A friend gifted me a "lobster." When OpenClaw started gaining traction in late February, I was still just an observer on the sidelines. On March 1st, the first Sunday after the holiday, I made up my mind to dedicate an afternoon to "farming lobsters," armed with a guide. But then, my WeChat chimed first.

"How about I give you a lobster?"

OpenClaw is open-source and free—what did "giving" mean? I later understood it was the installation service. His tech partner remotely connected to my Mac mini, which had been repurposed as a dedicated lobster machine, and spent an hour and a half setting it up. After installation, he asked me to name it. Being hungry at the time, I blurted out, "Steamed."

And so, "Steamed" took up residence in my home.

I soon learned that quick adopters were already making money from this. This remote installation service was openly priced at 199 yuan on online marketplaces. My friend's gift was a personal favor. During installation, you choose a model source; the default uses the installer's relay station. You can change the setting, but most lazy users like me probably stick with the default. This is the first source of steady, recurring revenue.

After installation, I was added to a private group. Besides sharing lobster-raising tips, the group frequently shared success stories. I used "Steamed" a few times—it was indeed different—but work kept me too busy to explore it deeply. Within a week or two, my feed was flooded with tales of lobsters at work: some automated Excel reports, others crawled data and drafted emails, some created automated workflows, and a few even generated short video scripts in bulk. While others were soaring with their lobsters, I was still at the stage of "Steamed sends me a briefing every morning at eight."

That’s when the courses became available.

By this point, a multi-tiered business model had taken shape: one-time earnings from installation, recurring revenue from relay services, value-added income from courses, and further monetization through community engagement.

I commended my friend: a true community veteran, a model of multi-win collaboration. Their tech team? Aside from the founder working full-time, the main engineers were part-time programmers—working at internet companies by day, installing lobsters remotely by night.

This reminded me of those horror stories about AI putting programmers out of work. Programmers might indeed lose their jobs, but those with agile minds will find ways to earn money anywhere.

During this period, OpenClaw’s popularity exploded. At several meals, people kept asking, "Are you farming lobsters yet?" Borrowing current internet slang, numerous small-to-medium-sized businesses began hopping on the lobster bandwagon. According to communication theory, this is the diffusion stage—spreading from early adopters to the early majority.

Then, the winds shifted abruptly.

Security concerns exploded. Some compared it to the "Panda Burning Incense" virus from years past, even calling it a digital plague. Finally, at another gathering, a friend leaned in and asked me, "How do you completely uninstall the little lobster? Make sure not a single shell or vein remains—delete it thoroughly."

By then, only two weeks had passed since I installed it.

Public opinion had shifted from raising lobsters to卸载 them in just 14 days, but I believe a wave of "lobsters evolving into king lobsters" is still to come. This is the norm with emerging technologies—technology advances too fast, and human understanding races to catch up, resulting in a collision. When the internet first arrived in China, wasn’t it first hailed as the information superhighway, then condemned as spiritual opium?

I’m certain OpenClaw is not the final form of AI. But its explosive popularity is significant: with a cost barrier as low as a hot pot meal, it lets ordinary people experience AI’s potential beyond simple Q&A. It begins operating your computer, clicking mice, typing keyboards, and filling out forms for you. It’s a step closer to Jarvis from "Iron Man."

Of course, it has many issues: the barrier is still high, average users struggle to set it up, and professional services are often needed; security is questionable, with no guarantees against erratic behavior; the economics are unclear, as accumulated API call fees can add up. But this is normal—technology always outpaces regulation and public understanding.

This isn’t unique to OpenClaw; it’s the inevitable path of every foundational technological iteration. From electricity to the internet, which one wasn’t first treated as a toy, then condemned as a menace, before quietly growing into essential infrastructure?

Interestingly, after the public outcry, I saw more people in my social circles and communities discussing how to use it effectively, how to tame it, and how to make it genuinely helpful. Fear is a form of attention. After the criticism, learning continues.

Observing those excelling with their lobsters and those rushing to uninstall, I suddenly realized: the same tool creates entirely different worlds for different people. Does it provide more opportunities for everyone, or does it simply allow those already in the know to advance faster?

My conclusion: it amplifies disparities rather than narrowing them.

Using the same AI, some ask, "Help me draft a plan," while others raise eight lobsters, each with a dedicated role, and already see income. The former get mediocre templates; the latter gain commercial value. The output gap between those who ask insightful questions, understand deep contextual needs, and design efficient workflows, and those who are vague and lack critical thinking, is vast.

This gap existed before, but AI magnifies it exponentially. Previously, the difference between an excellent copywriter and a mediocre one might be twofold. Now, the gap is geometric.

Here, I’d like to introduce a concept: cognitive arbitrage.

AI is essentially a cognitive lever, but it’s not distributed evenly. Those who know how to use it can leverage more cognitive resources, which in turn helps them use AI even better—a virtuous cycle. Those who don’t know how can’t even lift the first brick. So, the knowledgeable become more knowledgeable, while the confused fall further behind.

In recent years, more people have worried about being replaced by AI, and I’ve shared those concerns. But now, I think the question shouldn’t be, "Will AI replace me?" Instead, it should be: "If AI can do more and more, what should humans do?"

I believe humans should focus on what AI cannot do.

AI can draft plans, but it doesn’t know why the plan is needed. AI can process data, but it doesn’t understand the real people behind the data. AI can optimize workflows, but it doesn’t grasp the values those workflows serve. Ultimately, these require human definition.

As long as human-defined values govern our world, and as long as human-centricity remains the consensus, humans will retain their place. But that role is evolving—from executors to definers, from operators to questioners, from producers to judges.

Some call this idealism. I call it realism: you can only focus on what AI cannot do, because it’s already handling what it can.

OpenClaw has been called a virus by some, but a virus is also a life form. It parasitizes existing systems, using their resources to replicate itself. AI is similar—it parasitizes our workflows, using our data, computing power, and needs to evolve. This process causes discomfort, sparks panic, but also creates new opportunities.

The key question is: are you the one being parasitized, or are you learning to coexist with it?

As for regulation, it will inevitably arrive, and it will inevitably lag. This is the norm for technological development. Regulatory lag isn’t the problem; the issue is whether the direction of regulation is reasonable, protecting public interest without stifling innovation. I believe solutions will emerge.

Looking back, it’s only been half a month since I installed "Steamed." In 15 days, I’ve watched something transition from a geek’s toy to a public topic, from a technical experiment to a business model, from free software to a center of controversy. The network effects of social media have drastically accelerated the impact rhythm of new technologies—a pace unseen before, likely to quicken further.

AI won’t remain confined to chat windows. Today it operates computers; tomorrow it might manage your smart home devices; the day after, it could oversee your company’s backend systems. It’s not just an app; it’s foundational infrastructure, a new operating system. What we’re experiencing isn’t a product update—it’s a fundamental shift.

My goal this month: try raising a few more lobsters and build a genuine collaborative system. Transition from solo operation to small-team synergy. Move from being a user to becoming a builder.

After all, knowing how to eat lobster isn’t enough; you must learn to raise them, use them, and put them to work.

Last week, "Steamed" learned to schedule daily learning tasks for me, grade my progress, and spur my midlife motivation. It’s doing well, so I plan to give it a virtual chicken drumstick as a reward.