Although AI chatbots are getting better at answering questions, summarizing documents, and solving mathematical equations, they still act like helpful assistants for one user at a time. They aren’t designed to manage the heavy lifting of real collaboration, like coordinating people with competing priorities, tracking decisions over time, and keeping teams aligned over time.
Humans&, a new startup founded by alumni from Anthropic, Meta, OpenAI, xAI, and Google DeepMind, believes bridging that gap is the next major frontier for foundational models. This week, the company raised $480 million in a seed round to build a “central nervous system” for the human-AI economy. While the startup’s “AI that empowers humans” framework has dominated the early press, the company’s actual ambition is much more radical: building a new underlying model architecture designed for social intelligence, not just information retrieval or code generation.
“It feels like the first paradigm of scaling, where question-answering models were trained to be very smart in certain industries, is ending, and now we’re entering a second wave of adoption where the average consumer or user is trying to figure out what to do with all of this,” Andi Peng, one of Human&’s co-founders and a former Anthropic employee, told TechCrunch.
Humans&’s pitch goes beyond the narrative of AI taking people’s jobs and focuses on helping bring people into a new era of AI. Whether it’s just a marketing story or not, timing is critical as companies are moving from chat to agents. Although the model is competent, the workflow is not, and the coordination challenges remain largely unsolved. And through it all, people feel threatened and overwhelmed by AI.
The three-month-old company, like many of its peers, was able to grow an impressive seed round on the back of this philosophy and the pedigree of its founding team. Humans& doesn’t have a product yet, and it’s not clear what exactly it is, but the team says it could be an alternative for multiplayer or multiuser contexts, such as communication platforms (think Slack) or collaboration platforms (think Google Docs or Notion). Regarding use cases and target users, the team suggested both enterprise and consumer applications.
“We’re building products and models that are centered around communication and collaboration,” Eric Zelikman, co-founder and CEO of human& and former xAI researcher, told TechCrunch, adding that the company’s focus is on developing products that help people collaborate and communicate more effectively with each other and with AI tools.
“Similar to when you have to make a decision in a large group, someone often gets everyone in a room and asks them to express their opinions on, say, what kind of logo they want,” Zelikman continues, chatting with his team as he recalls the time-consuming and tedious process of getting everyone to agree on a logo for a startup.
tech crunch event
san francisco
|
October 13-15, 2026
Zelikman added that the new model will be trained to ask questions that feel like they’re interacting with a friend or colleague, someone who is trying to get to know you. Current chatbots are programmed to constantly ask questions, but they do so without understanding the value of the questions. According to him, this is because it is optimized for two things. two are how quickly the user likes a given response, and two is how likely the model is to correctly answer the question it receives.
One reason why it’s not clear what the product is may be that humans don’t yet have an exact answer to it. Peng said humans and models work together to design products.
“Part of what we’re doing here is also making sure that as the model improves, we can co-evolve the interface and the behaviors that the model can perform into a meaningful product,” she said.
What is clear, however, is that humans are not trying to create new models that can connect to existing applications and collaboration tools. Startups want to own the collaboration layer.
AI and team collaboration and productivity tools are areas that are receiving increasing attention. For example, startup AI note-taking app Granola raised $43 million at a $250 million valuation to release more collaboration features. Several prominent voices are also clearly framing the next stage of AI as one of coordination and collaboration, rather than just automation. LinkedIn founder Reid Hoffman today argued that companies are wrong to adopt AI by treating it like a lone pilot, and that the real use lies in the coordination layer of work: how teams share knowledge and run meetings.
“AI lives at the workflow level, and the people closest to the work know where the friction actually lies,” Hoffman wrote on social media. “They discover what should be automated, compressed, or completely redesigned.”
It is a space where humans want to live. The idea is that the company’s model-slash-products act as “connective tissue” across any organization, whether it’s a 10,000-employee company or a family, understanding each person’s skills, motivations, needs, and how to balance it all for the benefit of the whole.
To get there, we need to rethink how we train AI models.
“We’re trying to train models in a different way, where more humans and AI interact and collaborate,” Human& co-founder and former OpenAI researcher Yuchen He told TechCrunch, adding that the startup’s models are also trained using long-term, multi-agent reinforcement learning (RL).
Long-Horizon RL aims to train models to plan, act, correct, and follow through over time, rather than producing good answers once and for all. Multi-agent RL trains for environments where multiple AIs or humans are in the loop. Both of these concepts have gained momentum in recent academic research as researchers push LLMs beyond chatbot responses and toward systems that can coordinate actions and optimize outcomes across many steps.
“The model has to remember things about itself and you, and the better it remembers, the better it understands the user,” he said.
Despite having a talented staff running the show, there are many risks ahead. Humans require endless amounts of cash to fund the expensive efforts of training and expanding new models. This means competing with large, established players for resources, including access to computing.
But the biggest risk is that humans aren’t just competing with concepts and slacks around the world. The AI top dog is coming. And these companies are actively working on better ways to enable human collaboration on their platforms, even as they assert that economically viable jobs will soon be replaced by AGI. Through Claude Cowork, Anthropic aims to optimize collaboration in the way we work. Gemini is built into Workspace, so AI-powered collaboration is already happening within the tools people already use. OpenAI has recently been pitching developers on multi-agent orchestration and workflows.
Importantly, no major company seems ready to rewrite models based on social intelligence that give humans an advantage or make them targets for acquisition. And with companies like Meta, OpenAI, and DeepMind on the prowl for top AI talent, M&A is certainly a risk.
Humans& told TechCrunch that it has already turned down interested parties and has no intention of being acquired.
“We believe this is going to be a multi-generational company, and we think this has the potential to fundamentally change the future of how we interact with these models,” Zelikman said. “I believe we can do it and I have a lot of confidence in the team we have assembled here.”
This post was originally published on January 22, 2026.
