The AI industry has a new player with a contrarian pitch: what if the future of artificial intelligence isn’t about autonomous agents doing everything for us, but about technology that makes us better at working with each other?
Humans&, a newly launched frontier AI lab, has emerged from stealth with $480 million in seed funding and a $4.48 billion valuation.
The company is positioning itself as a “human-centric” alternative to the current AI race, where the goal seems to be building systems that can operate independently for as long as possible.
Instead, Humans& wants to create AI that acts as “connective tissue” within organizations and teams, strengthening relationships and collaboration rather than automating people out of the picture entirely.
From Google and Anthropic to Grok, The All-Star Founders of Humans&
The seed round was led by SV Angel and co-founder Georges Harik, with backing from some of tech’s biggest names: Nvidia, Jeff Bezos, and Google Ventures. For context, this is seed-round money sized like a late-stage war chest, the kind of capital typically reserved for teams with exceptional credentials and ambitious technical roadmaps.
The funding reflects investor confidence not just in the philosophy of Humans&, but in the team behind it. The founding group reads like an AI research supergroup, pulling talent from across the industry’s most influential labs.
Other co-founders include Andi Peng, who was a researcher at Anthropic and worked on reinforcement learning and post-training for Claude models; Georges Harik, who was the seventh employee at Google and assisted in developing the search engine’s first advertising systems; and Noah Goodman and Yuchen He, who are both AI researchers at X and developed the Grok chatbot and at Stanford University, specializing in psychology and computer science.
The team of about 20 people also includes experienced programmers from OpenAI, Meta Platforms Inc., and MIT.
It challenges to the current status quo is the company’s assumption that the current narrative of AI development, that it somehow equals freedom, is just that, a narrative. There is no question that the AI models are improving with regard to reasoning, coding, and operating independently.
“That was never my motivation,” co-founder Andi Peng said about the push toward autonomous AI. “I think of machines and humans as complementary.”
That philosophy translates into a surprisingly straightforward product vision: AI-powered instant messaging software that helps people coordinate, communicate, and solve problems together. Rather than an agent that disappears to complete tasks alone, Humans& envisions AI that lives inside how teams already work.
Moving from Single Sessions to Ongoing Partnerships
Making “collaborative AI” real requires solving some hard technical problems. Humans& points to needed innovations in long-horizon and multi-agent reinforcement learning, memory systems, and user understanding areas, where current AI often stumbles when tasks become complex, multi-step, and context-dependent.
The company is exploring two key directions. First, building AI that proactively asks for the information it needs and actually remembers it, moving beyond single-session interactions toward ongoing, assistive relationships. Second, developing multi-agent systems where multiple AI assistants can coordinate on complex work, mirroring how human teams naturally divide responsibilities.
The goal is software that can handle “long-horizon activities”, tasks that unfold over hours or longer, while supporting multi-agent collaboration and intelligently requesting inputs from team members when needed.
Betting on the Enterprise Adoption Gap
The launch comes at an interesting moment. Businesses are excited about generative AI, but many struggle to integrate it into daily operations in meaningful ways. Humans& is betting the biggest opportunity isn’t just smarter models, but better interfaces, better memory, and better integration into existing workflows.
The company frames this as building “deeper connective tissue” for organizations and communities, requiring rethinking both large-scale model training and how humans interact with AI systems.
What Success Looks Like
For now, Humans& remains early-stage, though unusually well-capitalized. The company has hinted at building neural networks designed to boost worker productivity through those long-horizon and multi-agent capabilities.
The real test will be whether this compelling philosophy translates into a product people actually use daily, rather than something they turn to occasionally. If Humans& succeeds, it won’t just be another AI assistant, it could define an entirely new category of AI designed not to replace teams but to become the glue that holds them together.
