Y Combinator · Startup School 2026
Alexandr Wang: “This is a Once-in-a-Civilization Opportunity”
Scale AI’s founder and Meta’s head of superintelligence research tells a room full of young founders why the next decade rewards vision and ambition over technical horsepower — and how a well-tuned swarm of agents can already outpace a team of 100 engineers.
Tap a timestamp pill below to jump the video to that moment.
The problem
When every investor tells you no, how do you know you’re right?
Alexandr Wang founded Scale AI in 2016 at 19, after dropping out of MIT after one year. The pitch was simple: AI training was hitting a wall not because of models or compute, but because there was no clean, efficient way to get labelled data at scale. The problem was obvious to Wang, who had trained models at MIT and knew the pain firsthand. It was invisible to almost everyone else.
For years after launch, every time Scale went out to raise, investors pushed back. The numbers were strong, the revenue was real — but VCs called data “unsexy” and questioned the business’s durability. Most of those same investors are now writing think pieces about data being the central business opportunity in AI. Wang made it through those years not by changing course, but by building a deeper conviction that the people saying no had simply never trained a model themselves. They were reasoning from press coverage; he was reasoning from first principles.
The challenge he’s raising for the founders in the room at Startup School 2026 is the same one Scale forced him to solve a decade ago: how do you identify a truth about the world early — long before it becomes consensus — and then hold that belief through years of market noise, investor skepticism, and people around you who just don’t see it yet?
“you need to develop conviction in a set of beliefs that nobody else agrees with”
“I think the key thing is you need to develop conviction in a set of beliefs that nobody else agrees with. Like I think if you look at all the most successful companies in the world, they were started at a time long before the sort of like core idea was popular, and they toil in obscurity for years and years before the idea or the space or the concept or the business becomes consensus. And the only way you’re going to be successful is if you’re able to identify these truths about the world early, long before everyone else. And I think that — I mean, one of the most surprising things — like, you know, Scale, we’ve been working on AI for a decade. You just can’t base your business decisions based on what everyone else is saying around you. Like if you go too much with the herd, you will get immensely confused and you will end up nowhere.”
How they solve it
Build an internal compass, find the steepest exponential, then ride it with agents
Wang’s answer to the conviction problem is not motivational. It’s structural. You have to develop what he calls an “internal compass” — a personal, rigorously maintained model of how the future will unfold — and treat that model as the signal while tuning out the rest as noise. The founders who built the most important companies of the last several decades did exactly this: they saw something the market had not priced in yet, and they stayed the course long enough for the world to catch up.
The second piece of the framework is identifying the right exponential to bet on. Wang frames it as finding the curve with both the steepest slope and the longest runway. A decade ago, that was Moore’s Law. Right now, it is AI progress. The starting point of the curve doesn’t matter — when Wang started Scale, the applications were cat detectors in YouTube videos. What matters is the rate of change and whether the curve has decades left in it. He’s emphatic that we are still near the bottom of the AI curve, not the top: “every wave is ten times bigger than the last.”
What makes this moment genuinely different from prior technology waves, Wang argues, is that AI has already shifted what the binding constraint is. The bottleneck is no longer intelligence — it is vision and ambition. The models are already powerful enough that the limiting factor on what gets built is whether someone has a clear picture of what they want the world to look like and the determination to make it happen. This is the “once-in-a-civilization opportunity” he is pointing at: not that AI is impressive, but that for the first time, cognitive leverage is cheap and ambition is the scarce resource.
“it’s probably a once-in-a-civilization opportunity to be a dreamer”
“I think it’s as a result it’s like one of the most incredible — it’s probably a like once-in-a-civilization opportunity to be a dreamer and to have a vision and to have ambition and to impose a view of how the future world should look by building something amazing. You know, one of the things that we were chatting about backstage is, when I started Scale, or you know 10 years ago, if you start 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 now I actually think with the power of agents and AI broadly speaking, it’s much closer to Goliath versus Goliath — but maybe the startup is like a mega Goliath that is vastly enhanced by the power of agents.”
The most concrete tactical claim Wang makes is about agentic loops. He describes a pattern he has seen work at Meta: define a tight feedback loop inside a business — one where the inputs, the metric, and the optimization target are clearly specified — then let a swarm of agents run it. Where humans used to sit on the edges of those loops (getting customers, making them happier, spending the resulting revenue to get more customers), agents can now occupy those edges and run the loop far faster. He has seen internally at Meta that the right agentic setup can outperform a team of 100 engineers “very handily, very easily.” The abstraction is not magic: it comes down to markdown files, cron jobs, and a well-defined eval. But the leverage, he says, is enormous.
“a swarm of agents can accomplish more than a team of 100 engineers”
“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 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 these agentic coordination problems — I think that is 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 maybe isn’t in distribution? —— Yeah. Mechanically, figuring out what the metric is and then yeah it just comes down to skills, markdown files, cron jobs. Goal. Yeah, like — I think it’s always funny how mundane everything is once you really dig into it.”
On the skills question — whether CS degrees still matter as the abstraction layer rises — Wang lands somewhere nuanced. He does not think Word Cell beats Shape Rotator. Systems thinking, he says, is never going out of style: the game is the same whether you are organizing 10 humans, 100 agents, or a trillion agents working together. What changes is that founders also need a deeper philosophical compass now, a coherent view of where civilization is headed, because the decisions that get made in the next decade will shape the world more than anything in the past century.
Takeaway
The quick version
- Build an internal model of the future that is independent of market consensus — conviction in a contrarian belief, held long enough, is what separates the companies that define categories from those that follow them.
- Find the curve with the steepest slope and the longest runway. Right now that is AI. The starting point looks boring; the endpoint is civilization-scale. Get on it early and stay on it.
- Agentic loops — a clear metric, a tight feedback cycle, a swarm of agents running it — are the highest-leverage bet for builders today. The implementation is mundane. The output is not.
- The scarce resource has shifted. Intelligence is becoming abundant. Vision and ambition are the new bottleneck: do you have a clear picture of what the world should look like and the drive to build it?
“The scarce resource isn’t going to be intelligence or agency. I really think it’s going to be vision and ambition. Do you have a clear view of what you want the world to look like in 5 to 10 years that it does not look like today?”— Alexandr Wang, Startup School 2026
Wang closed by offering every founder in the room $1,000 in Spark API credits. The invitation was pointed: use the cheapest frontier model on the market to build the thing nobody else believes in yet. Start on the exponential now — the starting point always looks boring.