听清。

Do you use those tools in marketing and legal inside Higgsfield itself?

在 Higgsfield 内部，你们在市场和法务上也用这些工具吗？

As I understand it, you're a dogfooding? company.

我理解你们是一家 dogfooding? 公司。

What kind?

什么类型的？

Dogfooding?.

Dogfooding?。

What does that mean?

这是什么意思？

Well, fast-growing.

就是发展很快的那种。

Ah, okay.

啊，好。

Right, so, at Higgsfield we don't vibe code.

对，我们在 Higgsfield 不搞 vibe coding。

Globally we have really strong engineers, and they still code more than they vibe code.

我们在全球范围招到的工程师都非常强，他们还是写代码的时候多，vibe coding 的时候少。

Vibe coding is probably more about someone who isn't an engineer building a product quickly.

vibe coding 大概更适合那种本身不是工程师、但要把产品快速搭出来的人。

But today's LLMs have seriously boosted productivity for senior engineers too.

但现在的大模型，对资深工程师的生产力同样有很大提升。

That's undeniably true.

这一点毫无疑问。

And vibe coding is a great tool - in fact, we're hiring vibe coders right now.

而 vibe coding 是个很棒的工具，我们现在反过来还在招 vibe coder。

It's a new kind of specialist: they're not classic programmers or engineers, they have a different mindset, a different approach.

这是一类新的专才：他们不是传统意义上的程序员、工程师，思维方式不一样，路数也不一样。

In some ways they have more - well, probably fewer - boundaries in their heads than a classic senior engineer.

他们脑子里的框框，某些地方更多——或者说，大概是更少，比传统的资深工程师少。

Obviously, to build a scalable large product for tens of millions of users, vibe coding is probably still a bit raw as an approach, but for prototyping, for new products, vibe coding is great.

当然，要做一个能扛住几千万用户的可扩展大产品，vibe coding 作为方法可能还有点糙；但拿来做原型、做新产品，vibe coding 非常好用。

So we're looking for vibe coders, we're hiring vibe coders.

所以我们在找 vibe coder，我们在招 vibe coder。

We were one of the first on the market to hire prompt engineers.

当年我们也是市场上最早招 prompt engineer 的公司之一。

That's similar to vibe coding as well.

那件事跟 vibe coding 也很像。

Just with different tools - ComfyUI, media-gen AI.

只是工具不一样——是 ComfyUI，是媒体生成类 AI。

That kind of tool does exist.

这类工具是有的。

And we're actively looking for vibe coders right now, exactly for that new type of product, for prototyping.

vibe coder 我们现在也在积极找，就是为了那种新类型的产品，为了做原型。

There you go.

就是这样。

And for everything else too, of course - for marketing, for legal.

当然其他事情上也用，市场、法务都用。

Personally, I don't have a legal education.

我个人是没有法律背景的。

And obviously modern language models let me ask lawyers better questions and get up to speed faster on a large volume of information.

现在的语言模型确实让我能向律师问出更到位的问题，也能更快地把一大堆信息消化掉。

And in marketing too, naturally, we use generative AI.

市场那边我们自然也在用生成式 AI。

But at the same time, globally, this is kind of an inside joke at the company.

但同时，整体上，这也算我们公司内部的一个梗。

Often when new people join us, the first thing they hear from me - say someone comes into marketing - is that I ask them to do outreach.

经常有新人进来，比如有人加入市场团队，他从我这儿听到的第一件事，就是让他去做 outreach。

Meaning writing cold emails, writing cold messages, in large volumes.

也就是大量地写陌生开发邮件、写陌生开发的私信。

And obviously the first thing that comes to mind is to use generative AI and some LLMs.

显然，人第一反应就是去用生成式 AI、用大模型。

And I often ask them not to do that - I ask the person to write by hand, to really live through the process and iterate on it as much as possible.

而我经常要求他们别这么做——我要求他亲手写，真正把这个过程走一遍，尽可能多迭代几轮。

Because you have to understand that if you're doing outreach, a huge number of other companies, startups and businesses competing for that same outreach are most likely all using LLMs.

因为你得明白，你在做 outreach，那一大堆跟你抢同一批人的公司、创业团队、企业，八成全都在用大模型。

And to stand out - if it's written by hand, that alone will make you stand out in that crowd.

想脱颖而出，只要是手写出来的，光这一点就能让你从这堆人里跳出来。

So I often ask employees to do the opposite: use less automation and more of that natural intelligence.

