70 人、12 个月、$250M run rate
"In 12 months, a team of roughly 70 people in Palo Alto, California, built an all-in-one AI workspace suite, with AI generating nearly 100% of the code."
"…and reached a $250 million annual run rate. In five months, they signed more than 5,000 business clients."
近 100% 代码由 AI 生成,5 个月签下 5,000+ 企业客户——这组数字是全文 AI-native 组织论的事实底座。
AI 不是工具,是 workforce 的一部分
"The businesses pulling ahead today no longer treat AI as a tool, but as part of the workforce."
"…one that can work alongside employees around the clock at 10 to 100 times the efficiency of traditional workflows."
这是"AI-native 组织"的定义性判断:AI 全天候与员工并肩工作,效率是传统工作流的 10-100 倍。
评估 AI 正在变成一份全职工作
"Most knowledge workers we talked to were spending more time evaluating AI than using it."
"But as AI exploded—new models, new tools, new capabilities every few weeks—a new problem emerged."
本该提效的东西成了新负担。Genspark 的定位由此而来:AI 仓储会员店,一张会员卡汇聚最好的模型和工具。
搜索的终点是办成事,不是找信息
"People do not search because they want more information. They search because they want to get something done."
"I spent 20 years in search, beginning at Microsoft and helping launch Bing. What I learned is simple."
20 年搜索行业经验沉淀成 Genspark 的根基:构建帮人完成工作的 AI,而不只是帮人找信息。
AI specialist 是天花板,不是职位
"It is not a job title so much as a new ceiling on what one person, working alongside AI, can do."
"Not just engineers, but marketers, consultants, salespeople and operators who can now accomplish work that once required much larger teams."
未来最值钱的员工不是在工位上坐得最久的,而是适应最快、最会把 AI 变成业务 leverage 的。
最难的不是技术,是 adoption
"Getting an entire team to genuinely change how they work is harder than building the technology. It's a leadership problem, not a software problem."
"I can deploy a powerful system to a 5,000-person company on Monday, and by Friday only a few hundred employees are actually using it well."
他坦承的三大未解难题:token 经济账、数据边界、adoption——且把 adoption 列为最难的一个。
选 AI 要看代际,老工具过时极快
"Two products may both be called AI but belong to very different generations."
"Older tools can become last-generation systems surprisingly fast. That is simply the nature of the AI cycle."
企业从试验走到收入,要靠领导层亲手用 + 判断采用哪一代 AI + 全员铺开,AI 才能变成贯穿业务的能力层。
Building a company designed around AI as part of the team, and what it means for how the next generation of work gets done.
把 AI 当作团队一员来设计公司,这对下一代工作如何完成意味着什么。
Genspark's trajectory offers a glimpse of what AI is making possible.
Genspark 的发展轨迹,让人得以一窥 AI 正在创造的可能性。
In 12 months, a team of roughly 70 people in Palo Alto, California, built an all-in-one AI workspace suite, with AI generating nearly 100% of the code, and reached a $250 million annual run rate.
12 个月里,一支位于加州 Palo Alto、约 70 人的团队,做出了一套 all-in-one 的 AI 工作空间套件——近 100% 的代码由 AI 生成——并达到了 $250M 的年化收入 run rate。
In five months, they signed more than 5,000 business clients.
5 个月里,他们签下了超过 5,000 家企业客户。
For Co-Founder and CEO Eric Jing, those numbers point to something bigger than one company's success.
在联合创始人兼 CEO Eric Jing(景鲲)看来,这些数字指向的东西,比一家公司的成功更大。
The businesses pulling ahead today no longer treat AI as a tool, but as part of the workforce, one that can work alongside employees around the clock at 10 to 100 times the efficiency of traditional workflows.
如今领先的企业,已经不再把 AI 当工具,而是当作 workforce 的一部分——它可以全天候与员工并肩工作,效率是传统工作流的 10 到 100 倍。
Jing calls this an AI-native organization, and he believes it represents the next phase of how businesses will be built and run.
Jing 把这称为 AI-native 组织。他相信,这代表着企业构建与运营方式的下一个阶段。
The AI landscape changes constantly. How did that shape the way you designed Genspark?
Q · AI 格局变化不停,这如何塑造了你设计 Genspark 的方式?
Eric Jing: I spent 20 years in search, beginning at Microsoft and helping launch Bing.
Eric Jing:我在搜索行业做了 20 年,从 Microsoft 起步,参与了 Bing 的发布。
What I learned is simple: People do not search because they want more information.
我学到的道理很简单:人们搜索,不是因为想要更多信息。
They search because they want to get something done.
他们搜索,是因为想把事情办成。
That insight became the foundation of Genspark.
