Meta eyes 20% cuts to fund AI

Google

Google launched Gemini Embedding 2, its first fully multimodal embedding model that unifies text, images, video, audio, and documents into a single semantic space with state-of-the-art performance. This advancement simplifies complex AI pipelines for RAG, semantic search, and multimodal applications across major AI development platforms.

Z.ai

GLM-5-Turbo is a foundation model purpose-built for OpenClaw agent workflows, offering superior tool calling, instruction following, and long-chain task execution. The release includes ZClawBench, a new benchmark for evaluating agent performance, showing substantial improvements over previous models.

Wall Street Journal

OpenAI's plan to allow adult content in its chatbot faces pushback from its own safety advisers over concerns about compulsive use and emotional dependency, yet the company is moving forward with the feature.

X

The world models category is fragmenting into five complementary approaches (JEPA, 3D spatial, learned simulation, infrastructure, active inference), each backed by major funding, that address different aspects of environment understanding rather than directly competing.

TechCrunch

Meta may cut 20% or more of its workforce to offset heavy AI infrastructure spending, joining a wave of tech companies using AI to justify major layoffs that some critics call 'AI-washing' rather than genuine automation-driven cuts.