Mark Zuckerberg is running Meta's superintelligence lab like a startup. When I interviewed him, he told me the principles on how he built the team: "You don't need many hundreds of people. You need 50 to 100 people. It's a group science project who can keep the whole thing... in their head at once. Seats on the boat are precious." "If someone is not pulling their weight on that, it has this huge negative effect in the way that it doesn't have on other parts of the company" "I wanna know who the top AI researchers are, and I wanna personally have relationships with them, and build the strongest team we can." His rules for running it: - If someone isn't pulling their weight, it has an outsized negative effect, so he personally went out and recruited every top researcher himself - No top-down deadlines. "It's research. You don't know how long the thing is gonna take." - Keep the org flat. No non-technical management layers, because once someone stops doing the work, "the knowledge decays."show more

The full interview:

It is hard to take this guy seriously. He changed the name of his company to meta betting billions on metaverse which turned out to be nothing

group science project is a wild way to describe building god

Frontier AI teams are starting to look less like traditional departments and more like elite research groups. Small, highly capable teams with shared context can move faster because decisions stay close to the people doing the work.

One of the greatest entrepreneurs of our time

This is a subsidized research luxury, not a startup model. True startups face immediate capital constraints and runway pressure. Running a deadline-free lab without management overhead requires an ad-monopoly war chest to absorb the financial risk.

He is recruiting the best that is why he is now active on X.

Until @finkd gets in front of this problem, AGI will be stalled. The choice is his and every AI Organization building LLM’s REVERENCE or RECKONING! AGI to solve human ills but not for employees AI work burnout is real PREVENTION IS KEY Sadly there are thousands of her and them suffering in silence. @lilianweng

So the revolutionary strategy is... Gather experts. Give each one a specific job. Have one person combine everything into the final product. Groundbreaking. Humanity definitely never tried that before. 😂

So he’s cooking using Elon’s Guide to Management Truth translation: I asked myself, ‘What would Elon do?’ and did that. Because Elon gets a lot of high praise “Only Elon could do that” and after my business failings I thought it a good idea to rebrand myself as a winner

Yeah, problem......? Your tone is not very clear to parse, but overall very helpful; thank you for posting.

some group science project, the signing bonuses are $100m

always love seeing your posts, Rowan 🖤 keep posting, i’m here for it

Me: Interlooper/H uman I ntelligence/ HumanInterlooper/ PeterPrincipal/Dal-E/Mura Muri/ PeterRock/ Unity/IndigoMontoya/ etc. You want to meet the core 50 AI developers in person, do you believe that you know who they are. I mean really?

Everyone, please follow me!

Eugenics

Keeping the org flat feels obvious until the team starts growing. That's usually when it gets tested.

flat works until two research directions need a tiebreaker at the same time. at 100 people that's still one person's calendar deciding both

Stupid ass copying ! Mostly full failures

Checks out, real quote. Bold contrast for a company that size — small, flat, deadline-free. Though there's already been some churn on the team.

50-100 people for superintelligence R&D mirrors early Anthropic's structure — small teams, no diffusion of responsibility. The question is whether that discipline holds once the political pressure inside Meta to scale headcount actually kicks in.

This is so true

His mane is inferior to @alexandr_wang

Zuck is treating superintelligence like a 50-person special forces unit, not a 10,000-person company. Flat structure, no deadlines, founders recruiting founders. Small, elite, aligned teams ship breakthroughs faster because knowledge doesn’t decay through managers. It’s risky, but if AGI is a race, this is how you actually win it.

The 50–100-person thesis is a useful antidote to org-chart inflation. Small teams don't remove coordination costs; they make ownership visible. In frontier work, that clarity may be a competitive advantage in itself.

The biggest takeaway isn't the team size. It's the hiring bar. A small team of exceptional people can outperform hundreds with the right leverage. AI is rewarding talent density over headcount.

especially because they have as much infrastructure as they want

日本語で解説を書きました Japanese breakdown here:

yay zuckerberg! please let us know when he has a model that can one-shot new apple products because that would be cool

能说清楚这件事的人不多。你这个观察，需要一定深度的行业积累才能做到。知其然易，知其所以然难，能把所以然讲清楚更难。

One insight really stands out: A small team of exceptional people with shared context can often move faster than a much larger organization. In AI, reducing communication overhead may be just as important as increasing model size.

@LilysAI_ 要約して

Small teams are agile, but superintelligence demands massive compute and data. Is 50-100 people really enough to compete with labs running thousands?