The war for AI talent has entered absurd territory. When Meta's Mark Zuckerberg attempted to recruit Andrew Tullock, a distinguished engineer from OpenAI, the offer was staggering: $1 billion in salary over six years, plus another $500 million in stock and bonuses. That's roughly $250 million annually—Cristiano Ronaldo money for a researcher. Tullock rejected it.
Instead, he chose to stay with Thinking Machines Lab, the startup he co-founded with Mira Muratti (former CTO of OpenAI) after leaving the ChatGPT maker six months ago. Thinking Machines has no product, no revenue, and no clear timeline to profitability. It raised $2 billion on reputation alone. When Meta made similar offers to other Thinking Machines employees, all of them declined. None defected from Anthropic either—only two researchers left that company when Meta came calling, suggesting people genuinely believe in their current missions more than Meta's Super Intelligence Team.
This reveals the brutal arithmetic of AI scaling. Money and GPUs are commodities; talent compounds. Zuckerberg understands this, which is why he's paying like a desperate billionaire. But the real signal is structural: if someone at Thinking Machines—an unfunded lab with no shipping product—turns down $1.5 billion, they must believe their equity stake will be worth far more. They're betting on reaching artificial general intelligence. And they're betting they can get there first with their team, not Meta's money.
The talent wars expose another truth: OpenAI, despite its dominance, is bleeding people. Ten researchers left for Meta. Thinking Machines and Safe Super Intelligence (Ilya Sutskever's startup) poached talent because they offered something money can't easily replicate—autonomy, mission alignment, and founder-level skin in the game. Meta can offer infinite capital. It cannot offer the feeling of being at the frontier with a small team betting the world.