Humans& Raises $480M to Build AI Models Designed for Coordination, Not Just Chat

Humans& Secures $480 Million Seed Funding for New AI Vision

Humans&, an AI startup barely three months old, has closed a $480 million seed funding round at a $4.48 billion valuation, marking one of the largest seed raises in startup history 1

1

2

2

. The round was led by SV Angel and co-founder Georges Harik, Google's seventh employee who helped build its first advertising systems 2

2

. Major investors include Nvidia, Jeff Bezos, GV (formerly Google Ventures), and Emerson Collective, signaling strong confidence in the company's vision for human-centric AI 3

3

4

4

.

The founding team brings formidable credentials from the AI industry's leading labs. Co-founders include Andi Peng, a former Anthropic researcher who worked on reinforcement learning and post-training of Claude 3.5 through 4.5; Eric Zelikman and Yuchen He, two former xAI researchers who helped develop the Grok chatbot; and Noah Goodman, a Stanford professor of psychology and computer science 2

2

. The startup's 20-person team also draws talent from OpenAI, Meta, Google DeepMind, and MIT 2

2

5

5

.

Building Foundation Models for Social Intelligence

Humans& positions itself at what it calls the next major frontier for foundation models: AI collaboration designed for coordination rather than isolated assistance 1

1

. Current AI chatbots excel at answering questions and summarizing documents, but they function as helpful assistants for one user at a time. They lack the capability to manage the complex work of real collaboration—coordinating people with competing priorities, tracking long-running decisions, and keeping teams aligned over time 1

1

.

"It feels like we're ending the first paradigm of scaling, where question-answering models were trained to be very smart at particular verticals, and now we're entering what we believe to be the second wave of adoption where the average consumer or user is trying to figure out what to do with all these things," Peng told TechCrunch 1

1

.

The company aims to build what it describes as a "central nervous system" for the human-plus-AI economy, developing new foundation model architecture designed specifically for social intelligence for AI 1

1

. This represents a departure from models optimized primarily for information retrieval or code generation.

Rethinking How AI Interacts With Teams

CEO Eric Zelikman explained that the new model will be trained to ask questions in a way that feels like interacting with a friend or colleague, someone genuinely trying to understand you 1

1

. Current chatbots ask questions constantly but without understanding the value of those questions, he noted, because they've been optimized for two narrow metrics: how much a user immediately likes a response and how likely the model is to answer correctly 1

1

.

The company envisions its product as a potential replacement for multiplayer contexts like communication platforms such as Slack or collaboration platforms like Google Docs and Notion 1

1

. Rather than building a model that plugs into existing tools, Humans& wants to own the collaboration layer entirely 1

1

.

Technical Approach: Multi-Agent Reinforcement Learning

To achieve its vision for AI for collaboration, Humans& plans to train its algorithms using reinforcement learning, a training approach commonly used to develop reasoning models 5

5

. The company has identified the need for innovations in "long-horizon and multi-agent reinforcement learning, memory, and user understanding" 2

2

.

Long-horizon processing refers to complex tasks that take large language models hours or more to complete, representing a major focus area for machine learning researchers 5

5

. The startup also plans to equip its software with support for multi-agent systems, meaning its neural networks will be capable of collaborating with other AI models on multistep tasks 5

5

.

Peng told Crunchbase News that the majority of the capital will be spent on compute for training models 4

4

. The frontier AI lab will work with Nvidia on hardware and software to support this intensive computational work 3

3

.

Market Context and Growing Competition

The Humans& raise reflects a broader trend of massive seed funding flowing to AI startups founded by researchers from major labs. Crunchbase data shows that more than 41% of the $38.4 billion invested in global seed funding in 2025 went to companies in AI-focused industry categories, up from 30% in 2024 4

4

. Just over $15 billion went to AI-focused seed rounds, up about 50% from 2024 4

4

.

The space for AI to enhance worker productivity through coordination is becoming increasingly competitive. The AI note-taking app Granola recently raised a $43 million round at a $250 million valuation as it launched more collaborative features 1

1

. LinkedIn founder Reid Hoffman has argued that companies are implementing AI wrong by treating it like isolated pilots, and that the real leverage is in the coordination layer of work—how teams share knowledge and run meetings 1

1

.

What Comes Next for the AI Startup

Humans& still doesn't have a product, and details remain scarce about what exactly it will be 1

1

. Peng explained that the company is designing 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 co-evolve the interface and the behaviors that the model is capable of into a product that makes sense" 1

1

.

The Times reported that the company plans to launch its first product early this year 5

5

. The team has hinted at both enterprise and consumer applications, though specific use cases remain undefined 1

1

.

The company's success will depend on whether it can deliver on its promise to rethink how foundation models are trained and how people interact with AI systems. As companies transition from chat to agents, the coordination challenge remains largely unaddressed, creating an opening for Humans& to establish itself as the standard for how AI supports human connection 1

1

4

4

.