所以我经常反过来要求员工少用自动化，多用点这种天然智能。

Hello.

您好。

By the way, yesterday I got a notification that there'd be a fireside chat today.

顺便说一句，我昨天收到通知，说今天有 fireside chat。

I bought a ticket and just flew in from Shymkent.

我买了机票，刚从 Shymkent 飞过来。

Well done.

厉害。

From a push notification in the Astana Hub app.

是看到 Astana Hub App 的推送通知吧。

Yes, yes, yes.

对对对。

Download the Astana Hub mobile app.

大家都去下载 Astana Hub 的手机 App。

It's very convenient.

非常方便。

Well done.

厉害。

Yes.

对。

My name is Vener.

我叫 Vener。

And here's the question I actually flew in for.

我这次飞过来，其实就是为了问这个问题。

This one is addressed to Murat.

这个问题是问 Murat 的。

I'm from Shymkent? myself.

我自己是 Shymkent? 人。

Right now we're building a B2B platform for sourcing construction materials and services.

我们现在在做一个 B2B 平台，用来找建材和相关服务。

At this point we already have a website, bots, and a chat.

目前网站、机器人、在线聊天都已经有了。

And while we were still coding, we were testing at the same time — during the tests we connected one of the suppliers and did $600,000 in sales.

而且我们一边开发一边测试，测试期间接进来一家供应商，就做出了 60 万美元的销售额。

So here's my actual question: at what stage could I come to you so that you'd invest in us?

所以我真正想问的是：我要做到哪个阶段，才能来找你们，让你们投我们？

When and how?

什么时候找、怎么找？

Well, we'd need your MRR to be somewhere around at least $30,000.

嗯，我们需要你们的 MRR 至少做到 3 万美元左右。

Mm-hmm.

嗯。

And traction — something like three to six months, so we can see the trend.

还要有增长势头，比如说三到六个月吧，这样能看出走势。

Yeah, yeah.

对对。

Basically we just need to understand whether you have market fit or not.

就是说，我们只需要判断出你们到底有没有 market fit（产品与市场的契合）。

You see?

明白吗？

Yeah, roughly like that.

对，大概就是这样。

Right.

好的。

So I need to do this through your email, the one that— Yes, you can send it to inbox@ma7.vc.

那是不是要发到你们的邮箱，就那个—— 对，可以发到 inbox@ma7.vc。

We'll definitely reply.

我们一定会回。

It'll land in our CRM.

邮件会进到我们的 CRM 里。

Got it.

好的。

Thank you.

谢谢。

Testing.

喂。

Testing.

喂。

Send me your pitch on Telegram and I'll forward it to Murat.

把你的 pitch 发我 Telegram，我转给 Murat。

For flying all the way in from Shymkent.

就冲你从 Shymkent 大老远飞过来。

Well done.

好样的。

Can I send mine too?

我也能发吗？

Sure, go ahead — I'll definitely take a look.

行，尽管发，我肯定会看。

The question first.

先说问题。

Good afternoon.

下午好。

Thank you very much.

非常感谢。

I have a question.

我有个问题。

When a startup raises investment, it has to explain to a person — well, to the investor and to people — what the thing actually is.

创业公司拿投资的时候，得跟人解释清楚——跟投资人、跟大家——自己到底是做什么的。

So this is a question about documentation.

所以这其实是个关于文档的问题。

Given how fast everything moves, how does documentation work, so that when a new person joins, they don't get handed a 30-page PDF about what we do — how do you adapt it quickly so that the communication problem is closed, and everyone understands and is in the loop, roughly speaking?

考虑到发展节奏这么快，文档是怎么做的？比如新人进来，不至于甩给他一份 30 页的 PDF 讲我们在做什么——怎么让它快速跟上，把沟通这件事解决掉，让所有人都明白、都跟得上，大致是这个意思。

How does that work at your company?

你们那边是怎么做的？

The question is for me.

这个问题是问我的。

Yes.

对。

Actually, I didn't quite get the question — is it about how to pitch an investor, or how to onboard a new employee?

不过我没太听懂问题——是问怎么跟投资人 pitch，还是问新员工的 onboarding 怎么做？

Well, let's say, hypothetically, here's a situation.

打个比方，假设有这么个情况。

Something went wrong — say, a bad deploy on the server — and you figured out how to fix it.

出了个岔子，比如往服务器上部署错了，然后你们想明白了该怎么解决。

And usually you do some kind of review so that mistakes like that don't happen in the future.

