humans&

humans& is a frontier artificial intelligence lab established in September 2025 with a mission to develop AI that enhances human connection and collaboration. The company was co-founded by a team of prominent researchers and engineers from leading AI institutions, including Georges Harik, an early Google employee; Andi Peng, formerly of Anthropic; Eric Zelikman and Yuchen He from xAI; and Noah Goodman, a Stanford University professor. This founding group possesses extensive experience from major labs like Google DeepMind, OpenAI, and Meta, having contributed to significant projects in AI reasoning, behavioral training, and agent development.

The company's formation is rooted in the belief that the next evolution of AI should not focus on replacing human labor but on augmenting it. Co-founder Andi Peng expressed unease with the industry's push towards fully autonomous systems, advocating instead for a model where machines and humans are complementary. Consequently, humans& is dedicated to building AI tools that facilitate teamwork and strengthen how people work together. The core of their strategy involves pioneering new training methods for AI models, with a focus on long-horizon reinforcement learning, multi-agent systems, memory, and a sophisticated understanding of user context.

Although a specific commercial product has not yet been released, the company envisions a tool akin to an AI-enhanced instant messaging platform. This system would be embedded within group conversations, assisting with tasks like research while also being able to ask clarifying questions and retain information over time—a departure from current chatbots optimized primarily for answering prompts. The substantial computational power required for this new training paradigm is a key driver of the company's financial strategy.

Keywords: human-centric AI, collaborative intelligence, frontier AI lab, multi-agent reinforcement learning, AI-enhanced communication, human-machine collaboration, AI reasoning, AI behavioral training, AI agents, long-horizon learning, user understanding AI