Raising Lobsters: China's National OpenClaw Mobilisation
Why are retirees, students, and housewives lining up to wire OpenClaw into their lives?
OpenClaw, the latest darling of AI agents. It exploded in popularity in early February, and now an increasing number of off‑the‑shelf tools are shipping with OpenClaw agents built in.
You probably already knew some of those; here’s what you didn’t: the way China is using OpenClaw looks nothing like what’s happening in the West.
In China, OpenClaw’s story shifted from a technology to an economic one.
On Chinese eBay right now, hundreds of freelancers are selling one service: come to your house and install/set up OpenClaw for you.
A district in Shenzhen has drafted a ten-point policy, called 'The Lobster Ten', rewarding people for building companies on OpenClaw.
If you register a one-person company in the district, the proposal includes three months of free computing, a 30% subsidy on AI token costs, months of free accommodation (yes, you heard it right), and more.
Even though this policy is still in public consultation, it's already the most specific and concrete government support for AI agents anywhere in the world.
China has done this before.
The speed. The subsidies. The frenzy.
The exact same playbook has been used since 1958, ie, mass mobilisation, local government competition, ordinary people thrown at a technical problem.
The very playbook that contributed to an estimated 30 million deaths in the Great Leap Forward has, in other eras, produced near‑total dominance in manufacturing and robotics.
Again, the mechanics underneath are the same; only this time, the target is achieving mass adoption of the mass AI agent (OpenClaw).
The questions I’m answering today are:
Is China going to succeed in this round of AI-agent competition?
With $1, how far can you start an OpenClaw business in the UK and US versus China?
And what are the angels you should watch out for while this develops?
The Four Steps Unique Chinese-Style Paybook
China’s approach to new technology is simple: move people, capital, and local officials until it either becomes an industry or fails, full of debt.
And the machine has been running the same moves since the late 1950s: pull in people with no formal expertise; set local governments against each other on speed; treat caution as weakness; and move fast enough that mistakes become learnings instead of disasters.
And this pattern repeats every few years, in every frontier industry. Many of us have seen this in textiles, batteries, EVs, solar panels, rare metals, robotics, and AI.
Those four moves have produced outcomes ranging from catastrophic to world-dominating. Sometimes both, depending on where you stand in the supply chain.
Where it went catastrophically wrong
The Great Leap Forward is the clearest proof of what happens when it’s pointed at the wrong problem, with no honest feedback loop, and the mobilization was too massive to halt halfway.
From 1958, Mao mobilised peasants to produce steel, and local officials were handed production targets to achieve.
There were many things wrong in that era; the local officials were cheating on the KPIs, while the majority of the output from backyard furnaces was mostly unusable pig iron, cast from melted farming tools.
In the end, it led to grain rotting in the fields and to 30 million people dying of extreme hunger.
The pattern survived. What changed was the target.
It was their very first experiments in nationwide mobilization.
How it worked when it worked: Made in China
In the 1990s, “Made in China” meant cheap toys and textiles.
The same four moves ran the whole transition. Tens of millions of rural workers left farming, moved into factories rading low-productivity plots for repetitive work in sectors where they had no prior experience.
Coastal cities and provinces competed to attract land plants by offering industrial land below cost and bundling it with roads, power, and ports. Places like Shenzhen and Shanghai turned into some of the most prosperous manufacturing hubs on earth.
Yes, the early factories were dirty, chaotic, and often inefficient.
But before the year 2000, “good enough and shipping” beat “perfect and slow” a thousand times for China.
Today, China accounts for roughly 29% of global manufacturing output; the rest of the world depends on it.
There were overbuilt factories for sure, but accidentally, it created a feedback loop. The more they build, the cheaper the next unit, the harder it becomes for anyone else to compete.
Scale was the right thing to back, and with scale comes this manufacturing superpower.
Where it is now: Robotics
That dominance isn't so much about cheap labour anymore.
In 2010, China barely registered in robot density rankings.
To achieve this, they followed the exact same moves.
Provinces like Guangdong and Jiangsu threw money at robotics demonstration zones. Mid-tier factories with little to no prior experience joined the automation gold rush.
Of course, the early adoptions were wrong in many ways, eg, in 2017, Bloomberg’s reporting makes clear that China’s robot push could be a ‘bubble’ if they don’t adjust the direction and how they subsidize the businesses.
China now hosts the world's densest hardware stack and accounts for more than half of global industrial robot output.
And no other country can replicate the entwined robotic and humanroid supply depth to threaten China’s position in the next few decades.
In 2026, the nationwide mobilization approach is pointed to OpenClaw.
2026 the National Frenzy of OpenClaw
If you are still new to the OpenClaw topic, this is my latest analysis:
We are watching this four‑step playbook run once again.
The mission code is “raising lobsters”.
Step #1: Mass mobilisation of non‑experts
In early March, nearly a thousand people queued outside Tencent’s Shenzhen headquarters for a free OpenClaw installation.
The crowd included a retiree in his sixties, a nine-year-old primary school student, housewives, and corporate employees who’d taken the day off work. People whose entire experience of automation was WeChat and online shopping.
Over twenty Tencent Cloud engineers ran the operation: loading OpenClaw, connecting AI models, and configuring cloud hosting.
Within days, a secondary market had already formed.
On Xianyu (China’s version of eBay), freelancers started listing house-call installation services:
I’ll sit at your computer and wire the Little Lobster into your inbox.
With some installers reportedly earning over $36,000 in a matter of days.
Many installers themselves weren’t technical experts; they’d learned the process online and jumped in.
Most of the buyers were people driven by the same fear:
I don’t fully understand this yet, but I can’t afford to be the person who missed it.
