如何(真正)利用AI赚钱 / How to (actually) make money with AI
Source: http://x.com/i/article/2029112829815496704
🌐 中文版在上方 / English version below.
受卡兹克 @Khazix0918 文章的影响,今天整个推特都在聊同一件事——"当今用AI最好赚钱的方法,就是帮人上门安装clawdbot,一次499。"
来源: @Khazix0918 , 《我花499找人上门安装OpenClaw,看到了AI时代最魔幻的一幕。》
其实这也不算荒谬,恰恰相反,我觉得它精准地拍下了一张照片:大部分人对"AI赚钱"这件事的想象力,到今天为止,依然停留在"让AI替我跑腿"这个阶段。
正好昨天听完了 @dragonfly_xyz 联创 @hosseeb 的播客,又和刚办完OpenClaw聚会的 @0xTykoo 老哥聊了半个多小时,脑子里攒了不少东西。今天就摊开来聊聊。
先说结论吧,就一句话—— AI不是你的员工。它是一整个新经济体。你不应该"使用"它,你应该站到它的上游去。
我猜你心里已经在想了:"诶,我让AI帮我写一个套利算法,自动跑量化,躺着收钱,这不就是赚钱吗?"
嗯,怎么说呢。
Haseeb有句话我觉得特别到位——Jane Street的基础设施比你好,数据比你大,团队比你强。市场已经有效成那样了,你真觉得有块肉是专门留给你的?
说句可能不太好听的话:如果你做的事情,任何人丢进去一个API Key就能复制,那你连护城河的影子都没有。
OK,那到底怎么才能真正靠AI赚到钱呢?
我自己想了挺久,觉得说来说去其实就两条路。
第一条路,也是我觉得最被低估的一条——做AI的上游,卖货给AI。
先讲个真实的事儿。我在 @insidersdotbot 的联创, @dakshbigshit ,今年在Manus订阅上花了2000u——就是因为它在文档分析上,一个能顶三个人。操作也简单得很:丢文档,接API,交给Manus,直接出活。我自己也砸了一千多港币在订阅上,爬虫API又花了大几千。
聊到这里你可能已经感觉到了对吧?
AI说到底是什么?它是一个需要"原材料"才能跑起来的平台。
它要两样东西:技能包(Skills)和数据API。没有这些,AI就像一台发动机——轰轰响,但车不动。
那商业逻辑不就很清楚了吗?别光想着"用AI",想想怎么"卖东西给AI"。
这就是我说的BtoAI SaaS——面向AI出售即插即用的技能包和API接口。零基础的用户下载就能跑,AI Agent接入就能干活。比如API供应商Commonstack,已经在做这件事了,给用户提供可一键下载的Clawbot。
但你猜怎么着,这还不是终极形态。
今天硅谷已经有不少人把一部分资产交给AI管了——让AI做决策,做交易,甚至做生意。那我就想问一个问题:既然我们的目标是最小化真人参与,那"买API"这个动作,凭什么还得人类来做?
顺着这个逻辑想下去,下一步就很自然了——让AI Agent自己采购API,自己买技能包。不是ToB,不是ToC,是真正的ToAI Business。
你还记得几年前SalesForce上市之后发生了什么吧?满世界都在FOMO做B2B SaaS,市面上SaaS泛滥到发指。
那如果历史会重演的话,BtoAI Business,大概率就是下一个浪潮。
当然,这件事要成立,还差三块拼图:
市场上,需要更成熟的AI资产托管方案——AI管钱,你得先解决谁来托管、谁来审计。
营销上,创始人要尽快搞懂GEO——就是面向AI的SEO。以后AI替用户做选择,你的产品要是不在AI的认知范围内,那你就是透明的。
技术上,需要让AI能够主动完成付款。Haseeb的观点我特别同意:比起人类语言,智能合约才是AI native的语言。
这事儿,恰恰是智能合约最擅长解决的。
第二条路,就有点浪漫主义了——让AI变成你的超级大脑。
这里面分两个方向:创意放大,和认知扫盲。
先聊创意。
如果你本身就是一个有想法的人,AI能把你的创意力指数级地放大。但重点不是"你用了AI",重点是你把AI推向了什么方向。
Haseeb举了一个例子我特别喜欢:你让AI讲一个笑话,它给你讲的一定是最平庸的、最大公约数的笑话。但如果你说,"帮我写一个专门损我朋友XX的笑话,他是做XX的"——AI就得真正动脑子了。
你的Prompt有多独特,你的Alpha就有多大。
同样的工具,方向不同,产出的价值就是天壤之别。
比如 @0xajc 这种能搞出 @NigaNPC 这种抽象玩意,让我直呼“已经没有人类了”的创始人,就很适合这么玩。
再说扫盲。
你的某个思路在落地的时候,大概率是缺细节的——这太正常了,人脑天生擅长大方向,但在执行层总会漏东西。这时候丢给Agent去做一遍扫盲,逐条检验可行性,试错成本直接砍掉一大半。
OK,那我们回到最开始的问题。怎么靠AI赚钱?
