“Raising Lobsters” has become so popular that I’m sure everyone knows it even better than I do.
Whether it is various independent self-media accounts or authoritative official channels, everyone seems to be paying close attention to raising lobsters—on one hand reminding people to use it cautiously, and on the other hand providing strong policy and financial support.
But as for me, up to this point, I still have no plan to personally try installing Lobster. The main reasons are as follows:
1. We already had our own “Lobster”
This is the main reason. Before Lobster appeared—or at least before I knew about it—I had already been using tools like Cursor and Qoder, which implement coding functionality through natural language.
These tools are already very useful, so I simply do not feel motivated to try Lobster. I am also quite busy in practice, and I do not think the time investment would be worthwhile.
2. Lobster has overly broad permissions
Lobster has all kinds of skills, and the permissions it requires are also extensive. That feels far too unsafe.
If you are not careful, it could easily lead to your computer being infected. For example, there is the ransomware shown below.
3. Lobster consumes a huge number of tokens
Cursor costs 138 per month, which I already feel is expensive, but it only operates on file I/O and does not require MCP to invoke so many different software tools.
Compared with Lobster, the workflow is much more cumbersome, and the token consumption burns through real money incredibly fast. That is why there have also been online reports of people being charged tens of thousands overnight.
4. The seriousness of programming
Any code generated by AI must be carefully reviewed line by line. Only by truly doing this can you have the confidence to commit it.
For every AI-generated modification, I also use Goland’s built-in “Show Diff” feature to compare changes carefully and make sure everything is correct. If it is a frontend page, I also make sure to conduct thorough self-testing.
Even if the AI makes mistakes, the code can simply be restored. Whether through local history or git commit history, whether something was changed incorrectly or deleted by mistake, restoration is always easy.
But Lobster, with its chain of pipeline-like operations, makes the process far too uncontrollable, and many operations are irreversible. The risk is simply too high.
Have you ever raised Lobsters? How did it feel?