通常这时候会做个复盘，免得以后再犯同样的错。

And when a newcomer joins, they're told about that mistake right away, because it might be some non-obvious pattern in behavior, something like that.

然后新人一进来，马上就会被告知这个错误，因为这可能是某种不太显眼的行为模式之类的东西。

How is that documented, so that in the future people understand it and don't repeat the mistakes that have already happened?

这些是怎么记录下来的，好让后来的人明白，不再重犯已经犯过的错？

Okay, I get it.

好，我明白了。

So it's an onboarding question then.

那这就是个 onboarding 的问题。

Broadly speaking, at Higgsfield our onboarding is pretty tough.

整体上说，Higgsfield 的 onboarding 相当狠。

When a new employee joins, we like to throw them into the deepest end we can.

新员工一来，我们喜欢直接把他扔到最难的火线上。

Meaning we rarely give them any documentation at all.

也就是说，我们基本不会给他任何文档。

What we'll give them instead is most likely a very big, critically important task, and we'll see how they handle it.

我们大概率会直接给他一个很大、又非常关键的任务，看他能不能扛下来。

When a new employee joins, it's really important to understand as fast as possible, in a very short time, whether they're a fit with the company or not.

新人进来的时候，最重要的是在尽可能短的时间里判断出，他跟公司到底合不合。

Right.

就这样。

Yeah.

嗯。

And, you know, we even had a case where a new employee joined, worked for one day, and at the end of that day we told him we were disappointed in him — even though he'd only worked one day.

所以我们甚至出过这种事：一个新员工来上班，干了一天，当天下班时我们就跟他说，我们对他很失望——他才干了一天。

Today he's a top specialist here, one of the best, but he loves to remind us how on his very first day we told him we were disappointed in him.

现在他是我们这儿数一数二的顶尖高手，但他老爱翻这笔旧账，说我们第一天就跟他讲对他很失望。

Yeah.

嗯。

So the best way to do onboarding is to throw the person into the deep end as much as you can.

所以 onboarding 最好的办法，就是尽可能把人扔到火线上去。

The documentation question is solved by the fact that the whole company is split into very small teams, and inside each team people coordinate themselves and keep their own documentation.

文档这件事，靠的是把整个公司拆成一个个很小的团队，每个团队内部自己协调、自己维护文档。

We don't have some giant global documentation for the entire company.

我们没有那种覆盖全公司的庞大统一文档。

Right. — And what do you work by?

就这样。—— 那你们是按什么方式在做？

By Scrum?

按 Scrum 吗？

Honestly, I don't know what that is.

说实话，我不知道那是什么。

I see.

明白了。

All right.

好。

Okay, that's it — the last one.

好，就到这儿了，最后一个。

Well, the young lady — two more questions.

那位女士，再来两个问题吧。

And let's wrap up with that.

咱们就在这儿收尾。

Hello, my name is Artur.

您好，我叫 Artur。

Thank you so much for today's chat.

非常感谢今天这场对谈。

I have a question for Yerzat.

我有个问题想问 Yerzat。

Yerzat, I've been following you for a very long time.

Yerzat，我关注你很久了。

Thank you so much.

非常感谢你。

You're literally that phrase, "You're living my dream."

你就是那句话的活样本——"You're living my dream"。

That's you.

说的就是你。

And my question is this: could you please tell us about your first sales?

我的问题是：能不能讲讲你们最早的那几笔销售？

Right now Higgsfield's sales system is basically clear.

现在 Higgsfield 的销售体系大致是看得明白的。

What I'd like to know is how Higgsfield built up its first sales right after the MVP was created.

我想知道的是，MVP 做出来之后，Higgsfield 是怎么跑出第一批销售的。

Was it advice from, well, from the book Marketing Warfare, or did you apply some kind of magic, roughly speaking, that moved Higgsfield a step or two forward — not to the global level yet, but a step or two, let's say.

是照着《营销战》（Marketing Warfare）那本书里的建议做的，还是说你们用了某种“魔法”，把 Higgsfield 往前推了一两步——先不说到世界级，但至少走出了一两步。

Thank you.

谢谢。

We spent a very long time discussing the go-to-market question.

GTM 这个问题，我们前面已经聊了很久。

It was all covered.

该讲的都讲透了。

The whole recording will be on YouTube.

完整录像之后会放到 YouTube 上。

Well, Yerzat, maybe literally a couple of words, but we went through this at length here — yes.