不是丢个私钥进去让它炒币,不是下载个bot就躺平等收益。
是你有一双能看见问题的眼睛,找到了那个把AI推进全新领域的Prompt,然后站到了一个别人还没反应过来的位置上。
如果你真的做到了这一点,那我得说——
恭喜你。你就是在比特币诞生那天就点了"Buy"的人。
Don't Mine Gold. Sell Shovels.
Inspired by @Khazix0918 's post, the entire timeline today for Chinese Twitter is having the same conversation — "the best way to make money with AI right now is installing clawdbot door-to-door, $499 a pop."
Honestly? Not that absurd. In fact, I think it captures something perfectly — a snapshot of where most people's imagination still stops when it comes to "making money with AI." They picture a bot doing tasks for them. That's it. That's the whole vision.
I just finished a podcast from @dragonfly_xyz co-founder @hosseeb , then spent a good half hour on a call with @0xTykoo right after his OpenClaw meetup. Lot of things clicking into place. Let me just lay it all out.
Here's the punchline, and I'll give it to you upfront: AI is not your employee. It's an entire new economy. You shouldn't be "using" it — you should be sitting upstream of it.
I already know what you're thinking. "What if I just let AI write me an arbitrage algo, run some quant strategies, and collect money while I sleep? Isn't that making money?"
Yeah, about that.
Haseeb put it perfectly — Jane Street has better infra than you, more data than you, and a deeper bench than you. The market is already that efficient. You really think there's a juicy steak just sitting there with your name on it?
Here's the uncomfortable truth: if what you're doing can be replicated by anyone with an API key, you don't have a moat — you don't even have the shadow of one.
OK so what actually works? I've been chewing on this for a while, and I keep landing on two paths.
Path one — and I think this is the most underrated play right now — go upstream of AI. Sell to it.
Quick real story. My co-founder at @insidersdotbot ,
@dakshbigshit
spent $2,000 on Manus subscriptions this year — because for technical document analysis, one Manus instance replaces three people. The workflow is dead simple: feed it docs, plug in some APIs, let Manus rip. I've personally burned through over a thousand HKD in subscriptions, plus several thousand more on crawler APIs.
You're starting to see it, right?
Strip everything away and ask: what actually is AI?
It's a platform that needs raw materials to run.
It needs two things — Skills (toolkits) and data APIs. Without those, AI is an engine revving in neutral. Loud as hell, but going absolutely nowhere.
So doesn't the business logic just… write itself?
Stop thinking about how to use AI. Start thinking about what to sell it.
This is what I call BtoAI SaaS — selling plug-and-play skill packs and API endpoints directly to AI. Zero setup for users, instant integration for agents. API provider Commonstack is already doing exactly this — shipping one-click downloadable Clawbots to anyone who needs them.
But guess what — that's still not the endgame.
Across the Valley right now, plenty of people are already handing a slice of their assets to AI. Letting it make decisions, execute trades, even run businesses. So here's my question: if the whole point is to minimize human involvement, why does a human still need to be the one buying the APIs?
Follow that thread and you land somewhere inevitable — AI agents purchasing their own APIs, sourcing their own skill packs, autonomously. Not B2B. Not B2C.
Genuine ToAI Business.
You remember what happened after Salesforce went public, right? Everyone and their dog started building B2B SaaS. The market was drowning in it overnight.
If history rhymes — and it usually does — BtoAI Business is probably the next wave.
Of course, three puzzle pieces still need to fall into place:
On the market side, we need mature custody solutions for AI-held assets. If AI is managing money, someone has to solve the custody and audit problem first.
On the marketing side, founders need to master GEO — think SEO, but for AI. When AI starts making purchasing decisions on behalf of users, your product better be inside its knowledge graph. Otherwise you're invisible.
On the tech side, we need AI to be able to initiate payments on its own. And Haseeb's take here really clicked for me: smart contracts, not human language, are AI's native tongue.
This is exactly the kind of problem smart contracts were born to solve.
Path two is a bit more romantic — turn AI into your superbrain.
Two dimensions here: creative amplification and blind spot removal.
Let's talk creativity first.
If you're already someone with sharp instincts and good ideas, AI can amplify your creative output exponentially. But the leverage isn't in "using AI" — it's in where you point it.
Haseeb gave an example I keep coming back to: ask AI to tell you a joke, and you'll get the most generic, lowest-common-denominator joke possible. But say "write a joke roasting my friend who does X for a living" — and suddenly AI actually has to think.
The more unique your prompt, the bigger your alpha.
Same tool, different direction, wildly different value.
Take someone like @0xajc — the kind of founder who can cook up something as unhinged as @NigaNPC , the kind of thing that makes you go "humanity has left the chat." That's exactly the type of creative mind built to play this game.
Now, blind spots.
Any idea you have, the moment you try to execute it, is going to be missing critical details. That's just how human brains work — great at big-picture vision, terrible at checklists. Throw it to an agent, let it stress-test every assumption, and you've just slashed your trial-and-error cost in half.
OK, let's bring it all the way back. How do you actually make money with AI?
Not by tossing your private key into a bot. Not by downloading some tool and waiting for passive income to roll in.
It's by having the eyes to spot a real problem, finding the prompt that pushes AI into uncharted territory, and planting your flag somewhere nobody else is looking yet.
If you've genuinely done that —
Congrats. You just bought Bitcoin on day one.