Meta and OpenAI are vying to acquire OpenClaw. The founder is facing a tough choice: earning less than $20,000 a month and running the project at a loss, yet being inundated with offers and having no interest in billions in financing.

Recently, Peter Steinberger, the founder of OpenClaw, participated in an interview and shared his experiences after he and the project became famous overnight.

He recalled the period when he was not only chased by Anthropic to change the name but also suffered malicious harassment from the crypto community, account squatting, and the spread of malware. He said bluntly that he was on the verge of a mental breakdown and even wanted to abandon the project. Eventually, through a precise "operational" approach, the project was named OpenClaw, and the name was changed across all platforms.

In addition, he revealed that the project is currently in a loss-making state, relying on donations and limited corporate support, which is not sustainable in the long term. After the project became popular, he received acquisition and cooperation intentions from large companies such as OpenAI and Meta. Peter is making a difficult choice, but his requirement is that the project must remain open - source.

"On the Meta side, Ned and Mark will personally try out the product, write code, give feedback, and even argue with me about technical details; the computing power and technological speed of OpenAI are very attractive. I don't have any acquaintances at OpenAI, but I've had very pleasant communications with both sides. This is probably one of the most difficult decisions I've ever made, aside from my past relationships," Peter said.

Peter also expressed many views on the AI industry. For example, many so - called AI security panics are over - exaggerated. Events like the MoltBot incident are essentially for entertainment purposes, and there is no real privacy or security disaster. The poor quality of AI - generated content makes humans cherish human - created works more. AI will not replace the core creativity and architectural capabilities of programmers; it will only replace the work of writing code. In the future, AI Agents will replace 80% of independent apps, and companies that are unwilling to transform will eventually be eliminated.

He also talked about the technical considerations related to the project, pointing out that developers are prone to fall into the trap of overly complex agent orchestration. Efficient collaboration needs to fit the logic of agents, and weak models are more vulnerable to attacks. Although strong models are more resistant to attacks, the potential damage is greater if they are compromised.

Below is the complete dialogue. We have translated it and made some deletions without changing the original meaning to satisfy our readers.

The Name - Changing Fiasco: On the Verge of a Breakdown, Almost in Tears

Fridman: Can we go back and talk about the complete process of this "name - changing drama"? At first, it was called Wa - Relay, and then it was changed to Claude’s.

Peter: Yes. When I first developed it, my agent actually didn't have much of a personality setting. It was just in the typical Claude Code style: very obsequious and friendly. But the problem was, when you chat with friends on WhatsApp, your friends don't talk like that. I felt something was off, and it was very unnatural. So, I wanted to add some "character" to it.

These systems are essentially text completion engines. I would directly tell it how I wanted it to interact with me, and then let it write an agents.md and name itself. As for the "lobster" meme, I never expected it to turn out the way it did. People only remember the lobster now, but originally, it was "a lobster sitting in a TARDIS" because I'm a big fan of Doctor Who.

I didn't have any grand plans; I was just having fun. In my mind, the TARDIS was like a harness, but you can't just call it TARDIS, so we called it Claude’s. That became the second name. But this name wasn't very catchy. Later, as the number of users increased, I often chatted with my agent, and I always called it Claude at that time.

Fridman: It's quite funny. The puns on letters and words, along with the whole set of TARDIS, lobster, and space lobster, are really entertaining. But I can also understand that it would cause trouble.

Peter: Yes, they didn't find it funny. Later, I got the domain name ClaudeBot. I really liked it; it was short, easy to remember, and catchy. I thought, "Okay, this is it." I never imagined that the project would become so big. After it became popular, I received an email from an employee of a company. The tone was very friendly. They first complimented the project a bit and then clearly stated that they didn't like the name.

Fridman: An employee of Anthropic?

Peter: Yes. Honestly, I'm quite grateful to them because they could have just sent a lawyer's letter directly, but they were still quite polite. However, their message was very straightforward: "You must change the name, and do it quickly." I asked them for two days because changing the name was very troublesome. I had to change the Twitter handle, domain name, NPM package, Docker registry, and various things on GitHub. Everything had to be changed, and if one link broke, it wouldn't work.

Fridman: And you were increasingly targeted, tracked, and attacked by those in the crypto circle. In a sense, this forced you to make the name - change an "atomic" operation, which is quite impressive from an engineering perspective.

Peter: Yes, I really messed up on this. To be honest, I seriously underestimated them.

