AI chatbots are getting better at answering questions, summarizing documents, and solving mathematical equations. However, they still largely behave as helpful assistants for one user at a time. They are not designed to manage the messier work of real collaboration, such as coordinating people with competing priorities, tracking long-running decisions, and keeping teams aligned over time.
Humans&, a new startup founded by alumni of Anthropic, Meta, OpenAI, xAI, and Google DeepMind, believes closing that gap is the next major frontier for foundation models. The company this week raised a 480 million dollar seed round to build what it calls a central nervous system for the human-plus-AI economy.
The startup’s framing of AI for empowering humans has dominated early coverage, but the company’s actual ambition is more novel. It aims to build a new foundation model architecture designed for social intelligence, not just information retrieval or code generation.
One of Humans&’s co-founders, Andi Peng, a former Anthropic employee, explained that the first paradigm of scaling involved question-answering models trained to be very smart in particular verticals. Now, she believes, we are entering a second wave of adoption where the average user is trying to figure out what to do with all these AI tools.
Humans&’s pitch centers on helping usher people into this new era of AI, moving beyond the narrative that AI will take their jobs. The timing is critical as companies transition from chat interfaces to AI agents. While models are competent, workflows often are not, and the coordination challenge remains largely unaddressed, leaving many people feeling threatened and overwhelmed by AI.
The three-month-old company has managed to raise its substantial seed round based on this philosophy and the pedigree of its founding team. Humans& still does not have a product, nor has it been clear about what exactly that product might be. The team suggested it could be a replacement for multiplayer or multi-user contexts like communication platforms or collaboration platforms. For use cases and target audience, the team hinted at both enterprise and consumer applications.
Eric Zelikman, co-founder and CEO of Humans& and a former xAI researcher, stated that the company is building a product and a model centered on communication and collaboration. The focus is on helping people work together and communicate more effectively, both with each other and with AI tools. He gave an example of the tedious process of getting a large group to agree on something like a company logo, implying their technology would streamline such collaborative decisions.
Zelikman added that the new model will be trained to ask questions in a way that feels like interacting with a friend or colleague who is trying to get to know you. He contrasted this with current chatbots, which ask questions constantly but without understanding their value, as they are optimized primarily for user satisfaction and answer accuracy.
Part of the lack of clarity around the product is that Humans& is designing it in conjunction with the model itself. Co-founder Andi Peng said the company is ensuring that as the model improves, the interface and the model’s capabilities co-evolve into a coherent product.
What is clear is that Humans& is not trying to make a new model that plugs into existing applications. The startup wants to own the collaboration layer entirely.
AI combined with team collaboration and productivity tools is an increasingly hot field. For example, the AI note-taking app Granola recently raised 43 million dollars at a 250 million dollar valuation as it launched more collaborative features. Several high-profile voices are also framing the next phase of AI as one of coordination and collaboration, not just automation. LinkedIn founder Reid Hoffman recently argued that the real leverage of AI is in the coordination layer of work, such as how teams share knowledge and run meetings.
That is the space where Humans& wants to operate. The idea is that its model and product would act as connective tissue across any organization, understanding the skills, motivations, and needs of each person and balancing them for the collective good.
Achieving this requires rethinking how AI models are trained. Yuchen He, a Humans& co-founder and former OpenAI researcher, said the startup is trying to train its model in a different way that involves more humans and AIs interacting and collaborating together. The model will be trained using long-horizon and multi-agent reinforcement learning.
Long-horizon reinforcement learning trains the model to plan, act, revise, and follow through over time, rather than just generate a one-off answer. Multi-agent reinforcement learning trains for environments where multiple AIs or humans are involved. Both concepts are gaining momentum in academic research as scientists push large language models beyond chatbot responses toward systems that can coordinate actions over many steps.
He also emphasized that the model needs a strong memory about itself and the user to enable better understanding.
Despite the stellar team, there are significant risks ahead. Humans& will need vast sums of cash to fund the expensive process of training and scaling a new model, putting it in competition with major established players for resources like computing power.
The top risk is that Humans& is not just competing with collaboration tool companies. It is challenging the leading AI companies themselves, which are actively working on better ways to enable human collaboration on their own platforms. Anthropic’s Claude Cowork aims to optimize work-style collaboration, Gemini is embedded into Google Workspace, and OpenAI has been pitching developers on multi-agent orchestration.
Crucially, none of the major players appear poised to rewrite a model based on social intelligence, which could give Humans& an advantage or make it an acquisition target. With companies like Meta, OpenAI, and DeepMind seeking top AI talent, mergers and acquisitions are a real possibility.
Humans& told reporters it has already turned away interested parties and is not interested in being acquired. Zelikman expressed the belief that this will be a generational company with the potential to fundamentally change how we interact with AI models, and that the team has faith in its ability to execute that vision.