The Person Who Co-Invented the Transformer Just Left Google for OpenAI
Two years ago, Google paid $2.7 billion to bring Noam Shazeer back. He just left for OpenAI.
If you don't know the name, here's why it matters: Shazeer is one of the eight co-authors of "Attention Is All You Need," the 2017 paper that introduced the Transformer architecture. Almost every AI system you've used in the last five years — ChatGPT, Gemini, Claude, Copilot — runs on that architecture. It's the foundational idea behind the modern AI boom.
The fact that he's now at OpenAI, on the eve of their IPO, is not a quiet personnel update.
How He Got Back to Google in the First Place
Shazeer left Google in 2021 to co-found Character.AI, the AI companion platform. He spent three years there building conversational AI at scale, working outside the lab that originally developed the Transformer.
Then in 2024, Google acquired Character.AI's technology and team for $2.7 billion. Shazeer came back as part of that deal, rejoining the team that had built much of the modern AI stack.
Less than two years later, he's gone again — this time to OpenAI.
What OpenAI Is Building Around This
Shazeer isn't the only big hire OpenAI announced this week. They also brought on Dean Ball, a former White House AI policy official, to lead a new "Strategic Futures" team. That team will focus on AI policy, governance, and the relationship between frontier AI labs and governments.
Taken together, these moves paint a clear picture: OpenAI is staffing up aggressively, on both the research and policy sides, before going public. They want the best builders and the best navigators in the room.
The Context Makes It More Interesting
This is happening at a particularly charged moment. Anthropic's most capable models were pulled from access by the US government this week — a dispute involving export controls and concerns about foreign access. OpenAI, meanwhile, is racing toward an IPO that could value the company near $1 trillion.
The AI talent market isn't just competitive. It's becoming a strategic asset in itself. Governments care who has access to the best researchers. Companies are paying extraordinary sums — literally billions — to secure them. And the researchers themselves are choosing where to place their bets on where the next breakthrough will happen.
Shazeer's move to OpenAI is a vote about where he thinks the frontier will be built.
Why Talent Signals Matter in AI
In most industries, personnel changes are worth noting but rarely decisive. In AI research, the best people aren't interchangeable — they shape the direction of entire architectures.
The Transformer paper had eight authors. Those eight people collectively changed the trajectory of the technology. Where they work matters, not just for morale or press releases, but because they tend to produce the ideas that get turned into the next generation of models.
When one of them moves between labs, it's worth paying attention to what they're betting on.
What This Means If You Use OpenClaw
OpenClaw runs on top of the models that come out of these labs — Claude, GPT, and others. The talent wars and research breakthroughs happening at Anthropic, OpenAI, Google, and elsewhere translate directly into more capable agents for anyone building or using AI tools.
When a Transformer co-author joins the lab that's about to go public with nearly a trillion-dollar valuation, it's a reasonable signal that the pace of capability improvements isn't slowing down. That's good news if you're building workflows around AI agents — the models they run on will keep getting better.
OpenClaw is designed to grow with that progress: as the underlying models improve, your agents get more capable without you needing to rebuild anything.