Zuck’s outrageous playbook

Hi, and happy Thursday.

It’s not too late to join our next masterclass:

Last Summer, Mark Zuckerberg pulled off a stunt that would have gotten most CEOs fired.

(… Or at least ex-Uber CEO Travis Kalnick).

Meta’s Llama 4 AI model underwhelmed everyone.

Worse, there were no good options to repeat the acquisition playbooks that were Instagram, WhatsApp and Oculus (now Meta VR). 

Meta looked to be well and truly out of the super intelligence race.

So what could Zuckerberg do?

What followed was something we’ve never seen before: 

The $14.3bn acquisition of a 28 year old and a few of his colleagues. 

Not a company, its customers, products, intellectual property or anything else - just the 28 year old and his colleagues. 

At the time, it felt as desperate and crazy as the Metaverse, so we all shrugged it off.

But 15 months later, that said individual (now 29) has launched Muse - beating even ChatGPT in viral growth.

Meta stock has risen about 21% since Muse’s launch - about $380Bn of value - essentially paying back that $14.3Bn investment over 26 times.

Muse is everything we’ve been talking about in this newsletter all year - AI employees - but easy-to-use, without the setup headaches of OpenClaw, and a price point you can’t beat: free. 

Like most AI agents, Muse’s effectiveness comes from how many of your accounts and systems you connect it to. Nevertheless, here are a few simple use cases you can try without handing over your life:

• Booking a business lunch. “Find a quiet restaurant near [location] for four people this Thursday at noon, and reserve a table.” explicitly supports reservations with approval.

• Purchasing a small item for your workspace. “Find a laptop stand compatible with [model], under $40 delivered, and buy it once I approve.” 

• Submit a service request. “Report the missed garbage pickup through my municipality’s website and give me the confirmation number.”

But how did Zuckerberg turn Meta’s AI story around?

And what takeaways are there for the rest of us?

Well, it started three years ago, when we learned that you don’t need to acquire a company to get what it has. 

Over the course of a November weekend in 2023, Microsoft almost hired everyone at OpenAI following the firing of CEO Sam Altman.

All Zuckerberg had to do was figure out how to close that loop for himself.

But how do you raid the top labs without instigating an event like Altman’s fire?

And how do you even get on the phone with those people you need to hire?

Zuckerberg’s answer?

9 figure signing bonuses, as in - $100,000,000.

And for that direct line? That’s where the aforementioned 28 year old, Alex Wang, comes in.

You might not know this, but the AI labs rely on “tier 2” suppliers who go to all kinds of lengths to get the data the AI labs need, without the labs having to “get their hands dirty.”

The leading such supplier?

Scale AI, founded and led by Alex Wang.

When you are the #1 supplier of data to the labs, you also have a direct line to the researchers at the labs using that data.

So what did Zuckerberg do?

Meta acquires a 49% stake in Scale AI for 14.3Bn and hires Wang and a few of his colleagues, while leaving   Scale AI to continue to operate, business as usual.

They called it win-win-win for Scale AI, its investors and Meta.

I’m not sure the other AI labs would have described it in the same way.

Zuckerberg and Wang proceeded to raid talent at a scale and vigor that was catnip for internet memes. After all, who would have thought a career in AI research would lead to out-earning the world’s top movie stars, sports players, well-known CEOs and just about everyone else.

15 months later, that outrageous playbook appears to have worked.

In Muse, Meta has a new “billion dollar” product and is firmly back in the super-intelligence race.

And there’s a lesson in this for the rest of us.

While corporate leaders are being told to replace or super-charge their people with Copilot, Zuckerberg’s answer to Meta’s most consequential problem was to spend billions recruiting exceptional people. 

Meta is among Anthropic's biggest clients, so staff having access to top tier Fable-class models was not sufficient.

Zuckerberg still needed to get the right humans in the room. And he was willing to pay extraordinary sums to do it. That should give the rest of us pause.

AI can make more of the work cheaper and faster. 

But figuring out what to build, seeing what everyone else has missed, and making it actually work?

The people who can do that are becoming more valuable. 

At Prescouter we’ve long called them “subject matter experts” - both our internal domain experts and specialists from our expert network. They’re people who’ve lived with the problem for years, if not decades. They know the right questions to ask, how to think about the problem and how to achieve the ultimate end goal.

In a very small way - comparative to the Muse story - we too are seeing the outsized returns and value that exceptional talent brings to the work we do for clients.

Best,

Dino