Humans&: OpenAI and DeepMind Veterans Bury the Chatbot Era

It seems we've reached a point of saturation. Every day a new language model comes out that writes poetry or Python code half a percent better than the previous one. But let's be honest: the interface "a human types in a line — a machine responds with text" is already starting to look archaic for complex business tasks. It was precisely on this sense of impasse that a group of engineers played who decided we've had enough of chatbots.

Enter the startup Humans& (yes, the name with an ampersand, hinting at the connection). And this isn't just another garage project. The roster of founders reads like a credits list for a blockbuster AI movie: here are people who went through the school of Anthropic, Meta, OpenAI, xAI and Google DeepMind. When specialists of this caliber are massively leaving their comfortable positions in Big Tech for one idea, it's worth paying attention. Usually it means they've sensed something that unwieldy corporations are either overlooking or unable to implement due to bureaucracy.

The main idea of Humans& — a paradigm shift from "chat" to "coordination". Right now large language models (LLMs) are brilliant loners. They can write an essay, but try to make three different models agree on planning a vacation or a complex logistics project without your constant intervention — and everything falls apart. Modern AI doesn't know how to collaborate effectively either with people or with other AI. We're still forced to be babysitters for neural networks, constantly correcting their direction of movement.

Humans& is building a new generation of foundational models tailored specifically for collaboration. This is a critically important step toward what the industry calls the "agentic future". Imagine a system where you don't write a prompt, but set a goal, and AI itself finds the necessary tools, connects with other agents, clarifies details with human colleagues and delivers the result. This is no longer about text generation, it's about understanding the context of collaborative work and the implicit social contracts that underpin any workflow.

Why is this important right now? Because the "one giant chatbot for everything" model is showing its limits. We're seeing a trend toward specialization and fragmentation. Ilya Sutskever left to create safe superintelligence, others are moving into robotics, and Humans& have claimed the niche of the "glue" that should connect disparate intellectual systems into a working mechanism. If current LLMs are a brain in a jar, then Humans& are trying to create a nervous system that allows these brains to work together.

The bottom line: The era of solitary chatbots is ending. The next big battle will not be fought over text quality, but over AI's ability to embed itself in complex human processes without friction. Humans& are betting that the next trillion dollars will be earned not by whoever makes the smartest chat, but by whoever makes that chat actually work.