After years of watching smart teams mistake sampling for safety, I no longer ask how many AI tests we ran, only which failures we have made impossible by design.
New research looks at how leading AI models hold up doing actual white-collar work tasks, drawn from consulting, investment banking, and law. Most models failed.
Humans&, a new startup founded by alumni of Anthropic, Meta, OpenAI, xAI, and Google DeepMind, is building the next generation of foundation models for collaboration, not chat.
The move comes as Substack has been investing more heavily in video and livestreaming, as it looks to compete with platforms like YouTube and Patreon for both creators and viewers.