Step #2: Local governments chip in
Once Raising Lobsters caught fire, city and district governments were eager to be the first. Trying to get a head start on this agent competition.
Shenzhen Longgang drafted “The Lobster Ten”, once qualified, you get three months of free compute, a 30% subsidy on the token costs, two months’ free accommodation, and 18 months of discounted office space. Tailored for one‑person companies building on OpenClaw.
Wuxi Xinwu published the “Lobster Twelve”: 12 measures, including subsidies of up to 5 million RMB (roughly $700,000) for major open‑source AI projects, rent‑free offices for 3 years, and all kinds of other allowances.
Other cities also launched their own “raise a lobster” packages. As long as you are a new business registered within the district and propose what you plan on building on top of the lobster, you have a great chance.
Each district is trying to become the centre of China’s Lobster farming.
Step #3: Revolutionary enthusiasm over technical caution
China’s internet and cybersecurity authorities issued formal OpenClaw risk warnings, including warnings about server compromise and sensitive-data leakage, while the financial institutions were also told to strictly control or stop deploying similar external platforms.
Local governments have read the same reports and pushed ahead anyway.
Their bet is simple: the upside of finding new business models outweighs the downside of some messy incidents.
They deliberately park the risks while maximising experimentation to create the AI era’s unique opportunities.
Step #4. Speed as ideology
Look at the timeline:
OpenClaw project launched in Nov. 2025.
Baidu integrates OpenClaw into its search app, which has 700 million monthly users in Feb 2026
March 2026, Longgang’s Lobster Ten and Wuxi’s Lobster Twelve drafts go public
The interesting part wasn’t that OpenClaw went viral. Plenty of things go viral.
The interesting part was that companies, local governments, and ordinary people looked at the AI agent and all decided it was worth financing nearly simultaneously. The speed at how fast raising lobster craze moves is, again, astonishing.
Now, how does this math work in the US and UK?
Follow One Dollar (China vs UK/US)
$1 in China Buys You…
A metro ride to a Tencent‑style booth where an engineer wires an agent into your laptop for free, or hiring someone to install the agent at your place.
And then local governments turn that same dollar into free cloud, subsidised rent, and small grants for one‑person companies, who can prove their ability and willingness to build on OpenClaw.
And then local governments turn that same dollar into free cloud, subsidised rent, and small grants… and most importantly: handed out at registration. Again, the roof above your head, the tokens to help raise your lobsters will be handed to you if you pass the registration, not after a pitch deck.
Because in China, the unit of bet is an experiment. Not a company.
That dollar lands before the idea is clear and even before anyone has proved the business model.
The bet is:
Tiny, messy experiments beat picking winners on paper.
The Same $1 In the UK and the US
In the UK, one dollar doesn’t buy you a tube ride to a free install (well… there was none to start with).
It buys you about 0.00000‑something of a grant officer’s time while they skim yet another Innovate UK application.
Founders spend months writing bids for the Innovatte UK scheme. And even when they win, the cash only arrives in arrears; ie, you spend first, then claim back eligible costs against time sheets and invoices.
The British government is holding the money so tight that they don’t want to see it actually leaving the Bank of England vault.
Things are better in the US.
In the US, at least a dollar ends up in the hands of people running experiments.
SBIR‑style grants routinely offer Phase I awards in the low‑six‑figure range and Phase II awards that can run into seven figures, explicitly designed to fund risky R&D at small firms. And the fund can be drawn down as the work starts (not just in the end).
That said, in both the UK and the US, most of the "in‑person" AI activity is still industry conferences, invite‑only workshops, and internal hack days. If you aren’t a developer or an enthusiast, you are likely nowhere near these events because you don’t want to embarrass yourself.
For a random retiree in New York, that dollar doesn’t bring you anywhere near an AI-agent.
Do you think the AI startup grants in your country are up to the expectations? Let me know in the comments.
So we’re back to the question from the intro.
Which Version Are We Watching?
China hasn’t moved far from the Great Leap Forward instinct — mass mobilisation, local competition, faith in speed — and that playbook has failed as spectacularly as it has succeeded. Beijing has already had to zero out entire sectors when experiments went too far: the peer‑to‑peer lending boom, where platforms imploded, and millions of retail investors were burned, and the bike‑sharing bubble, where piles of yellow and orange bikes ended up as scrap while most operators went under.
If OpenClaw goes the P2P / bike‑sharing route, we’ll see:
Angry users who lost data, time, or savings (some of them have already started to pay for the OpenClaw removal service).
Local governments are embarrassed, quietly shelving “Lobster” policies
The failure would also give an already slow gov like the UK a chance, pointing at China and saying, “See, this is why we were cautious.”
If it goes the solar / robotics route instead, we’ll see:
A generation of Chinese founders and operators who treat agents as basic infrastructure, the way we treat emails.
A long tail of one‑person and tiny‑team companies in China running at cost levels a UK or US firm can’t touch. At least their rent, compute, and early mistakes were effectively subsidised.
The difference in the OpenClaw case is that most of the waste is virtual and can be contained. Bad agents can be uninstalled, or clumsy workflows get abandoned; a thousand failed one‑person companies cost server time and a bit of grant money.
Because the whole point:
China isn’t trying to make existing companies more productive, but to manufacture a new class of entrepreneurs. With AI as the raw material.
Three questions worth watching as this develops:
Who actually shows up in person to participate?
Who is willing to front‑load risk in small chunks?
Are people actually running agents, or just installing them? (which is a universal question you should ask)
And if you haven’t read my deep‑dive on OpenClaw itself — what it is, how it works, why I changed my mind.
Read it next if you want the technical side behind this mobilisation story.