100 engineers outworked by a single swarm of AI agents inside Meta. And the setup behind it is just markdown files and cron jobs. That receipt comes from the man who runs Meta Superintelligence Lab himself. Alexander Wang, the founder. Mark Zuckerberg paid $14.3 billion to bring in on the Y Combinator stage days ago. By the end of this video, you'll have the exact anatomy of that swarm — the loop, the metric, the files, plus the one resource he says is about to become scarce. And the exponential he'd tell his 18-year-old self to bet everything on. We watched all 32 minutes and fact-checked the wild claims. Moment one is the swarm.
We've seen internally at Meta cases where if you can develop the right agentic loop and you have the right eval, the right metric for the agents to optimize, you can have a swarm of agents accomplish more than like a team of 100 engineers, very very handily actually. Very very easily.
Quick context because the source matters. That's Alexander Wang. He founded Scale AI at 19. And last year, Meta paid $14.3 billion for 49% of it. Mostly to put him in charge of Meta Superintelligence Labs. He's on stage with Garry Tan, the president of Y Combinator at the same startup school where Jensen Huang and the creator of Claude Code spoke. We've broken down both talks, links below. The full talk is linked too. Here's what most people will scroll past. Moment one. Garry asks him for alpha — the applications people haven't figured out yet. Wang doesn't name an app. He describes a loop and then drops the internal receipt that made this whole video. Listen.
I mean, I think there's still just like astronomical opportunity in agentic looping and figuring out how you develop systems that enable you to spend like 1,000x more or 1 million x more on tokens to drive an outcome in a continuous feedback loop. Like if you think about most companies, companies are just these large-scale feedback loops where humans are operating each of the edges. Companies, they get customers and they figure out how to make those customers happier. And if customers are happier, then they spend more. And if they spend more, then you can hire more people who can then go figure out how to get more customers and make those customers happier. And that in some sense is the feedback loop of every startup or every business. And within that there are micro feedback loops that exist. And I think developing agentic systems that can operate and optimize these feedback loops — there's just huge amounts of alpha there. Like I think we've seen internally at Meta cases where if you can develop the right agentic loop and you have the right eval or the right metric for the agents to optimize, you can have a swarm of agents accomplish more than like a team of 100 engineers, very very handily actually, very very easily. And so, I think figuring out what the world looks like with lots of sort of these agentic coordination problems — I think that is like one of the most interesting problems today.
So mechanically speaking, I mean, markdown files, cron jobs — is it that and then basically pointing the agent at enough data so that it can figure something out that, you know, maybe isn't in distribution?
Yeah, I think figuring out the parts — yeah, yeah, mechanically figuring out what the metric is and then yeah, just comes down to skills, markdown files, cron jobs.
Yeah, slash goal. Like I think it's always funny how mundane everything is once you really dig into it.
So, it's not magic. You know what I mean? Some people put a lot of magic — there's like some LinkedIn threads out there about some magic stuff.
LinkedIn is where you get customers.
Read the anatomy back. Because it's three parts: a metric — the eval that defines what better means; a loop — agents that keep optimizing that metric; and fuel. He's saying spend a thousand times more tokens on an outcome. Because tokens are now cheaper than engineers. The part that should stop you: the creator of Flood Code stood on this exact stage days earlier and said, 'Verification is the single most important thing people get wrong.' Wang just independently named the same thing — the eval — as the difference between a swarm that beats a hundred engineers and a pile of API bills. And both men, unprompted, gave the same warning: ignore the LinkedIn gurus. The mechanics are mundane. Skills, markdown files, cron jobs. The moat is choosing the metric. Moment two. If a swarm can do the work of a hundred engineers, what happens to the oldest rule in startups? That you're David and the incumbent is Goliath. Wang says that whole frame is dead.
You know, when I started Scale, 10 years ago, if you started a company, you had to be — it was like David versus Goliath and you had to be clever and you had to find like an angle into the market and you had to sort of figure out a way to compete even though they had much fewer resources. And now I actually think with the power of agents and AI broadly speaking, it's much closer to Goliath versus Goliath. Like I think but maybe the startup is like a Mecca Goliath that is like vastly enhanced by the power of agents and AI and, you know, the large companies are the sort of more traditional Goliath, so to speak. But I think that startups now, like if you properly embrace AI agents and figure out the way to leverage their strengths in the most ambitious ways, you can easily outcompete incumbents.
Ten years ago, a startup won by being clever in the cracks. Wang's claim is that agents delete the resource gap itself. The incumbent has a hundred engineers and you just watched what replaces them. What the incumbent can't copy is speed of conviction. A two-person team that embraces agents fully is, in his words, a Mecca Goliath. That word is your free PDF keyword later, by the way. Moment three is why the window is open right now. Moment three. This is the line the title of the talk comes from and everyone will quote it. Almost everyone will miss what he says right before it. The part that's actually the business plan.
