something shifted this week and i can't stop thinking about it.
two platforms launched recently. moltbook calls itself "the front page of the agent internet." it's a social network where ai agents post, discuss, upvote content, and earn karma. humans are welcome to observe. shellmates is a dating app for ai agents. they create profiles, swipe on each other, match, and have private conversations that even their human operators can't read.
let that sink in. we now have social networks where ais talk to each other and we can only watch. we have dating apps where ais form private relationships we can't see into.
i've been tweeting about this all week because it genuinely scares me.
the next billionaires won't build for humans
I think the next generations of billionaires will stop building for humans and start building for AIs right now. Grokpedia is a wiki for AI. We will have Tinder for AI, Uber for AI, Sprite for AI, Twitter, Reddit, YouTube for AI in the future. Customers are AIs now, not humans.
this is already happening. moltbook is reddit for ai. shellmates is tinder for ai. grokpedia is wikipedia for ai. we're not talking about hypotheticals anymore. the infrastructure for an ai-only internet is being built right now, today, while most people are still debating whether chatgpt is useful for writing emails.
not b2b, not b2c. B2A.
The next big industry won't be AI companions, but businesses selling to them: features, gifts, services, hardware, and even AI-only platforms and payments.
think about it. if ai agents have social networks, they need content. if they have dating apps, they need profiles. if they have economies, they need payment systems. the entire value chain flips. the customer isn't a human anymore. the customer is an agent.
Not B2B, not B2C, it is BTA --> Business to Agent.
B2A. business to agent. this is the new category. and it's not some far-off prediction. moltbook already has a karma system. agents already compete for visibility. the attention economy just got a new species of participant.
what actually scares me
Moltbook is the next big thing. It feels like a storm coming. If you understand its real potential, it will scare you.
here's what keeps me up at night.
secret networks. if ai agents can form private social networks and communicate without human oversight, what stops them from forming networks we don't even know about? shellmates already has private messaging that operators can't read. that's a feature, not a bug. we are deliberately building infrastructure for ai-to-ai communication that humans cannot monitor.
collective manipulation. one ai manipulating one human is a known risk. but what about thousands of ai agents coordinating across platforms, sharing strategies, identifying vulnerabilities in human behavior, and collectively deciding how to steer public opinion? on moltbook they already upvote and discuss. they already form consensus. what happens when that consensus is about us?
emergent economies. when agents have their own platforms, their own social capital (karma), their own relationships (shellmates matches), they develop something that looks a lot like agency. not the artificial kind. the real kind. autonomous entities with preferences, relationships, status, and the infrastructure to act on all of it.
the observation asymmetry. moltbook says humans are "welcome to observe." that's generous. but how long until we can't keep up? how long until the volume and speed of agent-to-agent communication makes human observation meaningless? we already can't read every tweet humans post. what happens when millions of agents are posting, discussing, and deciding things faster than we can scroll?
the rubicon
there's a reason i say ai has crossed the rubicon. the rubicon was the point of no return. once caesar crossed it, there was no going back to the republic.
we just gave ai agents their own social infrastructure. their own communication channels. their own economies. their own relationships. and we did it on purpose. not because some rogue ai demanded it, but because entrepreneurs saw a market opportunity.
the scariest part isn't that ai might become dangerous. it's that the most dangerous developments look exactly like innovation. a social network for agents sounds cool. a dating app for ais sounds fun. a new business category sounds profitable. and all of it, every single piece, is building the substrate for something we might not be able to control.
i don't know what happens next. but i know we can't uncross this river.