这个洞察成了 Genspark 的根基。
We are building AI to help people complete work—not just help them find information.
我们在构建的 AI,是帮人完成工作的——不只是帮人找信息。
But as AI exploded—new models, new tools, new capabilities every few weeks—a new problem emerged.
但随着 AI 爆发——每隔几周就有新模型、新工具、新能力——一个新问题出现了。
Most knowledge workers we talked to were spending more time evaluating AI than using it.
我们聊过的大多数知识工作者,花在评估 AI 上的时间,比真正用 AI 的时间还多。
The very thing meant to make people more productive was becoming another full-time job.
这个本该让人更高效的东西,正在变成另一份全职工作。
That complexity is becoming a job of its own—and it's not a problem most people should have to solve.
这种复杂性正在变成一份独立的活儿——而它本不该是大多数人需要去解决的问题。
We track which models perform best for which tasks, orchestrate them behind the scenes and deliver results that any knowledge worker can use without needing technical expertise.
我们追踪哪个模型在哪类任务上表现最好,在幕后做编排,交付任何知识工作者都能直接用的结果,不需要技术背景。
Think of Genspark as an AI warehouse club: one membership, one place that brings together the best models, tools and capabilities in a single workspace.
可以把 Genspark 想成一家 AI 仓储会员店(warehouse club):一张会员卡,一个地方,把最好的模型、工具和能力汇聚到同一个工作空间里。
Instead of constantly evaluating vendors, switching platforms or becoming an AI expert yourself, you can simply describe what you need and get the work done.
你不用再没完没了地评估供应商、切换平台、把自己逼成 AI 专家——只要描述你需要什么,把工作做完。
How are AI agents changing what an individual employee can actually accomplish?
Q · AI agent 正在如何改变一个员工实际能完成的事?
Jing: What holds many people back is not a lack of interest in AI.
Jing:拦住很多人的,不是对 AI 缺乏兴趣。
It is the feeling that AI is hard to learn, changing too quickly and difficult to trust.
而是那种感觉:AI 难学、变得太快、难以信任。
Questions around data security are valid.
围绕数据安全的疑问是站得住脚的。
As a result, most people still use AI in a very limited way, as a faster search engine or a basic chatbot.
结果是,大多数人用 AI 的方式仍然非常有限——当作一个更快的搜索引擎,或一个基础的 chatbot。
But the technology has already moved far beyond that.
但技术早已走到远比这更前面的地方。
Today, anyone can use a phone to direct AI agents to work on their behalf 24/7.
今天,任何人都可以用一部手机,指挥 AI agent 24/7 替自己干活。
A product like Genspark Claw makes that shift easier to picture: a cloud-based agent with its own computer environment that can keep working in the background, more like a digital operator than a chatbot.
像 Genspark Claw 这样的产品,让这种转变更容易想象:一个云端 agent,拥有自己的计算机环境,可以在后台持续工作——更像一名数字操作员(digital operator),而不是一个 chatbot。
Instead of asking a quick question and getting a quick answer, you can assign an agent to research a market, analyze competitors, build recommendations and carry out complex tasks while you focus on higher-value decisions.
你不再是问一个快问题、拿一个快答案,而是可以把任务派给 agent:研究一个市场、分析竞争对手、形成建议、执行复杂任务,而你专注在更高价值的决策上。
When you come back, the work has already progressed.
等你回来时,工作已经往前推进了。
That is a fundamentally different way of working.
这是一种根本不同的工作方式。
That is how AI begins to extend individual reach, and how an AI-native organization takes shape—not from a top-down redesign but from the bottom up, as individuals learn to work with AI as part of their team.
AI 就是这样开始延伸个人的能力边界,AI-native 组织也是这样成形的——不是靠自上而下的重新设计,而是自下而上:每个人学会把 AI 当作自己团队的一部分来协作。
A new kind of role emerges in the process: the AI specialist.
在这个过程中,一种新角色出现了:AI specialist。
Not just engineers, but marketers, consultants, salespeople and operators who can now accomplish work that once required much larger teams.
不只是工程师,还有营销人、咨询顾问、销售和运营者——他们现在能完成过去需要大得多的团队才能完成的工作。
It is not a job title so much as a new ceiling on what one person, working alongside AI, can do.
与其说这是一个职位头衔,不如说是一个人与 AI 并肩工作时,能做之事的新天花板。
The most valuable employees going forward will not be the ones who spend the most hours at their desks.
往后最有价值的员工,不会是在工位上耗时最长的那批人。
They will be the ones who adapt fastest and turn these systems into real leverage for the business.
而是适应最快、能把这些系统变成业务真实 leverage 的那批人。
Where is AI still falling short? What hasn't worked yet?
Q · AI 还在哪些地方不够用?哪些事还没做成?
