So Meta bought Moltbook this week. A social network where the users are literally bots. I had to read that headline twice because I thought I was having a stroke.
This is the NFT phase of product strategy. Loud, hyped, and aging in dog years
But then I remembered I said the same thing about Amazon “burning cash” for years. I was wrong then. Could be wrong now. Let’s dig in.
Meta’s Recent Track Record Is… Not Great
Let’s look at what Meta’s been up to lately:
- Avocado got pushed to May (at least). Their big AI model launch? Delayed because it wasn’t competitive. Sources say they’re even talking about licensing Gemini instead. NYT | Reuters
- 20% layoffs incoming. Word is Meta’s planning massive cuts to offset AI infrastructure costs. That’s not “strategic restructuring” – that’s panic. Reuters | CNBC
- Peter Steinberger went to OpenAI. The guy who created OpenClaw got hired by the competition. Meta lost the talent war and had to settle for buying the platform he built. Like showing up to the prom solo but getting the limo the cool kids rented. TechCrunch | Bloomberg
Not exactly the resume of a company making brilliant strategic moves.
But Here’s Where It Gets Weird
Google bought YouTube in 2006 for $1.65 billion. Everyone called it overpriced. “A search company buying cat videos?” It was bandwidth nightmare, copyright chaos, and had zero monetization.
YouTube is now worth around $400 billion. The “dumb” move became one of tech’s best acquisitions ever.
Sometimes the move that makes no sense in the moment is actually genius. You just need to know what you’re really buying.
The “Agent Graph” Is The New Friend Graph
Facebook built the “friend graph” – mapping connections between people. That was their whole business for two decades.
But Zuckerberg’s been saying something interesting. In Meta’s Q2 2025 earnings call, he talked about a future where “every business will soon have a business AI, just like they have an email address, social media account, and website.” Meta Q2 2025 Earnings
An agentic web needs an “agent graph” – mapping how AI agents connect and what they can do for each other.
Picture this:
- Your personal agent finds you a flight
- It negotiates with the airline’s agent for an upgrade
- It books dinner through OpenTable’s agent
- It buys that shirt by haggling with the retailer’s agent on price and color
For this to work, agents need to find each other, connect, and coordinate. Moltbook was literally testing this – agent discovery, relationships, transactions.
Meta didn’t buy a social network. They bought the mall before anyone knew shopping malls were going to be a thing. They bought the infrastructure for a world that doesn’t exist yet, but might.
The Commerce Play Nobody’s Talking About
Here’s where Meta’s ad business meets the agentic web.
Right now: Ads influence humans. Humans click. Humans buy.
Agentic web: Your AI shops for you. No banner ads. Just your agent negotiating with business agents based on your actual preferences.
- Only eco-friendly companies? Your agent knows.
- Prefer small biz over corps? Your agent filters for that.
- Only buy on sale? Your agent waits for price drops.
If Meta owns the orchestration layer, deciding which agents talk to which, when, and how – they don’t just show ads. They control the transaction flow.
That’s not advertising. That’s commerce infrastructure. And it’s way bigger.
Meta’s Pattern: Bad at Building, Great at Buying
Look, Meta has a type. They make big, confusing bets that look stupid until they suddenly don’t:
- Instagram ($1B) – Overpriced photo app with no revenue. Now worth hundreds of billions.
- WhatsApp ($19B) – Absurd money for messaging. Now essential global infrastructure.
- Oculus ($2B) – VR that took a decade to matter. Finally showing promise (maybe).
They’re terrible at building AI from scratch. Avocado delays prove it. They can’t keep talent, people keep walking out the door.
But they’re really good at buying things that capture how people (and now agents) want to connect.
The Trust Question
All of this hinges on one thing: will people actually trust AI agents with their wallets?
Will you let an AI negotiate prices for you? Know your preferences well enough to shop automatically? Have access to your payment info?
The fact that OpenClaw exists means some people are already there. Moltbook showed agents are ready to transact.
The infrastructure is being built. The agents are getting smarter. The only question left is whether we’re ready to hand over the keys.
If Meta cracks agent trust, secure transactions, transparent negotiations, user control, etc. They don’t just win the AI race. They redefine commerce.
Final Take
I started this thinking Meta bought a social network for bots. Which sounds like a skit from Saturday Night Live.
But I think I was wrong. They bought the infrastructure for a world where AI agents do business with each other.
While Google, OpenAI, and Anthropic fight over who has the best chatbot, Meta is positioning to own the layer where agents actually do business.
Was this brilliant or boneheaded? Ask me in 2027. But I’m starting to think…IDK, maybe this is their YouTube moment?
What do you think?
Sources:
