In the accelerating contest to define artificial intelligence's next chapter, Jeff Bezos' six-month-old venture Project Prometheus has drawn Kyle Kosic — a computer scientist who helped build the computational spine of xAI's Colossus supercomputer — into its growing ranks. The hire is less a single career move than a symptom of a deeper reordering: engineers and researchers are choosing not merely between employers, but between visions of what AI should become and where it should live. Prometheus, with its $6.2 billion in backing and a mission rooted in the physical world, is emerging as one o
Bezos' Project Prometheus Hires xAI Cofounder Kosic in AI Talent War
All eleven cofounders have now departed from xAI
Mark
So Kosic left OpenAI in November and was apparently building something in stealth. How does that timeline work with joining Prometheus?
Mimi
The reporting says he was working as a builder at a stealth startup starting in November, which is when Prometheus was founded. It's possible he was involved early, or the timing is just coincidental. The article doesn't clarify whether he was at Prometheus during those months or joined more recently.
Luke
Right—and that's a gap. We don't know if he was a founder, an early hire, or if he just joined now. The article says it's unclear what capacity he's joining in, which is a pretty big thing to leave unresolved.
Mark
Why does his xAI experience matter so much?
Mimi
Colossus is the supercomputer that powers xAI's models. Building and maintaining that kind of infrastructure—the chips, the cooling, the networking, the software stack—is incredibly specialized work. Prometheus is explicitly trying to build AI systems that operate in the physical world, which means they need serious computational infrastructure.
Luke
But we should note: the article doesn't say what Prometheus' current infrastructure looks like or what gaps Kosic is filling. We're inferring that his expertise is needed based on what Prometheus says it does. That might be right, but it's inference.
Mark
The article mentions Bezos raising $100 billion for an AI manufacturing fund. Is that the same as Prometheus?
Mimi
The reporting says the fund would also finance Prometheus, so it sounds like Prometheus is one piece of a larger bet Bezos is making on AI manufacturing. The $6.2 billion Prometheus raised is separate from this $100 billion effort.
Luke
And we should be careful here: the $100 billion is what Bezos is seeking, not what he's secured. That's a proposal, not a done deal.
Mark
What does the exodus from xAI tell us?
Mimi
All eleven cofounders have left. That's not normal. It suggests either the company's direction shifted, or the founders found better opportunities elsewhere, or both. Kosic's move to Prometheus could be part of that pattern—top talent seeing a different future.
Luke
But we don't have statements from the people leaving about why they left. We know they left; we don't know their reasons. The article presents it as evidence of competition, which is fair, but it could also mean xAI has internal problems we're not seeing.
Il Polso
- A talent exodus is quietly destabilizing AI's most prominent labs — all eleven of xAI's cofounders have now departed, and OpenAI's leadership is cycling through medical leaves, resignations, and restructured roles.
- Kyle Kosic's move from xAI and OpenAI to Prometheus is a signal flare: engineers are following mission clarity and capital concentration, not just prestige or salary.
- Project Prometheus is barely six months old yet has already hired hundreds across three continents, drawing from Meta, Google DeepMind, and OpenAI with a focus on AI that can act in the physical world — manufacturing, aerospace, automotive.
- Bezos is not waiting for the talent to settle in before expanding the bet — he has been courting sovereign and institutional investors in Singapore and the Middle East for a $100 billion AI manufacturing fund.
- The competition is converging on infrastructure: whoever builds the computational and physical backbone for real-world AI systems may define the industry's next decade.
In the accelerating contest to define artificial intelligence's next chapter, Jeff Bezos' six-month-old venture Project Prometheus has drawn Kyle Kosic — a computer scientist who helped build the computational spine of xAI's Colossus supercomputer — into its growing ranks. The hire is less a single career move than a symptom of a deeper reordering: engineers and researchers are choosing not merely between employers, but between visions of what AI should become and where it should live. Prometheus, with its $6.2 billion in backing and a mission rooted in the physical world, is emerging as one of those gravitational centers pulling talent away from the industry's established poles.
Jeff Bezos' AI venture Project Prometheus has hired Kyle Kosic, a computer scientist whose recent years read like a map of the industry's most contested terrain. After two stints at OpenAI — the second ending in November 2025 — and a year at Elon Musk's xAI, where he led infrastructure development on the Colossus supercomputer, Kosic is now joining a company that is barely six months old but already moving with unusual force.
