AI chatbots are getting better at answering questions, summarizing documents, and solving math equations, but they still largely behave as helpful assistants for one user at a time. They’re not designed to handle the more complex work of true collaboration: coordinating people with competing priorities, tracking long-term decisions, and keeping teams aligned over time.
Humans&, a new startup founded by alumni of Anthropic, Meta, OpenAI, xAI and Google DeepMind, believes that bridging this gap is the next major frontier for foundation models. The company raised this week a round of funding of 480 million dollars building a “central nervous system” for the human plus AI economy. The startup “AI to empower humans“The framing has dominated early media coverage, but the company’s real ambition is newer: to build a new core model architecture designed for social intelligence, not just information retrieval or code generation.
“It’s like we’re ending the first scaling paradigm, where question-answering models were trained to be very intelligent in particular verticals, and now we’re moving into what we think is the second wave of adoption where the average consumer or user is trying to figure out what to do with all these things,” Andi Peng, one of the co-founders of Humans& and a former Anthropic employee, told TechCrunch.
Humans’ talk aims to help people enter the new era of AI, moving beyond the talk that AI will take their jobs. Whether it’s just marketing talk or not, the timing is crucial: businesses are moving from chat to agents. The models are competent, but the workflows are not, and the coordination challenge remains largely ignored. And despite all this, people feel threatened and overwhelmed by AI.
The company, created three months ago, like several of its peers, managed to germinate its surprising seed thanks to this philosophy and the pedigree of its founding team. Humans& still doesn’t have a product, nor is it clear what exactly it might be, although the team has said it could replace multiplayer or multi-user contexts like communication platforms (think Soft) or collaboration platforms (think Google Docs and Notion). When it comes to use cases and target audience, the team hinted at enterprise and consumer applications.
“We’re building a product and model that is centered around communication and collaboration,” Eric Zelikman, co-founder and CEO of Humans& and former xAI researcher, told TechCrunch, adding that the product’s focus is on helping people work together and communicate more effectively – both with each other and with AI tools.
“For example, when you have to make a decision as a large group, it often comes down to someone bringing everyone together in one room, for everyone to express their different camps, for example on what kind of logo they would like,” Zelikman continued, giggling with his team as they recalled the tedium that took a long time to get everyone to agree on a logo for the startup.
Techcrunch event
San Francisco | October 13-15, 2026
Zelikman added that the new model will be trained to ask questions in a way that feels like you’re interacting with a friend or colleague, someone who is trying to get to know you. Today’s chatbots are programmed to constantly ask questions, but they do so without understanding the value of the question. He says this is because they have been optimized for two things: how much a user immediately likes an answer given to them and how likely it is that the model will correctly answer the question they receive.
Part of the lack of clarity on what the product is could be because humans don’t exactly have an answer to that question yet. Peng said Humans designs the product in conjunction with the model.
“Part of what we’re doing here is also making sure that as the model improves, we’re able to evolve the interface and the behaviors that the model is capable of into a product that makes sense,” she said.
What is clear, however, is that humans are not trying to create a new model that can plug into existing applications and collaboration tools. The startup wants to own the collaboration layer.
AI and team collaboration and productivity tools are an increasingly hot area – for example, there has been growing interest from AI note-taking startup Granola. $43 million funding round at a valuation of $250 million as it launched more collaborative features. Several high-profile voices also explicitly define the next phase of AI as one of coordination and collaboration, not just one of automation. LinkedIn founder Reid Hoffman argued today that companies are misimplementing AI by treating it like isolated drivers and that the real leverage lies in the work coordination layer — how teams share knowledge and hold meetings.
“AI lives at the workflow level, and the people closest to the work know where the friction really is,” Hoffman wrote on social networks. “They are the ones who will discover what should be automated, compressed or completely redesigned.”
This is the space where humans want to live. The idea is that its model product would act as the “connective tissue” within any organization – whether a company of 10,000 people or a family – that understands the skills, motivations and needs of each person, as well as how it can all be balanced for the good of the whole.
To achieve this, we need to rethink how AI models are trained.
“We are trying to train the model in a different way that will involve more humans and AI interacting and collaborating together,” Yuchen He, co-founder of Humans& and former OpenAI researcher, told TechCrunch, adding that the startup’s model will also be trained using long-term, multi-agent reinforcement learning (RL).
Long-horizon RL aims to train the model to plan, act, review, and track over time, rather than just generating a good one-time response. Multi-agent RL trains for environments where multiple AIs and/or humans are in the loop. These two concepts are gaining momentum recent academic work as researchers push LLMs beyond chatbot responses to systems capable of coordinating actions and optimizing outcomes across many stages.
“The model has to remember things about itself, about you, and the better its memory, the better its understanding by the user,” he said.
Despite the excellent team running the show, there are many risks ahead. Humans will need endless amounts of money to finance the expensive effort of training and scaling a new model. This means it will compete with established major players for resources, including access to compute.
The main risk, however, is that humans don’t just compete with the world’s Notions and Slacks. This is happening for the Top Dogs of AI. And these companies are actively working on better ways to enable human collaboration on their platforms, even as they swear that AGI will soon replace economically viable labor. Through Claude Cowork, Anthropic aims to optimize collaboration at work; Gemini is integrated into Workspace, so AI-driven collaboration is already happening in the tools people already use; and OpenAI recently introduced developers to its multi-agent orchestration and workflows.
Crucially, none of the major players seem ready to rewrite a model based on social intelligence, which either gives humans an advantage or makes it a target for acquisition. And with companies like Meta, OpenAI, and DeepMind seeking top AI talent, mergers and acquisitions are certainly a risk.
Humans& told TechCrunch that it had already turned away interested parties and was not interested in an acquisition.
“We think this will be a generational company, and we think it has the potential to fundamentally change the future of how we interact with these models,” Zelikman said. “We have confidence in ourselves to achieve this and we have great confidence in the team we have assembled here.”
This article was originally published on January 22, 2026.
