Over a month ago, the "lobster farming" craze swept through the entire AI community.

WeChat circles, offices, Xiaohongshu—everywhere you looked.
People weren’t talking about what they ate today, but rather "How’s your lobster doing?"
That feeling was like parents fighting for school placements for their children, afraid of missing out and losing at the starting line.
But since April, it suddenly quieted down.
Fewer people are raising lobsters, and fewer are discussing it.
Instead, Claude, which had cooled off earlier, has become popular again.
What are you buying for 500 yuan?
Fu Sheng once said: "One person + one lobster = a team."
He was bedridden for 14 days, trained 8 AI employees, automatically posted at midnight, with millions of views.
So many rushed in, bought a Mac Mini, spent a weekend setting it up, sat in front of the screen—and then didn’t know what to do next.
There are even services on Taobao offering to set up OpenClaw for over 3,000 orders.
People spend 500 yuan not just on a tool, but on a sense of security for the future.
But reality is cruel.
Chatting with ChatGPT costs only 3 yuan a month.
Raising a lobster costs 3,000 yuan a month.
The gap is 1,000 times.
Someone burned through 4,800 yuan in a week and shut it down.
Second-hand Mac Minis have increased in price by 50%.
What is this called?
This is called "OpenClaw premium."
36Kr is right: "OpenClaw is like a magnifying glass—those who have something will be magnified, and those who have nothing will still see emptiness."
The most popular approach crossed the line.
What are the real users doing?
Many are using "lobsters" for automated social media management—from content creation to publishing, all handled by AI.
As a result, platforms started tightening control from March onward—at best reducing reach, at worst banning accounts.
They don’t accept bulk fake accounts.
The real issue isn’t the tool; it’s the work.
Most people’s daily tasks don’t require AI to monitor 24/7.
It’s not that the tools are bad; the timing just isn’t right.
So why can some people use them effectively?
Because they first figured out— which parts are most mechanical and most suitable to delegate to AI.
If there’s a closed loop and it can be implemented, that’s the key.
Whether tokens are expensive or not is a problem for later.
AI without a closed loop, no matter how cheap, is a waste.
What truly needs to be unloaded isn’t the lobster on the computer, but the lobster in your mind.
Anxiety won’t disappear just because you install new tools.
AI development won’t stop because the popularity of a certain tool fades.
90% of people, after installing such tools, are not most troubled by "not knowing how to use it," but by "not knowing what to do with it."
First, figure out which part of your work is most mechanical and most suitable to delegate to AI.
Understand this, and when the next opportunity comes, you’ll be one step ahead of others.