Humans&, a new AI startup founded by former employees of Anthropic, Meta, OpenAI, xAI, and Google DeepMind, is betting that the next major leap in artificial intelligence lies not in smarter answers, but in better coordination. The company has raised a $480 million seed round to build what it calls a “central nervous system” for the human-plus-AI economy—a foundation model designed for social intelligence, not just information retrieval or code generation. While current AI chatbots excel at answering questions, summarizing text, and solving equations, they remain isolated tools, optimized for one-on-one interactions. Humans& aims to close the gap in real-world collaboration: managing team dynamics, tracking decisions over time, aligning diverse priorities, and maintaining continuity across complex workflows. The company’s co-founders, including Andi Peng (former Anthropic researcher) and CEO Eric Zelikman (ex-xAI), argue that AI is entering a new phase—beyond individual assistance and into collective action. “We’re moving from the first wave of scaling, where models were trained to be smart in narrow domains, to a second wave where users are trying to figure out how to use all these tools together,” Peng said. Despite having no product yet, Humans& is building a model and interface in tandem, with the goal of creating a system that feels like a collaborative teammate—someone who understands not just what you say, but who you are, your goals, and your relationships with others. The team envisions the model as a central hub for communication and coordination, potentially replacing or augmenting platforms like Slack, Google Docs, or Notion. The model will be trained using long-horizon reinforcement learning, enabling it to plan, adapt, and follow through on multi-step tasks. It will also use multi-agent reinforcement learning, simulating interactions between humans and AI agents to improve coordination over time. The system will be designed to remember past interactions, understand individual motivations, and help balance competing interests for the benefit of the group. This vision places Humans& at the intersection of AI, teamwork, and workflow design. The company’s ambition is to become the connective tissue across organizations—whether large enterprises or small teams—by understanding both human and AI roles within a collective. The timing is strategic. As AI moves from isolated tools to integrated agents, the coordination bottleneck is becoming a major challenge. LinkedIn founder Reid Hoffman has argued that the real value of AI lies not in automation, but in how it enables teams to share knowledge, run meetings, and align on decisions. Humans& is not just competing with collaboration tools. It’s challenging the giants of AI—Anthropic, OpenAI, Google, and Meta—each of which is developing its own approaches to team-based AI. Anthropic’s Claude Cowork, Google’s Gemini in Workspace, and OpenAI’s multi-agent orchestration tools are all pushing in similar directions. But Humans& believes it has a unique edge: a model built from the ground up for social intelligence, not just task completion. The company has already turned down acquisition offers and insists it is building a generational company. “We believe this has the potential to fundamentally change how we interact with AI,” Zelikman said. Still, the path is uncertain. Training a new foundation model requires massive capital and access to compute—resources dominated by the largest AI labs. And with top talent in high demand, the risk of acquisition remains real. But for now, Humans& is focused on building something new: a system that doesn’t just answer questions, but helps people work together.