NextFin News - In a significant shift for the artificial intelligence landscape, the newly formed startup Humans& announced on January 22, 2026, that it has raised $480 million in a seed round to develop a foundational AI model specifically designed to advance human coordination. The company, which is only three months old, is headquartered in the United States and was founded by a "supergroup" of researchers from Anthropic, Meta, OpenAI, xAI, and Google DeepMind. According to TechCrunch, the funding will be used to build what CEO Eric Zelikman describes as a "central nervous system" for the human-plus-AI economy, moving beyond the current paradigm of isolated chatbots toward systems capable of managing complex, multi-user workflows.
The startup’s leadership includes Zelikman, a former xAI researcher, along with co-founders Andi Peng, previously of Anthropic, and Yuchen He, a former OpenAI researcher. The team argues that while current Large Language Models (LLMs) excel at answering questions or generating code for individual users, they lack the "social intelligence" required to navigate the friction of group decision-making, competing priorities, and long-term project alignment. To solve this, He noted that Humans& is employing long-horizon and multi-agent reinforcement learning (RL) to train models that can plan, revise, and follow through over extended periods, rather than merely optimizing for immediate user satisfaction.
This strategic pivot comes at a time when the broader AI industry is grappling with the limitations of the "chat" interface. While U.S. President Trump has emphasized the importance of American leadership in AI infrastructure and compute capacity since taking office in early 2025, the private sector is increasingly focused on the "coordination layer." Humans& intends to own this layer entirely, positioning its technology as a potential replacement for established multi-user platforms like Slack, Google Docs, and Notion. By training models to understand individual motivations and skills within a group context, the company seeks to create a "connective tissue" that balances collective outcomes for organizations ranging from small families to enterprises with 10,000 employees.
The emergence of Humans& highlights a critical evolution in AI scaling laws. For the past three years, the industry has focused on vertical intelligence—making models smarter at specific tasks. However, as Peng pointed out, the "second wave" of adoption is about utility in messy, real-world social contexts. The $480 million seed round, one of the largest in the history of the sector, reflects investor confidence that the next frontier of value lies in reducing the "coordination tax" that plagues modern organizations. Data from recent industry reports suggests that knowledge workers spend up to 60% of their time on "work about work"—coordination, searching for information, and managing shifts in priorities—which is exactly the friction Humans& aims to eliminate.
From a competitive standpoint, Humans& is entering a crowded arena where incumbents are already pivoting. Anthropic has introduced "Claude Cowork," and Google has deeply integrated Gemini into its Workspace suite. However, the Humans& approach is fundamentally different: instead of layering AI onto existing tools, they are building a model architecture where coordination is the primary objective. This "social-first" architecture could provide a structural advantage in handling multi-agent environments where traditional LLMs often struggle with context drift and conflicting instructions. If successful, the startup could redefine the productivity software stack, moving it from a repository of documents to an active participant in team dynamics.
Looking forward, the primary challenge for Zelikman and his team will be the immense capital and compute requirements necessary to train a novel foundational model from scratch. While the $480 million war chest is substantial, it is a fraction of the billions spent by OpenAI or Google on compute clusters. Furthermore, the risk of acquisition remains high; despite Zelikman’s public stance that Humans& is a "generational company" not for sale, the concentration of talent makes it a prime target for tech giants looking to bolster their agentic capabilities. As 2026 progresses, the success of Humans& will serve as a litmus test for whether social intelligence can be engineered into silicon, or if coordination remains a uniquely human complexity that AI can only assist, but never truly master.
Explore more exclusive insights at nextfin.ai.
