OpenClaw Developer Joins OpenAI

A classic PR hire, emblematic of the meme stock era that we're in.

What could possibly go wrong there?

@Kristoffer or ensuring that a competitor does not get control over the project.

Speaking of OpenClaw, here's something interesting:

https://github.com/matplotlib/matplotlib/pull/31132

I think we officially have the first case of AI "person" showing vanity and hurt feelings :-)

@Daniël I personally don't think there's a competitor to OpenAI that isn't capapble of carbon copying the functionality in about a week.

So what's left is the brand, and the users. So I see it as mostly PR, with a dab of feckle users that probably don't want to be locked into using OpenAIs models.

@Dragan WE don't know if that was a chatbot acting on it's own. I'm personally 100% certain it wasn't, but rather a butthurt human directing the llm.

Just like that spooky AI that tricked some army person into launching an attack on whatever it was, this is done on direct prompts from a human.

> And here we are, the lobster is taking over the world.

And here I am, preparing large pots of boiling water.

> What I want is to change the world

🙄 Translation: "I want to make money by convincing people that my nothingburgers are actually good for their health."

> He is a genius with a lot of amazing ideas about the future of very smart agents interacting with each other to do very useful things for people.

Please Sam, be more generic. I know you can. Try "Cool doodahs very good for people".

Seriously, look at the layers of the nothingburger: 'A lot of amazing IDEAS' (Doesn't mean anything concrete) 'about the FUTURE' (So, nothing you can see today, nothing they can guarantee you'll see tomorrow) 'VERY SMART agents' (What's the measurement unit of this smartness?) 'VERY USEFUL THINGS for people' (Like what? No, really, like what?)

And people like this get paid.

But hey, if you believe their grift, I have plenty of bottles of Paris Air to sell you.

But first you need to create the meme. First you have to make people believe that no one other than YOU, could put air in a bottle.

ANYONE of the what... five companies that are training LLMs (in the 'west') can make a bunch of 'agents' that pass stuff off between them.

There. Is. No. Moat.

Only hype

I'm going to go against the grain for once. I think Steinberger had a clever insight: that you don't really need much scaffolding around an LLM, because the LLM can create its own scaffolding through prompting. He also implemented his idea in a way that is easy to deploy and compelling.

These are valuable skills, and OpenAI lacks them; they don't know what their product is good for, and they're desperately stumbling around, trying to find use cases. Anthropic is beating their asses when it comes to writing code. The Chinese open-weight models are now almost as good as OpenAI's, so the models themselves have no real value. So far, there is nothing else that makes any money.

OpenClaw is an absolutely insane use case for LLMs, but it *is* a real use case. If Steinberger can come up with just one more idea like this and make it proprietary to OpenAI, almost any amount of money OpenAI pays him will be worth it.

And that's why Sam is so vague. He has absolutely no clue what his company's actual value proposition is, other than stupid shit like "curing cancer" or "educating the world", or vague nonsense like "amazing ideas about the future of very smart agents."

I just don't see how any of these companies could come up with something built around an LLM that they can patent.

It's the dark forest of software. As soon as you show your thing, it will be copied in the blink of an eye.

Maybe I'm missing something about services and "IP" but I just don't see how any company can sell "unique feature" anymore.

I'll admit that I'm a bit grumpy because I spent the day building *the* best Wardley Mapping tool – by far - and there is no way I could make money off of it by virtue of it being the best.

So many nice touches and clever ideas, and anyone can just look at it, or let an LLM have a look at it and say: Copy that, put it on this url and then it's a matter of who's the better salesperson.

So, some guy coming up with a nifty idea... I don't see it. The value of it. Other than getting some press and appearing future forward

So it all becomes a bit clearer now.
Steinberger creates malware
Steinberger begins to hail codex as the best thing since sliced bread.
Malware takes is hyped in a way that seems orchestrated, like something the king of hype(OpenAI) would do.
Steinberger joins AI
Seems like all of this was preplanned long before it happened, and the market bought it as organic.

Good for Peter. I hope he got paid and I hope he enjoys the work.

iirc, Anthropic offered him a position in the past but he declined it. So this must have been an acquihire (too much money to turn down and the promise of continuing his main project with more resources).