Humans&, a new frontier AI lab, was founded by alumni from Anthropic, OpenAI, Meta, xAI, and Google DeepMind. The startup emphasizes that AI should empower people, not replace them.

While AI models are rapidly improving in reasoning, coding, and autonomous actions, humans& focuses on human progress through understanding, trust, and collaboration.

The company believes the next chapter of AI should center on people and their interactions.

Building AI for Coordination and Social Intelligence

Unlike existing chatbots that serve individuals, humans& aims to solve the coordination challenges of teams.

Current AI tools are not designed to track long-term decisions, manage competing priorities, or keep groups aligned over time.

Humans& wants to build a model that functions as the “connective tissue” for organizations, helping people communicate and collaborate efficiently.

CEO Eric Zelikman explained the focus is on large-group decisions, using the example of agreeing on a startup logo as a time-consuming but common challenge.

A Foundation Model Designed for Human Interaction

Humans& is developing a new foundation model that prioritizes social intelligence over information retrieval or code generation.

The AI will ask questions in ways that feel natural, similar to a colleague or friend trying to understand you.

Co-founder Andi Peng highlighted that humans& is designing the product in tandem with the model so improvements in AI capabilities can evolve the interface and behaviors.

The company does not plan to integrate into existing collaboration tools, aiming instead to own the collaboration layer itself.

Advanced Training Techniques for Real-World Collaboration

To enable meaningful coordination, humans& is employing long-horizon and multi-agent reinforcement learning (RL).

Long-horizon RL allows the model to plan, act, revise, and follow through over extended periods. Multi-agent RL trains the AI to operate in environments with multiple humans and AIs interacting simultaneously.

Yuchen He emphasized that a model with strong memory and understanding of users can better anticipate needs and support collaboration over time.

Founding Team and Investor Support

Humans&’s founding team includes experts who have led major AI projects across industry and academia, including xAI, Anthropic, Google DeepMind, OpenAI, Meta, Reflection, AI2, Stanford, and MIT.

Key members include Alexis Ross, Andi Peng, Ani Nrusimha, Charlie George, Diyi Yang, Eric Zelikman, Georges Harik, Jeremy Berman, Niloofar Mireshghallah, Noah D. Goodman, Ray Ramadorai, Rob Li, Saurabh Shah, Taylor Sorensen, Varuna Jayasiri, Weisi Duan, Yuchen He, and Ziang Li.

The startup raised a $480 million seed round led by SV Angel and co-founder Georges Harik. Other investors include Nvidia, Jeff Bezos, GV (Google Ventures), Emerson Collective, Forerunner, S32, DCVC, Human Capital, Liquid 2, Felicis, CRV, and several others. Despite the large funding, humans& has made clear it is not seeking acquisition and intends to maintain independence to pursue a generational vision.

Humans& in the Competitive AI Landscape

Humans& is entering a market where AI-enabled collaboration is gaining attention. Competitors like Anthropic, OpenAI, and Google are embedding AI into existing productivity tools.

Startups like Granola are adding collaborative AI features to note-taking and workflow apps. Humans& differentiates itself by aiming to create a new model architecture focused on social intelligence and team coordination rather than incremental improvements to existing tools.

PHOTO: HUMAN&

This article is a summary of two original articles. The full versions can be read at the following links:

https://humansand.ai/

https://techcrunch.com/2026/01/25/humans-thinks-coordination-is-the-next-frontier-for-ai-and-theyre-building-a-model-to-prove-it/

https://techcrunch.com/2026/01/20/humans-a-human-centric-ai-startup-founded-by-anthropic-xai-google-alums-raised-480m-seed-round/

This article was created with AI assistance.

We make every effort to ensure the accuracy of our content, some information may be incorrect or outdated. Please let us know of any corrections at [email protected].