那 Yerzat 你就真的说两句吧，不过这块我们在这儿已经拆得很细了——是吧。

This is about what I talked about earlier, about go-to-market, how we did it.

这就是我前面讲过的，关于 GTM，我们当时是怎么做的。

You'll be able to go back and watch that.

那段回头可以再看一遍。

We covered that topic really well.

这个话题我们讲得非常充分了。

Mm-hm.

嗯。

Uh, hello, my name is— I feel like that boom-barrier director.

您好，我叫——我现在感觉自己像那个管道闸的主任。

Uh, hello, my name is Mosmazhan?.

您好，我叫 Mosmazhan?。

I've been into startups since I was 15, I've worked in IT for 16 years, and I'm 18 now.

我从 15 岁起就迷上了创业，在 IT 行业干了 16 年，现在 18 岁。

So, I have a slightly banal question.

我想问一个可能有点老套的问题。

What do you think about EdTech startups?

你们怎么看 EdTech 这类创业公司？

Is there potential there over the next five years?

未来五年这个方向还有潜力吗？

It's probably going to be a fairly general question.

这个问题问得可能比较笼统。

Thank you.

谢谢各位。

That's Education — well, EdTech is lying low right now, isn't it?

这属于教育领域——EdTech 现在正处在蛰伏期，对吧？

So what does that mean?

那这意味着什么？

It means there are hundreds, thousands of online schools out there, and their valuations and multiples have come down to almost 1x.

意味着市面上有成百上千家在线学校，而它们的估值和倍数几乎已经掉到了 1 倍。

That is, the valuation equals roughly 1x revenue.

也就是说，估值大概等于收入的 1 倍。

And that means, roughly speaking, about 5x EBITDA.

换算过来，大致相当于 5 倍 EBITDA。

It means it gets valued like an ordinary business.

意思就是，它被当成一门普通生意来估值。

That's the situation today for all EdTechs, except EdTechs that actively use AI.

今天所有 EdTech 都是这个处境，除了那些真正把 AI 用起来的。

There the multiples are still somewhere between 15 and 30.

那一类的倍数依然在 15 到 30 之间。

That's everything you need to know about EdTech today.

关于今天的 EdTech，你要知道的就这些。

If you don't have AI in there, it's just an ordinary business.

如果里面没有 AI，那它就只是一门普通生意。

And venture — you can't even call it a venture case anymore.

至于风险投资——它已经称不上是风投标的了。

And VCs aren't interested in it without AI.

没有 AI，VC 根本不感兴趣。

Yerzat, do you want to say a couple of words about EdTech?

Yerzat，EdTech 这块你想说两句吗？

Well, yes, if we're talking about EdTech, then everything Murat said is super important information that you need to understand.

说到 EdTech，Murat 刚才讲的都是非常重要的信息，你们得吃透。

The one thing is, EdTech is a great space in the sense that these days any distribution comes down to content.

唯一想补充的是，EdTech 是个很棒的方向，因为现在任何分发本质上都是内容。

If you know how to make great educational content about some technology product in some new field, that's a really serious force.

如果你会围绕某个新方向的技术产品做出优质的教育内容，那是一股相当强的力量。

If you can, for example, build a YouTube channel with good reach about some technological, innovative thing, that's a very strong asset.

比如说，你能做一个 YouTube 频道，讲某项技术创新，还能有不错的触达量，那就是一项非常强的资产。

And it's worth a lot — we work with tech bloggers a lot, and a blogger who averages 50,000 views with a US audience can charge around $100,000 for a single integration.

这种资产非常值钱——我们经常和科技博主合作，一个面向美国观众、平均播放量五万的博主，一条视频的植入就能报到十万美元左右。

And that's not some mega-large channel.

而这还算不上什么超大号频道。

I mean bloggers who talk about, say, Anthropic's Claude to an average audience of 50,000 people — that is, 50,000 views on a single video.

我说的就是那种讲 Anthropic 的 Claude 之类的博主，平均受众五万人，也就是单条视频五万播放。

Those are actually totally achievable numbers.

这个量级其实完全够得着。

It's not millions of views, and even a blogger like that charges $100,000 for one integration.

不是几百万播放，就这种体量的博主，一次植入也要十万美元。

And we, for example, we're working with bloggers like that right now.

我们现在合作的就是这样的博主。

And those numbers just really impressed me.

这些数字确实让我挺震撼。

And there's nothing hard about starting a channel like that.

而做这么一个频道其实没什么难的。

All you need is the internet and access to YouTube, some minimal setup, and at most some curiosity about AI.

