Recently, the whole internet has been asking: Have you been raising lobsters? This is not a crayfish that can be eaten, but an open-source intelligent agent OpenClaw that has become popular in the AI community, named after its icon of a red lobster. It can directly operate the computer and work automatically, making it a 24-hour online digital employee. The popularity of "raising lobsters" lies in the fact that it marks a crucial step for AI to move from a conversational mode of "being able to speak and speak" to an executive mode of "being able to do and do". This has made concepts such as' one person companies' within reach, greatly stimulating the public's imagination and enthusiasm for participation. But if you are an ordinary individual user, don't rush to try it out yet. In recent days, companies such as Tencent and Baidu have launched a free installation campaign for "Lobster", and users participating in the campaign have formed long queues. There are also technicians who provide on-site installation services for hundreds of yuan each time, reportedly earning a considerable income. However, many technology novices are not aware that "lobster" is not a disposable consumption that can be used immediately after installation. OpenClaw itself is open source and free, but its operation heavily relies on API calls from the underlying large model, which are charged based on token consumption. Token is the smallest unit used by AI to read and understand text, which can be understood as' chunks'. Behind a diligent and useful "lobster" is a massive amount of token training consumption. In actual cases, someone's high-end version of "Lobster" spent nearly 30000 yuan in one month under high-frequency calling. Due to the 24-hour operation of "Lobster", some users woke up to find that their hundreds of yuan recharge had been depleted overnight. What is more alarming than economic costs is the safety risks hanging over the heads of shrimp farmers. As an executable intelligent agent, the core capability of "Lobster" is to perform operations, requiring users to give higher permissions. If deployed improperly, it may allow personal devices to run naked on the public network and become "chickens" for hackers. Encountering cyber attacks, sensitive data breaches, loss of account assets, and deletion of important files are all lessons that "shrimp farmers" have already experienced in reality. The related topic "The first batch of shrimp farmers have started uninstalling" has become a hot topic. In response, China's official cybersecurity agencies have issued multiple warnings and reminders. Of course, the rise of OpenClaw is undoubtedly an important milestone in the development of AI technology. The automation future it presents is fascinating, and there are indeed users who are suitable for "raising lobsters". One is professional technical personnel. They have development capabilities and can independently solve the deployment and troubleshooting of intelligent agents. Secondly, there are clear demanders. They have high-frequency repetitive workflows that can be completed by 'lobster', and 'lobster' employees can create practical benefits. The third is those who can withstand risks. They can withstand potential risks such as data breaches and property damage, and can take security measures such as physical isolation and backup recovery. On the contrary, for ordinary users who have no programming foundation, no clear application scenarios, and no risk tolerance, blindly following the trend of "raising lobsters" does more harm than good. Instead of spending time and money raising a potentially uncontrollable 'lobster' now, it's better to stay focused and wait for the moment when future technology truly matures. After all, there is a threshold for trying new things, and good food is not afraid of being late. The charm of technology lies in empowerment, not in creating anxiety, and "raising lobsters" is not yet a universal game for all. OpenClaw is still in its early stages, and its best application scenario is still enterprise level automation processes or serving practitioners with clear needs in self media, cross-border e-commerce, and other fields. The rapid evolution of "raising lobsters" from a technological concept to a nationwide trend is driven by people's urgent expectation for intelligent improvement of productivity. We can maintain our attention and expectations, pay attention to the growth of "lobster", and hope that it will continue to mature and evolve in practical applications in different industries, ultimately growing into a universal pet with controllable risks and easy adoption. (New Society)