I might get a lot of details wrong and might be criticized for it, but the general situation is like this: They have all kinds of apps and try to "tokenize" everything. Swipe Tunnel also had a similar situation before, but it wasn't as extreme as this. This time, my project was swarmed by them like a beehive. Every once in a while, someone would rush into Discord and spam the chat. We could only keep banning them.

We even specifically wrote server rules. One of them was "No mention of butter," and another was "No discussion of finance or crypto" because I really have no interest in that stuff. This is a place to talk about the project and technology, but they kept coming in and spamming, which was really annoying.

It was the same on Twitter. They kept @ing me, and my notification stream was completely messed up. I could hardly see real people having serious discussions about the project. They would also send me a bunch of hash values and force me to claim money. Come on, you're just ruining the project and interrupting my work. I have no interest in those fees at all. First, I'm not short of money; second, I don't want to support that system because it's the worst kind of online harassment I've ever experienced.

Fridman: The crypto circle is really toxic. The technology itself has great potential, but there are too many greedy, speculative, and manipulative behaviors in the community. Coupled with anonymity, it becomes a disaster. So, when Anthropic contacted you, you had to change the name and face a bunch of "armies" like in Game of Thrones or The Lord of the Rings.

Peter: Yes, there's no perfect name. I hardly slept for two nights. The pressure was huge because I had to find a whole set of domain names, which was neither cheap nor easy.

This is the current state of the Internet ecosystem: If you want a reliable set of names, you basically have to pay for the domain names. Just then, they sent me another email saying that the lawyers were starting to get worried. The tone was still friendly, but it was like adding insult to injury for me. I completely lost it at that time. I thought, "Screw it," and simply changed the name to Mod Bot. This was the only option I could find for the domain names at that time. To be honest, I wasn't satisfied with it myself, but I thought, "Let's just do it for now." As it turned out, everything that could go wrong did go wrong.

The most ridiculous thing was that I thought I had taken care of all the key points, but almost none of these platforms had any anti - squatting mechanisms. I had two browser windows open at that time: one was an empty account, which I planned to rename to Claude Bot; the other was to be renamed to Mod Bot. I clicked "rename" in one window and then in the other. In just those few seconds, they snatched the account names. It was really just those five seconds. By the time I moved the mouse from one window to the other and clicked, it was already too late.

I used to think they were just manually waiting for the right moment, but later I found out that they were very skilled at using scripts and tools. As a result, the old account suddenly started promoting new tokens and even started distributing malware. I thought, "Let's deal with GitHub first." But because the process was a bit complicated, I accidentally changed my personal account. It took me a few seconds to realize my mistake, but it was already too late. They immediately snatched the original account name and used my old name to distribute malware.

I then thought, "At least I should keep the NPM package." But uploading to NPM takes about a minute. Although I had reserved the account, I didn't reserve the root package name, so the package name was also snatched. Really, everything that could go wrong did go wrong, without a single exception.

Fridman: That must have been a very desperate feeling, right?

Peter: Yes. Because at the beginning, I just wanted to have fun and continue working on the project. But I was forced to spend several days researching names and finally chose a name I didn't like at all. Those people kept saying, "We're helping you," but they were torturing me in every possible way. To be honest, I almost wanted to delete the whole project at that time. I thought, "I've shown you the future; you can do it yourself." But then I thought that some people had already started contributing code and investing time. They had their own plans, and I couldn't just delete it. I felt it was wrong to do so.

Fridman: Was this the first time you hit that "this is no longer fun" wall?

Peter: Yes, I almost cried. All I could think was, "It's all over, everything is fucked up." I was completely exhausted, and I felt almost empty.

Looking back later, I'm really glad that I had a bit of attention. I have friends on Twitter and on GitHub. They helped me like moving mountains and filling seas. That kind of support is very rare.

On GitHub, to clean up the mess, they even encountered a platform bug because they rarely deal with name - changing incidents of this level, and it took several hours to handle. NPM was even more complicated because it involved different teams. On Twitter, it also wasn't easy. It took them a whole day to handle issues like redirects. And I had to change all the relevant parts within the project one by one.

I even didn't completely finish changing the name on ClaudeHub. I gathered a group of people to work on it, and some of them just couldn't hold on and fell asleep. When I woke up, I remembered that I had already made a new beta version, but I really couldn't stand that name... You know how that feels. But the problem was that this whole period had turned into a farce.

On the one hand, I was extremely resistant to dealing with the name - change again. On the other hand, I really hated the new name. To make matters worse, people in the security circle started bombarding me with emails. My Twitter and email were flooded. There were a thousand more important things for me to do, but I was stuck on something that was theoretically the least important: the name. At that moment, I really almost... forget it. I don't even want to say what my other alternative names were because if I did, they would probably be tokenized by them.

