OpenClaw's Crash: How the AI Agent Craze Exposed a Toy Disguised as a Tool
At the start of 2026, if you were active in the internet circle, you definitely couldn't avoid a "tech lobster" named OpenClaw.
Within two months, it amassed 270,000 stars on GitHub, directly trampling Linux's historical record; big companies scrambled to grab it, and it flooded social media feeds, making it seem like if you didn't "raise a shrimp," you'd be embarrassed to greet your peers.
But internet memory is that magical.
Just two months later, the situation reversed dramatically. The first wave of people who followed the trend to "raise shrimp" had already started paying others to uninstall it; the Ministry of Industry and Information Technology directly issued a security warning; various tech media quickly turned against it, calling out "insufficient capability, too high cost."
To many seasoned insiders, this tech carnival that came and went quickly was completely expected. To put it bluntly: OpenClaw is essentially a toy-level AI concept machine, light-years away from a true productivity tool.
Why do we say that? Today we won't discuss obscure code, only business and common sense.
OpenClaw's explosive popularity scene
Cold eyes from insiders: You thought you bought a "fully automatic microwave," but it's actually a pile of "loose parts"
For people who don't understand technology, saying a few words to AI in plain language and having it automatically write code or send emails seems like magic. But on hardcore tech communities like Reddit, this lobster has been ridiculed.
A senior developer hit the nail on the head: "For someone who has never used the command line or any workflow tool, this may look magical; but for someone who knows what they're doing, it's almost useless."
This is simply a difference between buyer's and seller's photos.
The marketing claims it's a "24/7 digital assistant." But when you actually want it to do something (like find a recipe and add it to a shopping cart), you find you first have to handle a complex environment installation, apply for API keys, configure various browser permissions, start a local instance... After this whole process, non-technical people would have long broken down.
To use an analogy, you wanted to buy a "fully automatic microwave" that heats food with one press, but after opening the package, you found heating tubes, a circuit board, and a thick operating manual for high-voltage electricity. In comparison, similar lightweight Agent products can solve the problem with two taps on a phone.
Comparison of expectation and reality
Even more critical is the cost spiral.
Some peers' real-world tests show that just to make this lobster actually run, you'd burn through over 10,000 RMB in token fees per month. Overseas users easily start at $10,000 a month.
If you're a regular person who doesn't understand engineering optimization, the commands you issue are likely ineffective, leading to 80% of token fees evaporating into thin air. It's like hiring a top-university intern at a high salary, only to have them stare blankly at their desk 80% of the time, and spend the remaining 20% repeatedly asking you "Boss, what exactly do you want?" This economic model simply doesn't balance in business.
No wonder industry insiders pointed out sharply: This is a psychological warfare by API providers, using an "AI Agent" disguise to make non-technical people pay for expensive compute consumption.
Cost out-of-control scenario
Disillusionment of outsiders: You want it to sweep the floor, it burns down your house
Insiders are dismissive, but outsiders are truly despairing. This despair comes from the huge gap of the "last mile."
At a recent global developer conference, a particularly ironic scene unfolded: Big tech engineers were excitedly gesticulating at booths, while ordinary people around them held their laptops, looking confused and asking, "I've installed it, but what can I do with it?"
It was packaged as "usable by everyone," but in reality, it remains "only for geeks."
Contrast at the developer conference
Even more frightening are the security risks. Even high-end players who know their stuff can trip up. Forbes revealed a laughable inside story: A Meta AI security director wanted to use OpenClaw to clean up her inbox, and even added "confirm before acting." But this seemingly smart AI directly deleted her important emails without a word.
You want it to help sweep the floor, but it spontaneously uses your solid wood furniture as firewood, thinking it cleaned up nicely. With this kind of "blind box assistant" that frequently fails and has extremely vague security permissions, would you dare let it handle your company's core business?
AI assistant fail scene
The "automated future" in the marketing is plump, but the reality of errors and downtime is harsh. This shows that the distance from AI technology as a "showcase concept machine" to a "construction-site excavator" spans an entire Pacific Ocean. OpenClaw merely uses a down-to-earth "lobster" shell to try to cover up the fact that it's not ready.
The gap of the last mile
The end of tech self-indulgence is the beginning of AI landing
Looking back at this "lobster mania," it's actually a mirror that reveals truth.
It reflects the restlessness of the entire tech circle: People are too easily stirred by new concepts and brainwashed by marketing rhetoric. From the metaverse and Web3 back then, to today's various AI myths, the plot is strikingly similar.
The value of technology is never about the heat of trending topics, but about solving messy, muddy problems in the real world.
OpenClaw's rapid decline is a great thing. It reminds all entrepreneurs and investors with a mess: Stop indulging in the illusion of "tech self-indulgence, users pay."
The future of AI should not be a "big boss" that requires careful catering and burns money like water; it should be subtle business infrastructure that moisturizes everything.
Only when the tide recedes, when we stop blindly chasing "disruption" and instead strive to lower the barrier, cut costs, and improve security, will the true AI era open its doors to us.
The truth after the tide recedes
Top 2 from juejin.cn, machine-translated. The original thread is authoritative.
Why, after technological progress, do later generations always want to belittle the technological pioneers to nothing? In a while, will someone belittle the current harness engineering to nothing? If you look at the brick phone from the perspective of the smartphone. Thinking you're sober is purely arrogant xx.
However, OpenClaw can also be said to be the prototype of various current Agents, such as the current Qianwen Office, WorkBuddy and other applications, especially Qianwen Office, whose various settings are almost exactly the same.