Andi Peng leaves humans& six months after its $480 million seed

Andi Peng left humans& on July 29th, roughly six months after the frontier AI lab emerged with a $480 million seed round and a $4.48 billion valuation. Peng co-founded humans& with Eric Zelikman, Yuchen He, Georges Harik and Stanford professor Noah Goodman, and her departure removes a co-founder whose work at Anthropic and MIT directly supported the lab's human-centric AI thesis.

Andi Peng @TheAndiPenguin https://x.com/TheAndiPenguin left humans& on July 29th, roughly six months after the frontier AI lab emerged with a $480 million seed round and a $4.48 billion valuation. Peng announced her departure on X https://x.com/theandipenguin/status/2082892263697580226?s=46 on July 30th, saying she was "sad to be leaving humans&" and had sent a farewell note to colleagues the previous day. The message said July 29th would be her last day. Her note framed the split as amicable. Peng thanked the research team, described her colleagues as a family and offered to stay in touch, including by providing permanent tattoos of the company's ampersand. She closed with: "Go humans " Peng co-founded humans& with Eric Zelikman, Yuchen He, Georges Harik and Stanford professor Noah Goodman. She wrote that their early discussions about building a human-centric AI lab brought together questions that had driven her research career, including how AI systems can serve people and support human relationships. Before humans&, Peng was a researcher at Anthropic https://andipeng.com/ working on reinforcement learning and post-training for Claude models from Claude 3.5 through Claude 4.5. She had previously studied at MIT's Computer Science and Artificial Intelligence Laboratory, where her work examined human feedback, aligned preferences and human-machine collaboration. Peng earned bachelor's degrees in cognitive science and global affairs from Yale and spent two years as an AI resident at Microsoft Research. That background made Peng one of the clearest links between humans& and the post-training work behind a leading commercial model family. In January, she told The New York Times https://vi.web-platforms-vi.nyti.nyt.net/2026/01/20/technology/humans-ai-anthropic-xai.html that she viewed machines and people as complementary, drawing a distinction between her goals and the industry push toward systems that can operate autonomously for hours at a time. Investors backed the founding team before a product Humans& introduced itself on January 20th https://humansand.ai/ as a human-centric frontier AI lab focused on communication, collaboration and coordination. Humans& said its research would require advances in long-horizon and multi-agent reinforcement learning, memory and user understanding. The launch came with a $480 million seed financing https://techcrunch.com/2026/01/20/humans-a-human-centric-ai-startup-founded-by-anthropic-xai-google-alums-raised-480m-seed-round/ at a reported $4.48 billion valuation. SV Angel and Harik led the round. Nvidia, Jeff Bezos, GV, Emerson Collective, Forerunner, S32, DCVC, Human Capital, Liquid 2, Felicis and CRV were among the named backers. The size of the round reflected the market value assigned to the founders' records. Zelikman previously worked at xAI on reinforcement learning for Grok and co-authored research on models learning to reason from their own generated rationales. Harik was one of Google's earliest employees and worked on its advertising systems. Goodman studies computational cognition, language and social reasoning at Stanford. He and Peng supplied the academic and human-AI interaction side of humans&'s founding thesis. At the January announcement, TechCrunch reported https://techcrunch.com/2026/01/25/humans-thinks-coordination-is-the-next-frontier-for-ai-and-theyre-building-a-model-to-prove-it/ that humans& had about 20 employees and had yet to release a public product. The founders described possible software that could replace or complement multiplayer tools such as Slack, Google Docs and Notion, with an AI model tracking long-running decisions and helping groups coordinate. That made humans& a team-and-thesis investment from the outset. Peng's departure removes a co-founder whose work at Anthropic and MIT directly supported the lab's argument that models should learn from people over extended interactions. It also comes before humans& has publicly demonstrated whether that research program can produce a collaboration product with a market beyond the investor interest surrounding its founders. Peng's public message did not distance her from the lab's premise. She wrote that the world remains early in its path toward superintelligence and that progress will require smart people working together. Her internal farewell similarly said she remained motivated by the impact humans& set out to pursue, even as she left the founding group tasked with delivering it.