Written by: Haotian

As a "shrimp farmer," after enduring several weeks of painful torment, I would like to share a few insights about shrimp farming for your reference:

1) Everyone knows that shrimp farming is about "increasing efficiency," but the harsh reality is that most ordinary people are currently "wasting time" on shrimp farming. There are many unexpected troubles, such as Claude's account being suspended, API forwarding quotas being blocked, Openclaw suddenly upgrading with "memory" loss, etc. This consumes a significant amount of time, and the slight increase in efficiency is far from proportional;

2) On the Twitter timeline, I see various posts selling AI anxiety. Just take a look; there are claims that a single command can make Claude control your entire computer, that with one prompt, AI will permanently take over your work and you can just lay back, blah blah blah. The reality is, not only can you not just lay back, but you often find yourself coding late into the night for one feature, constantly fixing this bug and encountering that bug. It's hard to imagine how those who are not even willing to perform practical operations for five minutes can confidently shout about AGI arriving and AI overturning everything;

3) Raising lobsters can indeed fulfill many people's dreams of OPC (One Person Company), but the upper limit of large model capabilities is equal for everyone, but the level of understanding in harnessing large models varies greatly among individuals. Don't think everyone can become Peter Steinberger, Matt Schlicht, or Andrej Karpathy; the approach, framework design, iterative experience, capability level, and delivery results between a skilled developer and an ordinary person can be vastly different;

4) Raising lobsters essentially involves building your own exclusive AI OS, theoretically connecting multiple types of large models in parallel, managing the number of digital employees, organizing the task hierarchy of proactive and reactive collaboration, nesting skills, and controlling the friction level of collaborative task delivery, etc. The more overlapping demands there are, such as Cron, jobs, real-time data scanning, Trading skills, the greater the probability of conflict and optimization challenges. It's important to know that shrimp farming has never been an issue of "model capabilities," but rather a problem of "engineering implementation optimization."

5) Configuring Opus 4.6 and Gemini 2.5 Flash lite for all digital employees might allow them to do the same things, but the former is like hiring seasoned Wall Street elites, while the latter feels more like hiring a slave from a slum. They may achieve the same function, but the cost and delivery results are fundamentally incomparable. The truth is, the time you spend troubleshooting bugs can be matched by someone starting out using "cash capability." Shrimp farming is indeed very costly, and the scary thing is that most people understand all this yet still have to compromise by continuously optimizing with inferior models;

6) Managing digital employees is like stacking Lego; the more employees, the more skills, and the more complex the job scenarios, the higher the likelihood of a collapse occurring in an instant. For certain tasks, having sufficient capability is enough; do not be dissatisfied and impose demands that exceed your cognitive limits. It is advisable to put more effort into memory engineering, Git versioning design, and model hallucination removal; otherwise, there will be many moments that could bring you to the brink of collapse. You might feel great satisfaction one moment finishing a cool requirement, and the next moment everything falls apart leaving you in tears. Don't ask me how I know this;

That’s all.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。