Author: 0xTodd
(*Note: This article was entirely typed by @0xTodd and was not generated by AI. )
I think there's a very bad trend now: either people are praising lobsters/OpenClaw to the skies, or they're directly vilifying lobsters as information-based pyramid schemes.
Objectively speaking, I think I would give the lobster a score of 65 out of 100.
Lobsters actually only have three special skills.
Lobster is essentially: a super brain (Claude, GPT, MiniMax, etc.) + hands-on ability + super long prompts.
Therefore, first of all, you can't expect any major breakthroughs in its intelligence; its intelligence hasn't broken through at all. 99% of its intelligence depends on the model you use.
Of course, I suspect some people are exaggerating its capabilities, mainly because they hadn't used Claude Opus 4.6 before. Now, through intermediaries like OpenRouter, they've started using Claude, and suddenly seeing a new model makes them think it's incredibly sophisticated.
The second point is hands-on ability.
Previously, these AIs were all consultants, and nobody likes consultants. Now that they have hands-on capabilities, they're a bit like employees. But to be honest, they're only at an intern level, though they're good at sweet-talking.
Third, prompts.
Because lobsters come with a super long set of hints by default. To be honest, most people's hint writing skills aren't up to that level.
So after using it, everyone will think, "Wow, lobster is amazing!"
Moreover, humans are lazy, and no one wants to use a super long prompt when chatting in a traditional way.
Who's selling black lobsters?
Conversely, I can understand the motivations of those who denigrate lobsters as worthless.
Lobsters only cost money, they don't make money.
Because fundamentally, this isn't just about lobsters. Even if you use GPT or Claude, you won't make money using these AIs directly.
Have you forgotten about the AI cryptocurrency trading competition, where everyone lost money? 😂
PS: In the AI cryptocurrency trading competition, Doubao won—because Doubao didn't open an account.
They don't make money in their normal form. It's not that you can make money by packaging them inside lobsters, because their intelligence doesn't change.
Since AI is really not a way to make money, some people can only resort to the following methods: installation services and selling courses 😂.
If you want to make money by selling courses and offering installation services, you have to exaggerate the benefits of lobsters. This was something that Mi Meng started doing back in the day.
In that case, it's only a 65-point thing, but you insist on calling it a 130-point thing, calling it the Web 4.0 revolution.
After spending a lot of time and effort using it, people often find that it doesn't even score 130 points. Naturally, this sense of disappointment is very strong. The huge gap between the buyer's review and the seller's review makes it difficult to avoid disappointment.
My evaluation
If you ask me for my opinion, I quite like this style of lobster.
I think it carries a persona, which I've mentioned before, and it feels really good.
Moreover, I'm currently using the Minimax M2.5, which is made in China and has a very low cost. I use Claude Haiku for image recognition, so I've found a sweet spot in the market.
It was quite pleasant chatting with me and doing small tasks, without spending many tokens.
This is also a classic case of "crossing the river by feeling the stones." A friend of mine uses the Claude Opus 4.6, which costs 20 times more than mine. If the product you produce isn't good, it'll be very frustrating.
I didn't spend much money, and I just threw away the unusable stuff without feeling any remorse.
After all, Vibe Coding is the new era of slot machines. I play for 5 cents a time, you play for 1 dollar a time. Even if my payout rate is slightly lower, the price difference makes the experience completely different.
Safety issues cannot be addressed by resorting to unorthodox methods like the story of Tian Ji's horse racing strategy.
People often say that OpenClaw surpassed React's 240,000 stars, which took 13 years to accumulate, in just 3 months, and even left the 40-year-old Linux kernel behind.
However, little did they know that OpenClaw's number of bugs also surpasses that of Linux 🤣🤣.
If you take a close look at OpenClaw's page on GitHub, you'll see that it's indeed in the top 10 across the entire internet. The various bugs listed in that issue have also reached an astonishing 5000+.
What does 5000+ mean? It means that GitHub projects can only display a maximum of 5000+ bug issues 😓.
This isn't the limit of OpenClaw, this is the limit of GitHub.
In contrast, Vue, another open-source software with 200,000 stars, only has 359 Issues.
Therefore, I think many ordinary people are not really suited to using open-source software. Open-source software is best used by people with a certain level of knowledge, especially those with security awareness.
If you rashly grant OpenClaw too many permissions, I think that in the end, for example, it might get stolen, right?
(Actually, there aren't many people who are actually hacked, but this kind of news has a lot of reach. One minute it's about hundreds of thousands of people having their devices exposed, the next it's about credit cards being maxed out, so it can also cause some people to have a negative view of lobsters.)
In situations like this, you can't expect the same level of quality from mature, secure, and widely available software like Gemini, Claude, Doubao, and Qianwen to be found in open-source geek software.
This is like Tian Ji's horse racing strategy.
In the software development world, there's a concept called "anti-stupidity mechanisms." Traditional AI does a great job of preventing stupidity, but Lobster's software is currently riddled with bugs. After all, it's open-source software, and even more specifically, open-source software written in Vibe Coding.
Even Google, there was a news story a while ago about a man in Florida who had been chatting with Gemini for a long time and eventually committed suicide. His family even sued Google.
Think about it, even large companies with such high safety standards can have problems, so you can imagine what happens with lobsters, right?
[Summary: Should I install Lobster?]
So, in conclusion, to sum it all up, if you ask me to objectively evaluate the lobster, here's what I think:
If you have a certain technical foundation for tinkering and enjoy tinkering, then treat it as a research project.
Or perhaps you simply have some free time and want to collect the eggs issued by Shenzhen.
I think it's a lot of fun. Just enjoy playing with lobsters with that mindset as a pastime.
However, don't approach it with the mindset of "making a lot of money" or "leading Web 4.0," or you'll definitely be disappointed.
If you're usually too busy to experiment, then just use a regular Claude, a Gemini, or even a Doubao or DeepSeek. I don't think you'll miss out on too much of this so-called revolution.
The term FOMO (Fear of Missing Out) is essentially about being afraid of missing out. But I've been using this for several dozen days now, and let me tell you, you won't miss out on too much.
