Today I saw someone forwarding a message in a group chat, saying that after installing Feishu, you can one-click install OpenClaw and also get 4 million Tokens for free, so I excitedly downloaded it to try out.
Minimalist Installation
Compared to the hassle I had previously when installing on Ubuntu or Docker, this installation process on Feishu can be said to be revolutionary.
Download the App and register, and as soon as you open the interface, a prompt for one-click installation of Little Lobster will automatically pop up. Just click once, wait two minutes, and it’s done.
The Default AI Is a Bit Silly
I hadn’t even had a chance to ask which model this built-in AI actually uses. First, let me test the network connection by asking it to search online for interesting news on HackerNews today.
As a result, there was no response for a long time. After waiting for about 5 minutes, a bunch of messages came out consecutively, saying the network timed out.
Then I changed the request, asking it to check what’s trending on Weibo Hot Search today. But it kept getting stuck in the previous task, still repeatedly trying to access HackerNews without success.
I reminded it for the third time to check Weibo Hot Search instead of continuing with the previous task, and it finally got the idea.
But obviously, after messing around for a while, it still couldn’t find the Weibo Hot Search. It gave a bunch of reasons, which all made sense, but obviously, Weibo Hot Search is not only available on Weibo itself—many other websites also pull this information.
After much deliberation, it finally found a mirror version of Weibo Hot Search.
Then I asked it to follow the same approach and search for mirror information of HackerNews again, and this time it actually found it.
Since it had already found the information, I then proposed a third request, wanting to test its ability to analyze article content, as well as test whether it could search shopping websites.
And then, there was no “and then.” The 4 million Tokens had already been used up!!!
Since I hadn’t had a chance to ask what model it is, here I let Doubao roughly calculate the price of 4 million Tokens.
A Game for the Rich
To be honest, the effective time of this trial from start to finish didn’t exceed 5 minutes. I didn’t expect that the model consumption could be this fast. Previously, when I used my self-hosted Little Lobster for one or two hours, it seemed to only consume three or four million Tokens, and that was with intensive conversations and large amounts of long text input and output. According to this consumption rate, not many people can really afford to keep a Little Lobster.