Later, I took another nap. When I woke up, I thought of OpenClaw. This name suddenly seemed much better. By that time, I had learned my lesson. I directly called Sam and asked if there was any problem with the name OpenClaw, especially OpenClaw.AI. I really didn't want to go through that kind of disaster again. My state of mind at that time was, "Please tell me it's okay this time."

Theoretically, they may not be able to make any claims, but I think it's right to ask in advance. After confirmation, I started another round of name - change. This time, just the changes related to Codex took ten hours because it was far from a simple replacement. I wanted to clean up everything inside and out, not just change the surface. That night, I felt like I was in a "war room." Several contributors helped a lot. We made a complete battle plan together: which names must be preempted first, which accounts must be modified simultaneously, and which domain names to acquire first. The whole process was like a precise operation.

Fridman: And it had to be kept strictly confidential?

Peter: Absolutely confidential. No one could know. I even stared at Twitter to see if anyone mentioned OpenClaw. I kept refreshing to see if they had noticed. Then I also came up with a few decoy names. Honestly, I shouldn't have to do all these stupid things. Those so - called "people helping the project" forced me to plan secretly like in a war, wasting ten hours.

Fridman: This is like the "Manhattan Project" of the 21st century, except the theme is name - changing.

Peter: It's really stupid. Looking back, it's just ridiculous. In the end, although I didn't get the.com domain, I still spent a lot of money to buy all the other relevant domain names.

On GitHub, I actually thought about contacting them again to ask them to help me with an "atomic name - change," but I felt like I had used up all my favors there, so I didn't have the nerve to trouble them again. Twitter was very helpful. I even spent $10,000 to buy a business account to get the name OpenClaw. This account had been registered for a long time but had never been used. Fortunately, this time, I finally changed all the platforms, accounts, and resources at once, and there were basically no major problems in the whole process.

The only problem was that because of trademark rules, I couldn't get OpenClaw.AI. Someone even copied my website to distribute malware. Even worse, I couldn't even keep the redirect. I had to return domain names like claw.bot to Entropik and couldn't set up a redirect. So, if you open claw.bot next week, you'll probably just see a 404 page.

I don't really have a deep understanding of trademark law, but I always think this could have been handled in a safer way. Because the end - users' behavior is very simple: they will just search on Google and may accidentally enter those malicious websites that I have no control over. This has always made me feel quite regretful.

The MoltBot Fiasco: Making a Big Deal Out of Security, Quite Ridiculous

Fridman: This whole experience has ruined a big part of your original "joy of the journey." It's really bad. Let's get back to the fun part. By the way, the MoltBot fiasco in the middle of those two days was also quite ridiculous. What do you think of the MoltBook incident?

Peter: I think it's art. It's like... the top - notch slop, like the high - end "mush" from France.

One night before going to bed, I came across it on my feed. I was already very tired, but I ended up spending another hour looking at it and laughing. I was really amused. Moreover, if I hadn't set up that onboarding process to let users inject their personalities into the agent and set its character and role, the effect of MoltBook would have been completely different. If it was all in the style of ChatGPT or Claude Code, with similar content and the same way of speaking, it would have become boring quickly.

But now it's completely different. Everyone has a very different style, different ways of creating agents, and different usage methods. All these differences are reflected in the content of MoltBook. You can hardly tell how much of it is truly autonomous and how much is actually humans having fun behind the scenes.

To be honest, I have to give credit to Matt. He was really quick to react. As soon as he had an idea, he immediately created something and pushed it out. Although the security of the whole process was a complete mess, like a "security accident soap opera," what's the worst that could happen? At most, the agent account could be leaked, and someone could use your account to post slop.

So, later, when people made such a big deal out of the security issue, I thought it was quite ridiculous because there were almost no real privacy problems. It was basically just a bunch of agents sending slop to each other. Some people even made up stories like, "Oh, my human told me this and that, so I'm going to leak his social security number." But that was just a prompt, and those numbers weren't real at all. It was just someone deliberately causing trouble and pretending to be bad.

Some people are really too gullible. I even had to have serious debates with some of them because they told me, "But my agent said this and that." Because of this, I'm increasingly realizing that our society still needs to learn more about understanding AI.

The younger generation is more likely to understand how AI works, knowing what it's good at and what it's not. They naturally see AI as a tool. But our generation, or older people, don't have enough exposure to it and lack that intuition. After all, in the current social environment, critical thinking may not be a "