你需要的只是网络和能上 YouTube，一套最简配的设备，顶多再加上对 AI 的一点好奇心。

And you can build a tech channel about AI — obviously you need to run it in good English for a US audience.

然后你就能做一个讲 AI 的科技频道——当然，得用过得去的英语来做，面向美国观众。

And advertisers will simply swarm you.

广告主会直接把你围起来。

We literally fight over them.

我们真的是在抢人。

Take a blogger with 50,000 views — we're the ones talking him into taking our $100,000.

一个只有五万播放的博主，反倒要我们去劝他收下这十万美元。

We can immediately offer him something like a $4 million annual contract, and he might not even answer our message, because he's got another 30 startups writing to him.

我们可以当场开出四百万美元的年度合同，他甚至可能不回我们消息，因为还有三十家创业公司在找他。

What you have to understand is that the market is so overheated, that there's a huge amount of investment going into AI startups like Higgsfield.

你要明白的是，市场就是这么过热，像 Higgsfield 这样的 AI 创业公司拿到了海量投资。

And then all that money has to go into GTM, into distribution.

而这些钱最后都得投进 GTM、投进分发。

And that's exactly where it goes — to YouTube bloggers, to the top educational influencers.

它们流向的正是 YouTube 博主，是头部的教育类影响者。

It's a fairly large niche.

这是个相当大的细分市场。

And the barrier to entry there — you can do it from Kazakhstan and comfortably earn serious cash there, especially by our standards.

至于门槛——你在哈萨克斯坦就能做，而且能轻松挣到一笔可观的现金，尤其按我们这儿的标准。

Yes, thank you for the answer.

好，谢谢你的回答。

And with that we're wrapping up the Q&A session, right?

Q&A 环节我们就到这里，对吧？

And to close out today's fireside chat, I want to once again urge our audience, our founders and IT specialists, to keep learning constantly and to launch startups.

在今天这场 fireside chat 的最后，我想再一次呼吁在座的观众、创始人和 IT 从业者：持续自学，去做创业。

Obviously it's going to be very hard, you'll run into a mountain of obstacles, but today we're doing everything we can to improve your chances of success.

当然，这条路会非常难，你们会遇到一堆障碍，但我们今天做的一切，就是为了把你们成功的概率抬上去。

So thank you for the fact that we now have a role model like Higgsfield.

所以要感谢的是，我们终于有了 Higgsfield 这样一个榜样。

In the person of Yerzat, of the whole Higgsfield team.

有 Yerzat，有整个 Higgsfield 团队。

Thank you that we have venture investors like Murat Abdrakhmanov.

感谢我们有 Murat Abdrakhmanov 这样的风险投资人。

And thank you very much for taking part today.

也非常感谢各位今天到场参与。

Given that for them? every hour of speaking is worth a great deal.

毕竟对他们来说，每一个小时的分享都非常宝贵。

And I'm really glad we have a big live audience here that genuinely values information this valuable and experience this valuable.

我也特别高兴，今天现场坐着这么多观众，是真的珍惜这么有价值的信息、这么有价值的经验。

So subscribe to all our social media, apply to all our programs, and go for it.

所以请关注我们所有的社交账号，报名我们所有的项目，放手去做。

We'll always be there to support you.

我们会一直支持你们。

Thank you very much.

非常感谢。

Thank you so much to our distinguished speakers, and to Magzhan for the brilliant questions and the brilliant answers.

非常感谢各位嘉宾，也感谢 Magzhan，问题问得精彩，回答也精彩。

I think you could write a book off today's session, a proper handbook, and sell it by the million?.

我觉得就凭今天这场对谈，完全可以写成一本书、一本实用手册，然后卖上几百万册。

Thank you to everyone who took part.

感谢所有参与的人。

That wraps up today's fireside chat.

今天的 fireside chat 到此结束。

In just a moment, uh, we'll put up a QR code for feedback.

接下来会给大家展示一个反馈用的二维码。

We'd be very grateful if you answered a few questions and shared your thoughts, so that we can keep making events like this better, better, and bigger.

如果你们能回答几个问题、说说自己的看法，我们会非常感激——这样我们才能把这样的活动办得越来越好、越来越多。

Thank you, everyone.

谢谢大家。

Dear guests, we invite you to join us for a group photo.

各位嘉宾，请一起来合张影。

Take a chance with me.

跟我冒一次险吧。

Will you dance with me?

愿意跟我跳支舞吗？