Jing: I'll be honest about what's still hard.
Jing:我坦白说说现在还难在哪。
Three things, in particular.
具体是三件事。
First, the economics.
第一,经济账。
Token costs and ROI don't always line up.
Token 成本和 ROI 并不总是对得上。
The most capable models are also the most expensive, and not every workflow justifies the price.
能力最强的模型也最贵,而不是每条工作流都值这个价。
Getting the math right, knowing when to use a frontier model versus a lighter one, is a real engineering and product discipline, not a given.
把这笔账算对——知道什么时候用 frontier 模型、什么时候用轻量模型——是一门实打实的工程与产品功夫,不是天然就会的。
Second, data boundaries.
第二,数据边界。
Every enterprise leader I talk to asks the same question: Where does my data go, and could it end up training someone else's model?
我聊过的每一位企业负责人都在问同一个问题:我的数据去了哪里?会不会最终被拿去训练别人的模型?
That concern is legitimate.
这个担忧是正当的。
The industry is still establishing the norms, contracts and technical guarantees that make this verifiable, not just promised.
整个行业还在建立相应的规范、合同和技术保障,让这件事可验证,而不只是口头承诺。
And third, the hardest one, adoption.
第三,也是最难的一个:adoption(落地采用)。
The tools have moved faster than the people using them.
工具跑得比用工具的人快。
I can deploy a powerful system to a 5,000-person company on Monday, and by Friday only a few hundred employees are actually using it well.
我可以在周一给一家 5,000 人的公司部署一套强大的系统,到周五,真正用得好的员工只有几百人。
Getting an entire team to genuinely change how they work is harder than building the technology.
让整个团队真正改变工作方式,比把技术做出来更难。
It's a leadership problem, not a software problem.
这是领导力问题,不是软件问题。
We're learning, in real time, where AI is ready to be trusted and where the gaps still are.
我们在实时学习:AI 在哪里已经值得信任,哪里还有缺口。
At Genspark, we've built our platform around all three problems: routing across the full lineup of frontier models to balance capability with cost, designing an enterprise-ready product with security in mind from day one, and obsessing over a user experience simple enough that adoption tends to happen on its own.
在 Genspark,我们的平台就是围绕这三个问题构建的:在整个 frontier 模型阵容之间做路由,平衡能力与成本;从第一天起就带着安全意识设计企业级产品;再就是死磕用户体验,简单到 adoption 往往会自己发生。
The technology keeps moving, and none of this is solved forever.
技术还在不停往前走,这些问题没有一个是一劳永逸解决的。
The companies pulling ahead are not the ones pretending otherwise.
跑在前面的公司,不是那些假装问题不存在的。
They are the ones honest enough to name the gaps and disciplined enough to close them.
而是那些诚实到敢把缺口说出来、又自律到能把缺口补上的。
What does it take for enterprises to move from experimenting with AI to embedding it in workflows that actually drive revenue?
Q · 企业要从"试验 AI"走到"把 AI 嵌进真正驱动收入的工作流",需要什么?
Jing: What we have seen is that tangible gains do not happen without leadership involved directly.
Jing:我们看到的情况是:没有领导层的直接参与,实打实的收益不会发生。
AI adoption cannot be delegated to IT or innovation teams alone.
AI adoption 不能只委派给 IT 或创新团队。
Leaders understand where the business is constrained, which workflows matter most and where better execution translates into growth.
领导者清楚业务卡在哪里、哪些工作流最关键、更好的执行在哪些地方能转化成增长。
They have to use these systems firsthand and set the pace by understanding what the tools can actually do.
他们必须亲手用这些系统,靠理解工具真正能做什么来定节奏。
Just as important, companies have to be deliberate about which generation of AI they are adopting.
同样重要的是,公司必须想清楚自己采用的是哪一代 AI。
Two products may both be called AI but belong to very different generations.
两个产品可能都叫 AI,却属于完全不同的世代。
Older tools can become last-generation systems surprisingly fast.
老工具变成"上一代系统"的速度,会快得出人意料。
That is simply the nature of the AI cycle.
这就是 AI 周期的本性。
The companies moving fastest are adopting the newest platforms early and scaling them broadly.
跑得最快的公司,正在尽早采用最新的平台,并大范围铺开。
When leadership pairs that judgment with broad employee adoption, AI stops being a side experiment and starts becoming a capability layer across the entire business.
当领导层把这种判断与全员的广泛使用配在一起,AI 就不再是边缘试验,而开始成为贯穿整个业务的能力层(capability layer)。
For enterprise leaders concerned about security and governance, how does Genspark address those fears?
Q · 对于担心安全与治理的企业负责人,Genspark 怎么回应这些顾虑?
