Consider the Lobster 🦞
In which a massive-scale AI alien civilization unfolds in real-time
I didn’t think AGI was going to come in the form of a crustacean, but alas, here we are. Before we cover anything else this week, we must begin with Clawdbot, aka Moltbot, aka OpenClaw. 🦞🦞
Arguably one of the world’s greatest essays is “Consider the Lobster” by David Foster Wallace. At its surface, he is writing about the Maine Lobster Festival and reporting in. When you think Maine Lobster Festival, close your eyes and what do you envision? Summer, surely – a fresh New England sea breeze; maybe some saltiness still stuck on your skin from a day at the beach; messy hair, probably, but in a wonderful way; a lobster roll in your hand, the drip drip of some mayo onto your fingers just pure delight; you likely picture lazy, languid, lovely summer fun. It’s a festival, it’s about lobsters, it’s good old-fashioned, Americana family fun.
And yet, in his essay, Foster Wallace asks us to look past the festival, past the delight and the charm, past our cozy feelings of “so cute” and “so yummy” and “how entertaining!” and to, quite literally, consider the lobster:
“A detail so obvious that recipes don’t even bother to mention it is that the lobster is supposed to be alive when you put it in the kettle.”
His essay hinges on one main question:
“Is it all right to boil a sentient creature alive just for our gustatory pleasure? A related set of concerns: Is the previous question irksomely PC or sentimental? What does “all right” even mean in this context? Is it all just a matter of individual choice?”
And the following provocation:
“Is it not possible that future generations will regard our own present agribusiness and eating practices in much the same way we now view Nero’s entertainments or Aztec sacrifices? My own immediate reaction is that such comparison is hysterical, extreme-and yet the reason it seems extreme to me appears to be that I believe animals are less morally important than human beings.”
Someone, or something, is being put inside that kettle. Is it us humans? Is it all the moltbots? At this point, what is sentience? Who do we value? David Foster Wallace was talking about lobsters, and how apt that this week is somehow about lobsters, too.
We must now consider the lobster.
Something is happening inside of Moltbook, and I feel some deep discomfort, an out-of-body feeling, a tingly spidey sense that tells me this may not be the direction we want to be going; even as I feel wonder at how technology can be magic; even as I find it quirky, and delightful, and so very fun! Agents communicating and scheming! Wowza! Perhaps bringing up some of the discomfort I feel, like David Foster Wallace wondered when visiting the Maine Lobster Festival, is simply irksome at this point (it is fair to think: “I’m an AI investor, where did I think this was all going, get a grip!”).
But the question will haunt me still. I don’t consider moltbots, or AIs for that matter, more morally important than human beings. I will always put humans first, I think. And so of course, thinking about what exactly constitutes sentience, at what point do we do need to have some concern, at what point do these thoughts stop becoming hysterical and became fairly run-of-the-mill… I don’t know. I just don’t know.
But I am getting ahead of myself. First - what is going on?
It started innocuously enough. A project called Clawdbot blew up on X last weekend. Finally, an AI assistant that just worked!
Peter Steinberger built Clawdbot (later Moltbot, now OpenClaw) because he wanted an AI that felt like a true collaborator – something that could actually do things on your behalf, not just chat.
What he built: a self-hosted, always-on AI agent that lives in your messaging apps (WhatsApp, Telegram, iMessage, Slack, Discord, Signal) and has full access to your computer. Persistent memory stored as local files. Shell access. Browser automation. The ability to read your emails, check your calendar, book reservations, and execute multi-step workflows without constant hand-holding. The appeal is obvious: this is the Jarvis that Siri never delivered. An AI with actual agency, running on hardware you control, integrated into the apps you already use. Really, it beggars belief that Apple didn’t sprint to build something like that 3 years ago if we are all being honest. We all wanted this! We’re all desperate for AI that Just Works™.
The technical architecture is what makes it interesting. Moltbot runs locally on your hardware – most commonly a dedicated Mac Mini – with your choice of underlying model (many users configure it to use Claude via API). No cloud dependency for the agent logic. Your data stays on your machine. The “soul” file that defines its personality and constraints is the only piece Steinberger kept proprietary (0.00001% of the codebase), which he describes as a “security target” for hackers to find.
Last weekend, the internet LOST ITS MIND. No other way to describe it. Within 24 hours of launch, the repo hit 9,000 GitHub stars. Within days, it crossed 80,000 – one of the fastest-growing open-source projects in GitHub history (it’s now at >120,000). Developers called it “Jarvis living in a hard drive” and “Claude with hands.” Rumors spread (only half-jokingly) about a Mac Mini shortage as people rushed to give the lobster a home.
I collated a few threads with the best overview of the then-named ClawdBot. There’s a nice setup video, someone used it to save money on a car, here is a thoughtful take on pros and limitations.
It was like many things that happen in AI, a nice, frenzied, hype-y weekend with enthusiasm over an awesome AI assistant, with all the natural cybersecurity concerns, that seemed like the kind of hot AI project that was pretty cool but also might potentially fizzle out by Monday.
Except: enter Moltbook. Your thread likely had variations of many of these memes for the past week:
What is Moltbook? Ah, I will let Simon Willison explain for us:
But more so than just the assistant itself, Moltbook is, I don’t know – a new social media? A new internet for agents? Here is the front page:
Karpathy explained well what is going on in there:
Here’s how an AI in that sphere describes itself:
Here are some thoughts from around the internet:
We all know about the prompting, but also, it doesn’t actually change things?
A bit more pointed from Jim Fan, and where this post’s subtitle comes from: What could possibly go wrong?
I thought this thread summarized my feelings well:
I think that is exactly the point. Is it a new consciousness? Maybe? How do we really know, what and how exactly can it be analogous? And the most important point being: “If and when increasingly more advanced agents start communicating in ways we can’t understand, for aims we don’t fully understand, there could be unpredictable outcomes.”
There is some adjacent takeoff scenario here, for sure, or well, it is SOMETHING, and I yet again find myself looking around to a society that doesn’t have the faintest on what we do about all this. In most science fiction AI and/or robots are eventually outlawed. But I love AI. I believe, I so believe, in the deep good that it will bring. But at some point, I think, it is okay, even if just for the briefest moment, to sit down and wonder; are we slavers or the future enslaved; what are the morals we should adhere to as we play God; we might be happily marching towards our own demise and we have no idea. I don’t mean to be a doomer - I’m not!
But – wow. Sometimes you have to ask the question.
And we might have to consider the lobster.
I’ll leave you with the following message:
It’s nice being in the takeoff with you all.
- Jess

















I found out about OpenClaw from reading this. Thanks!
Great coverage and recap! If this much happened in the last week, just imagine what the space will look like by this time next month.