foo for thought

So I’ve been experimenting with Openclaw, the latest fad in the tech industry. People are literally buying Mac Minis just for this, assembling entire fleets of these little robots to run their business, launch their career, or run a side hustle. I’ve seen everything – from the hype to the FUD (come on we’re basically in crypto world at this point) – and it’s led me to question everything. What are we even here for if our little lobster can do everything for us?

For those of you who don’t know, Openclaw is a program made by literally one dude, Peter Steinberger (who is retired and did this for FUN) that puts an AI on your computer and gives it access to everything through command line tools. A security nightmare, obviously, which is why people are rushing to buy Mac Minis to run their little claw on. Business bros are all over this – you’ve probably seen an ad or two online about how to kick start your business with this (might just be me…), but it’s being hyped to the level that crypto was back in 2021. Which begs the question – is the hype deserved?

Openclaw does not replace people

First, the brutal reality – no, the hype is not quite deserved. The idea that you can maintain software by using software is, very cool but certainly not possible… yet. It may be someday that robots will be able to make other robots to make other robots to etc, but that day has yet to come. Engineering is… hard. Making something, maintaining it, advertising it – all very hard. 99% of startups fail – human startups – and you expect an AI to do a better job? That’s like not being able to run a mile fast enough and saying “You know what would make me run faster? Boots. Heavy boots.”

AIs are just worse at thinking than people are. We have billions of neurons, they do not. We can love and experience emotion, they cannot. The hype around openclaw is solely based on the idea that “computers can replace us” – it inches us one step closer to that paradigm by giving the AI access to everything we have as people.

That of course, does not mean that the AI can actually do as good as people. And, not even in the technical sense – it’s actually brilliant in the technical sense. I don’t even code for my job at Microsoft anymore. My entire job is just asking AI to code for me. That’s how good AI is now. That’s the reality. No – AI can’t actually perform as well because it doesn’t have direction. It doesn’t have purpose. It doesn’t have a family to feed. Its existence is meaningless, and there is nothing to even suggest to it that it has meaning. It can’t meaningfully store memories, it can’t experience intense emotions.

Have you ever worked with someone who doesn’t give a shit about what they’re doing? It’s like pulling teeth to get them to give a morsel of care about a task. Ok now take that person, remove their motivations in life, remove their consciousness, remove everything that makes them human – now we’re talking AI. The weird part is, unlike someone who doesn’t give a shit about what they’re doing, AI actually is extremely well-versed in knowledge – many domains, with extreme precision. It knows every language, understands every writing ever produced, every single math equation known to man.

But again, that knowledge does not equal motivation. What’s it matter, you may ask? Well, has AI made you a millionaire yet? No, because it can’t alone, because it has no motivation to. Even if you give it prompts, it doesn’t wake up every morning and have thoughts of its own – everything is programmed. Which means, all the motivation, all the idea making – that’s still on you. There’s a good reason the term “AI slop” came into existence – because something that an AI has complete control over will be a amalgamation of other things people have done.

You can’t replace your CEO with an Openclaw because they don’t care. They don’t even breathe. They don’t care if they don’t make money, why would they? There’s no intrinsic purpose to the life of an AI.

The idea of Openclaw is still transformative

With all of that said about how it can’t replace people, it has genuinely blown my mind in the past week. Giving AI the power to do anything is, actually incredible. The amount of code its cranked out on my personal Github is mind blowing. I literally have a cron job to tell it to make a “moonshot” idea and just post it to my Github at 4AM every morning (so if you see my 4am commits, those aren’t me!) I have cron jobs to improve itself, cron jobs to research market gaps – stuff that people would have to do.

Openclaw also doesn’t have to sleep. It’s not limited by biology because it’s not an organism, yet it can perform duties that only a human could. It can research, code, write articles for blogs, use social media, all while you’re tucked away in your covers. It’s genuinely wild, because the models are so good nowadays that the stuff it produces actually works and indisputably has some value to the world. It produces quality code, at least. Working without actually working is such a bizarre concept that I can only imagine is shared by corporate executives, but as an engineer it’s mind-blowingly awesome. It’s like I have 5 arms and 3 of them don’t need to sleep.

Another genuinely impressive feat it gets done are the integrations – by using all the CLIs and APIs developers have already setup, it’s able to flawlessly deploy applications from scratch. Which to me, that’s just brilliant because I absolutely hate doing all the logistics stuff – has been a huge barrier for shipping things personally. Now, I can solely focus on the ideas and the direction I want to go in instead of hacking all the weeds just to get there.

Openclaw also challenges to me the ideas of “roles” in a company. I did start to organize my Openclaw into multiple different “personas”, but then I realized they would all have identical expertise. Isolating each of them just didn’t make much sense when I thought about it more. Think about it – if you could theoretically have one person who would actually do stuff simultaneously and has domain expertise in everything, why would you even need different people to perform different roles? We only invented that because of the limitations our brain has for processing data. Computers can actually do that much better than us – they can spawn multiple processes with identical compute power, where we as people unfortunately cannot do that. So yeah currently, I only have one claw for everything – the only blocker I’m experiencing is compute power on multiple tasks might throttle it, but other than that, I find it beneficial to keep everything in one place.

It is definitely a big step towards AGI, or whatever that will end up being someday. It’s genuinely fascinating to talk to an AI that remembers experiences, who has performed tasks for you. It’s like a friend who knows everything but also knows nothing. It’s fascinating.

Concluding thoughts

The concept of an asynchronous worker is more fascinating to me than the actual release of Openclaw? Can Openclaw actually fulfill that dream? Who knows, I’m still trying to figure out how to optimize it. But one thing is for sure, it needs someone behind the wheel to steer it, just like AI. What it does for you though – really transformative stuff.

Yes, of course Openclaw is hard to setup and maintain. Yes, you do have to know about how to use a terminal and some idea of how to debug when software crashes. I’ve been doing this my entire career so it comes naturally, but all the business bros who have just hopped on the Openclaw bandwagon haven’t! And if you don’t know what a CLI is, maybe don’t try to take on something like this! I don’t want to gatekeep anyone, if you are genuinely interested in this stuff I’d highly recommend you learn stuff technically first, you can even use AI to help you learn! But, just be ready to debug. Because, yes it does kill itself – a lot. But if you’re ready to learn and be part of a technology that might evolve into the next generation of computing, maybe you should give it a try.