News archive.
Earlier AI news and analysis, kept for reference. Current coverage lives in News & Analysis.
DFlash Lifts LLM Throughput Up to 15x on NVIDIA Blackwell
Block-diffusion speculative decoding now available on vLLM and SGLang. NVIDIA released 20 checkpoints and recipes; swapping from EAGLE-3 requires config-only changes.
Power costs 40% of AI factory OpEx. Here's how to cut it.
Power efficiency is now the primary constraint on AI factory profitability. NVIDIA and researchers outline full-stack optimizations that can cut energy per token without sacrificing speed or scale.
Meta's 7mm Steel-Can Batteries Double Ray-Ban Runtime Without Chemistry Changes
Meta engineered ultra-narrow steel-can batteries for AI glasses by replacing wound electrodes with die-cut stacks, cutting impedance and enabling dual-battery systems. Here's how the form factor constraint became a power advantage.
Transformers.js Tests Cross-Origin Storage to Cut AI Model Duplication
Google and Hugging Face are experimenting with a new browser API that lets web apps share cached AI models and WebAssembly files across different websites. Early tests show 177 MB savings per duplicate download.
IBM Open-Sources CUGA: Build Agents in One File, Two Dozen Examples Included
CUGA is a lightweight agent harness that handles orchestration, tool binding, and state management so you write only a tool list and prompt. IBM shipped 24 working single-file apps to prove the pattern works.
Anthropic launches Claude Tags for enterprise team workflows
Anthropic rolled out Claude Tag, a collaborative tool designed for multi-team agentic workflows. Details on what it does and who benefits.
Broadcom launches $35B AI platform with Anthropic partnership
Broadcom announced a $35 billion AI infrastructure platform integrating Anthropic's Claude models. The move signals a major chipmaker bet on enterprise AI deployment.
Anthropic replaces Slack bot with Claude Tag, a persistent AI teammate
Anthropic is retiring its Slack app in favor of Claude Tag, a new persistent AI agent that learns from your workspace, monitors tasks, and works autonomously. Here's what changes for teams already using Claude in Slack.
OpenAI funds open source maintainers through Daybreak initiative
OpenAI launches Patch the Planet, a Daybreak-backed program to provide financial support and resources to open source maintainers. Details on eligibility and how to apply.
Anthropic sues US government over model export restrictions
An AI startup has filed suit against the federal government claiming it blocked deployment of Anthropic's latest model. The legal challenge raises questions about how US export controls apply to frontier AI systems.
Anthropic Ships Claude Tags for App Builders
Anthropic released Claude Tag, a feature letting developers organize and route requests within Claude API calls. Details on how it works and which models support it.
OpenAI joins effort to build shared AI safety standards
OpenAI is supporting evaluation frameworks and safety practices through the Appia Foundation. Learn what shared standards mean for your AI deployments.