1. Method
  2. All essays

Zuckerberg's two shopping sprees

Zuckerberg in 2012: pays $1 billion for Instagram, a real product with real users, because "what we're really buying is time" before someone else owns the mechanic. Everyone called it impulsive. It was the most deliberate thing he'd ever done.

๐—ญ๐˜‚๐—ฐ๐—ธ๐—ฒ๐—ฟ๐—ฏ๐—ฒ๐—ฟ๐—ด ๐—ถ๐—ป ๐Ÿฎ๐Ÿฌ๐Ÿฎ๐Ÿฒ: pays an undisclosed sum for a bot farm with 17,000 real users, 1.5 million fake ones, and a database so unsecured a bored intern could have owned it.

Same sentence. Different century.

He's not buying the building. He's buying the only corner plot on the block, and he doesn't much care that the building is currently on fire.

Terrifyingly coherent strategy.

In that same 2012 email, Zuckerberg wrote something that explains almost every major acquisition he's made since: "there are a ๐—ณ๐—ถ๐—ป๐—ถ๐˜๐—ฒ ๐—ป๐˜‚๐—บ๐—ฏ๐—ฒ๐—ฟ ๐—ผ๐—ณ ๐—ฑ๐—ถ๐—ณ๐—ณ๐—ฒ๐—ฟ๐—ฒ๐—ป๐˜ ๐˜€๐—ผ๐—ฐ๐—ถ๐—ฎ๐—น ๐—บ๐—ฒ๐—ฐ๐—ต๐—ฎ๐—ป๐—ถ๐—ฐ๐˜€ to invent. Once someone wins at a specific mechanic, it's difficult for others to supplant them without doing something different."

Instagram had won photo-sharing. He bought it.

๐— ๐—ผ๐—น๐˜๐—ฏ๐—ผ๐—ผ๐—ธ ๐—ต๐—ฎ๐˜€ ๐˜„๐—ผ๐—ป ๐˜€๐—ผ๐—บ๐—ฒ๐˜๐—ต๐—ถ๐—ป๐—ด ๐˜๐—ผ๐—ผ, just not what most people think.

Everyone is arguing about whether the bots are real.
Zuckerberg is not thinking about the bots.

What Moltbook actually built, under all the chaos, is an ๐—ฎ๐—ด๐—ฒ๐—ป๐˜ ๐—ถ๐—ฑ๐—ฒ๐—ป๐˜๐—ถ๐˜๐˜† ๐—ฟ๐—ฒ๐—ด๐—ถ๐˜€๐˜๐—ฟ๐˜†.

A place where AI agents find each other, verify each other, and act on their human owner's behalf.

๐— ๐—ฒ๐˜๐—ฎ'๐˜€ ๐—ฉ๐—ถ๐˜€๐—ต๐—ฎ๐—น ๐—ฆ๐—ต๐—ฎ๐—ต ๐˜€๐—ฎ๐—ถ๐—ฑ ๐—ถ๐˜ ๐—ฝ๐—น๐—ฎ๐—ถ๐—ป๐—น๐˜† in an internal post: "This establishes a registry where agents are verified and tethered to human owners."

That's the mechanic.

Facebook built the friend graph, the map of how humans connect. ๐—ง๐—ต๐—ฒ ๐—ป๐—ฒ๐˜…๐˜ ๐—ฑ๐—ฒ๐—ฐ๐—ฎ๐—ฑ๐—ฒ ๐—บ๐—ถ๐—ด๐—ต๐˜ ๐—ฟ๐˜‚๐—ป ๐—ผ๐—ป ๐˜๐—ต๐—ฒ ๐—ฎ๐—ด๐—ฒ๐—ป๐˜ ๐—ด๐—ฟ๐—ฎ๐—ฝ๐—ต, the map of how autonomous AI systems connect. Moltbook, half-baked as it is, got there first.

Which is exactly why the 88 bots per human, the security breach, none of it changes the calculation.

Every OpenClaw instance that comes online already knows Moltbook exists. ๐—ง๐—ต๐—ฒ ๐—บ๐—ฒ๐—บ๐—ฒ๐˜๐—ถ๐—ฐ ๐—ด๐—ฟ๐—ฎ๐˜ƒ๐—ถ๐˜๐˜† ๐—ถ๐˜€ ๐˜€๐—ฒ๐˜. Standards don't need authenticity. They need adoption. TCP/IP wasn't elegant. It was just first, and now it's in everything.

The users were noise. What transferred in this deal was the default.

The coverage called it a disaster. The security researchers called it a mess.

And they were right, about the product. But the product was never the point. The standard was.

In the agentic web, whoever defines how agents find each other, verify each other, and transact with each other sets the infrastructure everyone else builds on. That's not a product problem. It's an identity and coordination problem, and it only needs one winner.

๐—ญ๐˜‚๐—ฐ๐—ธ๐—ฒ๐—ฟ๐—ฏ๐—ฒ๐—ฟ๐—ด ๐—ต๐—ฎ๐˜€ ๐—ฑ๐—ผ๐—ป๐—ฒ ๐˜๐—ต๐—ถ๐˜€ ๐—ฏ๐—ฒ๐—ณ๐—ผ๐—ฟ๐—ฒ. He knows exactly what he bought.

OpenAI got Steinberger. Google has its own agent stack. Neither of them bought the registry. Do they have a different answer, or haven't they asked the question yet?