We are very excited to announce our investment in humans&, a human-centric frontier AI lab founded by Eric Zelikman and Georges Harik. We partnered with them at the earliest stages when they were just getting their idea off the ground (before they’d even found an office).
What convinced us to partner at this earliest stage was the combination of exceptional research credentials in exactly the areas that matter, proven experience shipping models used by billions, and the conviction to pursue a fundamentally different vision for AI.
A Different Vision for AI
No one changes the world alone. While AI models are rapidly learning to reason better and act more autonomously, real progress happens when humans understand one another, build trust, and work together. We believe that’s where the next chapter of AI will begin.
humans& is reimagining AI not just as a tool for automation, but as the connective tissue that strengthens how people interact together. This means fundamentally rethinking how models are trained and how people interact with AI. Instead of optimizing for single-turn task completion, they’re training models from the ground up to pursue long-horizon goals through genuine collaboration by helping users clarify intent, propose plans, negotiate tradeoffs, and maintain context over time.
Eric Zelikman pioneered reasoning models with his Self-Taught Reasoner (STaR) paper before any major industry releases, and his early work on training language models as user simulators directly informs humans&’s approach to multi-turn collaboration. At xAI, he showed he could scale research insights to production, leading post-training for Grok 2 and developing the agent RL recipe for Grok 4 with small teams.
Georges Harik brings a rare combination of product sense and research chops. As Google’s 7th employee, he wrote the business plan for AdWords and helped define AdSense’s targeting engine (products generating hundreds of billions in revenue), while his recent research on Quiet-STaR and probabilistic programming demonstrates continued technical leadership at the frontier.
The broader founding team has led major efforts across xAI, Anthropic, Google DeepMind, OpenAI, Meta, Stanford, and MIT, with deep expertise in the exact challenges ahead: long-horizon RL, multi-agent collaboration, human-AI interaction design, and value alignment.
Why We’re Excited
While the big labs continue to optimize for benchmark performance and single-turn task completion, humans& is pursuing a fundamentally different vision: AI that strengthens how people work together in organizations and communities rather than just automating tasks away.
That’s a future we believe in. AI’s greatest impact will come from making human connection and collaboration deeper and more effective. We’re excited to partner with humans& on this journey!
