World Labs’ Fei-Fei Li on Creating Large World Models
Bloomberg Live / YouTube
At Bloomberg Tech 2026, Fei-Fei Li laid out her case for spatial intelligence as the frontier beyond language AI. World Labs has now raised $1.23 billion and shipped Marble, a product that generates physically faithful 3D environments — simultaneously producing visual output and collision geometry a physics engine can compute on. The taxonomy she argues for covers renderers, simulators, and planners. The human decision layer does not appear in it.
Watch on YouTube
How Human AI Simulation Unlocks Decision Intelligence
Fast Company, June 15, 2026
A global ice cream manufacturer ran an end-to-end product innovation simulation across three markets in under two hours, using AI consumer agents trained on real behavioral data. The article describes a five-layer architecture — people, data, agents, models, workflow — that lets teams experience simulated futures rather than speculate about them. The closest thing in circulation to a practical description of designed decision architecture. Layer 3 in the wild, without the name.
Read on Fast Company
From Language AI to Physical AI: First General World Foundation Model Unveiled in China
CGTN, June 14, 2026
Beijing’s Academy of Artificial Intelligence unveiled Physis-v0.1 at its annual conference, describing it as the world’s first general world foundation model. The goal: AI that understands physical laws, spatial relationships, and cause and effect — not just pattern matching against text. The article notes future demand in robotics, scientific research, simulation, and digital twins. It does not ask who governs the decisions these systems will surface.
Read on CGTN
Why Does Work Slow Down Even When Every Task Is Fully Automated?
UC Today, June 11, 2026
Automation removes task friction. It does not remove decision friction. Approvals, ownership, and prioritization still determine throughput — and most enterprises have no system designed to compress those loops. The article calls this decision latency, documents where it forms, and argues that the next productivity frontier is not faster tasks but faster decisions. A clean, accessible articulation of the Layer 3 problem from outside the AI industry entirely.
Read on UC Today