Bezos and physicist Vikram Bajaj founded Prometheus last November with a specific ambition: building AI systems that can understand and operate in the physical world. The target industries — manufacturing, aerospace, automotive, computing — are places where AI that can manipulate real environments, not just process language or images, could be transformative. To pursue that mission, Prometheus has hired hundreds of researchers across San Francisco, London, and Zurich, raised $6.2 billion in funding, and seen Bezos travel to Singapore and the Middle East in search of as much as $100 billion for an AI manufacturing fund.
Kosic's infrastructure background, including his master's in machine learning from Georgia Tech and his hands-on work building the computational backbone for large-scale AI at xAI, makes him a natural fit for Prometheus' hardware-intensive vision. His specific role has not been disclosed, but the hire reflects the company's need for engineers who have already built at scale.
The broader context is one of visible disruption at the industry's established players. All eleven of xAI's cofounders have now left the company. At OpenAI, the chief marketing officer stepped down for health reasons, the head of AGI development took extended medical leave, and the chief operating officer moved into a special projects role. A robotics and hardware engineer resigned citing concerns about the company's government partnerships.
These are not coincidental departures. They reflect a recalibration among the people who build AI — a growing sense that the next meaningful phase of the technology will be shaped not by scaling language models further, but by grounding AI in the physical world. For engineers like Kosic, Prometheus represents a convergence of capital, mission, and timing that is difficult to ignore.
Jeff Bezos' artificial intelligence venture Project Prometheus has brought on Kyle Kosic, a computer scientist who spent the last four years moving between two of the industry's most prominent labs. Kosic left OpenAI in November 2025 after a second stint there, having previously worked at the company from 2021 to 2023. In between those periods, he spent a year at Elon Musk's xAI, where he led infrastructure work on the company's Colossus supercomputer. His move to Prometheus, first reported by the Financial Times, signals the intensity of a talent war now reshaping the artificial intelligence sector.
Project Prometheus itself is barely six months old. Bezos and physicist Vikram Bajaj founded the company last November with an explicit mission: building AI systems capable of understanding and operating in the physical world. The startup is targeting manufacturing, aerospace, computing, and automotive industries—sectors where AI that can actually manipulate the real environment, not just process text or images, could unlock enormous value. To staff this ambition, Prometheus has already hired hundreds of researchers and engineers across San Francisco, London, and Zurich, drawing them from Meta, OpenAI, and Google DeepMind. The company received $6.2 billion in funding last November. More recently, Bezos traveled to Singapore and the Middle East seeking as much as $100 billion for an AI manufacturing fund that would also finance Prometheus' operations.
Kosic's background makes him a natural fit for this infrastructure-heavy mission. He holds a master's degree in computer science from Georgia Tech with a specialization in machine learning. His work at xAI on Colossus—the supercomputer infrastructure that powers the company's AI models—gave him hands-on experience building the computational backbone that large-scale AI systems require. At Prometheus, he is expected to focus on similar infrastructure challenges, though the company has not yet disclosed his specific title or reporting structure.
The hiring reflects a broader scramble for talent that is now destabilizing some of the industry's most established players. xAI, despite its prominence and Musk's backing, has seen all eleven of its cofounders depart. Ross Nordeen and Manuel Kroiss resigned last month. OpenAI, meanwhile, is undergoing significant leadership churn. Kate Rouch, the company's chief marketing officer, stepped down due to health issues. Fidji Simo, who leads AGI development, took medical leave for several weeks. Brad Lightcap, the chief operating officer, transitioned to a special projects role reporting directly to CEO Sam Altman, with his previous responsibilities distributed to chief revenue officer Denise Dresser. In March, Caitlin Kalinowski, who worked on OpenAI's robotics and hardware team, announced her resignation, citing concerns about the company's partnership with the U.S. Department of War.
These departures are not random. They reflect a fundamental shift in where ambitious engineers and researchers believe the next phase of AI development will happen. Prometheus, backed by one of the world's richest individuals and pursuing a specific, concrete vision of AI in the physical world, represents a different kind of bet than the scaling-focused work at OpenAI or the autonomous-vehicle focus at xAI. For engineers like Kosic, the move signals where the capital, the mission clarity, and the opportunity are converging. The talent war is not just about salary or prestige anymore—it is about which company will actually build the systems that matter next.
Citazioni salienti
Project Prometheus is building AI systems designed to understand and operate within the physical world, with applications across manufacturing, aerospace, computing, and automotive industries— Project Prometheus mission statement (via Financial Times reporting)
Vuoi la storia completa?
Leggi l'originale su The Economic Times ↗
Technology