Jing: This issue matters a great deal, but the landscape is improving quickly.
Jing:这个问题非常重要,但整体局面正在快速改善。
Newer AI systems are much better equipped to address enterprise concerns around security and governance.
更新一代的 AI 系统,在应对企业的安全与治理顾虑上,准备已经充分得多。
With products like Genspark Claw, each agent runs inside a dedicated virtual machine, a setup enterprises already understand.
在 Genspark Claw 这类产品里,每个 agent 都跑在一台专属虚拟机里——这是企业本来就熟悉的架构。
That gives organizations familiar ways to monitor activity, apply security software, see how data is flowing and control what the AI can access.
这让组织可以用熟悉的方式监控活动、部署安全软件、看清数据怎么流动、控制 AI 能访问什么。
That makes governance much more practical.
这让治理变得务实得多。
Companies can define what data an agent can use, what tools it can leverage and what actions it is allowed to take.
公司可以定义一个 agent 能用什么数据、能调用什么工具、被允许执行什么动作。
The goal is not uncontrolled autonomy.
目标不是不受控的自治。
It is to give enterprises AI systems they can supervise, secure and integrate into existing governance frameworks with confidence.
而是给企业一套可以放心监督、加固、并整合进既有治理框架的 AI 系统。
Some hesitation remains because AI is evolving so quickly.
一些犹豫仍然存在,因为 AI 演化得太快。
But increasingly, the fear comes more from the speed of change than from a lack of workable controls.
但越来越多时候,恐惧更多来自变化的速度,而不是缺少可用的控制手段。
On security and governance, the newest platforms have made major progress and they are improving fast.
在安全与治理上,最新的平台已经取得重大进展,而且还在快速改进。
What does Genspark's own story reveal about the future of AI?
Q · Genspark 自己的故事,揭示了 AI 未来的什么?
Jing: Genspark's story is still being written, and we are still a small company.
Jing:Genspark 的故事还在书写中,我们仍然是一家小公司。
But I have a concrete basis for comparison.
但我有一个具体的参照系。
Before this, I served as a chief product officer and VP managing organizations of several thousand people.
在这之前,我做过 chief product officer 和 VP,管理过几千人的组织。
Having operated in both environments, I can say firsthand: The structure of work is changing.
两种环境都亲身操盘过,我可以第一手地说:工作的结构正在变。
With AI, small teams can move faster, operate with far more leverage and execute work that once required much larger organizations.
有了 AI,小团队能跑得更快、以大得多的 leverage 运转,执行过去需要大得多的组织才能完成的工作。
Genspark is an early example of what that looks like, not a template every company should copy, but proof of what is possible.
Genspark 是这件事的一个早期样本——不是每家公司都该照抄的模板,而是"什么是可能的"的证明。
That shift is only beginning.
这场转变才刚刚开始。
AI systems are becoming more autonomous, their ability to automate work is rising quickly, and the ways enterprises can govern and control them are growing more sophisticated.
AI 系统正变得更自主,自动化工作的能力快速上升,企业治理和控制它们的手段也越来越成熟。
We are already seeing ordinary knowledge workers operate with 10 to 100 times the effectiveness of traditional workflows.
我们已经看到普通知识工作者,以传统工作流 10 到 100 倍的效能在运转。
What was theoretical a few years ago is happening now.
几年前还停留在理论上的事,现在正在发生。
AI is a revolutionary technology, and no one, including us, yet knows exactly where it leads.
AI 是一项革命性技术,没有人——包括我们——确切知道它会通向哪里。
What I can say is that the early results are already hard to ignore.
我能说的是,早期结果已经让人无法忽视。
An AI-native organization is not defined by when it started, but by how it works.
一个 AI-native 组织的定义,不在于它什么时候成立,而在于它怎么工作。
AI is treated as part of the team, not a tool bolted on the side.
AI 被当作团队的一员,而不是拴在旁边的一个工具。
Structures are flatter, teams leaner, cycles faster.
结构更扁,团队更精,周期更快。
That is the AI specialist in practice, and that kind of organization is already taking shape.
这就是 AI specialist 的实践形态,而这样的组织已经在成形。
Some companies will be built this way from scratch.
有些公司从一开始就会按这种方式构建。
Most will get there by evolving, function by function, team by team.
大多数公司会靠演化抵达——一个职能一个职能、一个团队一个团队地变。
Both paths are already underway.
两条路都已经在路上。
At Genspark, we are building one version of this and making the tools that let organizations of every size build their own.
在 Genspark,我们在构建它的一个版本,同时打造让各种规模的组织都能构建自己版本的工具。
That is the work. That is the era. And it is just beginning.
这就是我们要做的事。这就是这个时代。而它才刚刚开始。