In an era where artificial intelligence excels at answering questions and summarizing documents, a critical gap remains: AI chatbots still struggle with the complexities of real-world human collaboration, such as coordinating diverse priorities, tracking long-term decisions, and maintaining team alignment.

Addressing this frontier, a new startup named Humans&, founded by alumni from Anthropic, Meta, OpenAI, xAI, and Google DeepMind, recently secured an impressive $480 million seed round.

Humans& aims to construct a “central nervous system” for the human-plus-AI economy, moving beyond AI as mere assistants. The company plans a novel foundation model architecture designed for social intelligence rather than simple information retrieval or code generation.

Co-founder Andi Peng explained that the industry is moving from a phase of specialized question-answering models to one where consumers seek practical applications of AI.

Humans&’s pitch emphasizes empowering people and enhancing collaboration, rather than replacing jobs, addressing a growing need for coordinated AI workflows in complex team environments.

Building AI for Real-World Teamwork

Despite being only three months old and without a finished product, Humans& has leveraged its founding team’s pedigree to raise significant capital. While details remain sparse, the model could potentially replace multiplayer collaboration tools like Slack, Google Docs, or Notion. CEO

Eric Zelikman emphasized that their model will focus on communication and collaboration, helping teams work together effectively with each other and with AI tools.

He illustrated this with a simple example: deciding on a startup logo, a task current AI chatbots handle poorly because they lack social awareness and multi-user understanding.

Humans& is uniquely designing its product and model together, ensuring that as the model evolves, the interface and features adapt seamlessly.

The vision is to create a collaboration layer where the AI understands individual skills, motivations, and needs, acting as the “connective tissue” across organizations, teams, or even families.

Training involves long-horizon and multi-agent reinforcement learning (RL), enabling the model to plan, act, and adapt over time in multi-user environments while improving memory and user understanding.

Challenges and the Road Ahead

Humans& faces significant hurdles. Training and scaling a new foundation model requires enormous resources, putting the startup in direct competition with tech giants for compute and infrastructure.

It also competes with existing collaboration tools and major AI platforms that are embedding enhanced collaboration features—like Anthropic’s Claude Cowork, Google’s Gemini in Workspace, and OpenAI’s multi-agent orchestration workflows.

Unlike competitors, Humans& is rewriting its model core to prioritize social intelligence, potentially giving it a competitive advantage but also making it a likely acquisition target.

Reports indicate the company has already rejected buyout offers, expressing confidence in its team and long-term vision to transform AI-driven human collaboration.