Who's behind the Black Lobster? Who's hyping up the Lobster?

Writing by: 0xTodd

I think there’s a bad trend right now: either praising Lobster/OpenClaw to the heavens or outright dismissing it as a scam.

Objectively speaking, I’d give Lobster a 65 out of 100.

Lobster actually has three main strengths:

Basically, Lobster is: external AI (Claude, GPT, MiniMax, etc.) + hands-on ability + super long prompt words.

So, first, don’t expect its intelligence to be groundbreaking; it hasn’t improved its IQ. Its IQ 99% depends on the model you use.

Of course, I suspect some people hype it up mainly because they hadn’t used Claude Opus 4.6 before. Now, through intermediaries like OpenRouter, they’re using Claude, and suddenly, they think it’s smarter.

Second is hands-on ability.

Previously, these AIs were just advisors, which nobody liked. Now that they have hands-on capabilities, they’re kind of like employees. But honestly, they’re only at intern level, just good at sweet-talking.

Third, prompt words.

Lobster comes with a set of super long prompts by default. Honestly, most people’s prompt-writing skills can’t reach that level.

So, after using it, people think: “Wow, Lobster is amazing!”

And humans are naturally lazy; no one wants to craft such long prompts when chatting traditionally.

Who’s criticizing Lobster?

But on the flip side, I can understand why some people dismiss Lobster as useless.

Lobster costs money but doesn’t make money.

Because, fundamentally, it’s not just Lobster’s problem. Whether you use GPT or Claude, you can’t make money just by using these AIs directly.

Remember the AI crypto trading contest? Everyone lost money, right?

PS: The AI crypto trading contest was won by Doubao — because Doubao didn’t even open an account.

They can’t make money in their normal form, so putting them into Lobster doesn’t change that. Their IQ remains the same.

Since AI can’t really make money, some people resort to methods like: installation services and selling courses.

If you want to make money selling courses and installation services, you have to hype Lobster to the max. That’s been the game since the Mimon era.

In that case, it’s just a 65-point tool, but you hype it up to 130 points, claiming it’s a Web 4.0 revolution.

In reality, after all the fuss, people realize: “Damn, it’s not even 130 points.” Naturally, this leads to a strong sense of disappointment. The gap between buyer’s expectations and reality is huge, and it’s very easy to mess up.

My opinion:

If you ask me to rate it, I actually like Lobster’s concept.

I think it has a persona, which I’ve mentioned before, and I think it’s very good.

Right now, I’m using Minimax M2.5, a domestic product with very low cost, and for image recognition, I use Claude Haiku, so I’ve found a sweet spot.

Chatting with me, doing small tasks, not using many tokens — it’s quite pleasant.

This is also a small “crossing the river on a pony” problem. Some friends use Claude Opus 4.6, which costs 20 times more than mine. If their output is bad, it’s very frustrating.

I haven’t spent much money, and if the output is unusable, I don’t mind throwing it away.

After all, Vibe Coding is like a new era slot machine. I play for 5 cents each time, and you play for 1 dollar. Even if my hit rate is slightly lower, the price difference makes the experience totally different.

Security issues can’t be a game of Tian Ji’s horse racing.

Everyone says OpenClaw surpassed React in just 3 months, which took 13 years and 240,000 stars, even leaving behind the 40-year-old Linux kernel.

But little do they know, OpenClaw’s bug count also surpasses Linux.

Check out the OpenClaw GitHub page — that star count is already in the top 10 worldwide. The bugs listed in issues have reached an astonishing 5,000+.

What does 5,000+ mean? It’s because GitHub only displays up to 5,000 issues.

That’s not the limit of OpenClaw; it’s the maximum display limit of GitHub.

In comparison, another open-source project with 200,000 stars, Vue, has only 359 issues.

So, I think many ordinary people aren’t really suited to use open-source software. It’s better suited for those with some technical background, especially with security awareness.

If you recklessly give OpenClaw many permissions, it could be stolen, right?

(Actually, not many have been stolen, but such news spreads easily. Sometimes they say hundreds of thousands of ports are exposed, or credit cards are hacked, which can create negative impressions of Lobster.)

At such times, you can’t expect Gemini, Claude, Doubao, or Qianwen — all mature, secure, and widely used software — to match the level of an open-source geek tool.

That’s a mismatch.

There’s a saying in software development: “Anti-stupidity mechanism.” Traditional AI does a good job of preventing stupidity, but Lobster currently has a bunch of bugs. After all, it’s open-source, even written with Vibe Coding.

Even Google, recently, had a news story about a guy in Florida who chatted with Gemini for a long time and ended up committing suicide. His family even sued Google.

Think about it — even a company with such high security standards can have issues. So, what about Lobster?

Summary: Should I install Lobster?

In conclusion, if you ask me to objectively evaluate Lobster, here’s what I think:

If you have some technical skills and enjoy tinkering, treat it as a research project.

Or, if you just have free time and want to try your luck, go ahead and play with Lobster.

I think it’s fun enough as a pastime. Just enjoy it with that mindset.

But don’t expect to make big money or lead a Web 4.0 revolution — you’ll definitely be disappointed.

If you’re busy and don’t have time to tinker, just use regular Claude, Gemini, Doubao, or DeepSeek. This so-called revolution won’t leave you too far behind.