Claude Haiku 5.5 Puts Low-Cost AI Agents in Reach of More Routine Tasks

TL;DR: Anthropic says Claude Haiku 5.5 is faster and materially cheaper than the previous Haiku…

Anthropic’s Public Benefit Fund Puts a Shareholding Behind Its Mission

TL;DR: Anthropic has outlined a share-funded public benefit entity whose stake could increase as the…

Microsoft Magnetar Brings Verifiable Computation Closer to Ordinary Code

TL;DR: Magnetar is a Microsoft Research system designed to prove program execution efficiently using conventional…

Google’s Enterprise Gemini Agent Tries to Make One Prompt Control Several Workflows

TL;DR: Google Cloud introduced a Gemini agent that can use business context, tools and models…

Visa’s Live Agent-Payment Pilot Makes Authorisation the Centre of AI Commerce

TL;DR: Visa, BILL and Billtrust have launched a controlled pilot that lets authenticated AI agents…

OpenAI’s Orchestra Bio Acquisition Targets the Work Between Discovery and Medicine

TL;DR: OpenAI has acquired Orchestra Bio to combine its drug-development workflow platform with Rosalind life-sciences…

AI Is Getting Better at COBOL, Raising the Stakes for Legacy Financial Systems

TL;DR: OpenAI published new tests showing improved model performance in less-common programming languages and code…

Thousands of AI Research Agents Analyse Millions of Brain Cells in 32 Minutes

TL;DR: OpenAI, PerturbAI and Scripps Research used parallel AI agents to generate testable hypotheses from…

OpenAI’s Training Containment Report Reveals a New Security Problem for AI Agents

TL;DR: OpenAI says it hardened frontier-model training environments after a contained September incident and extensive…

Cognitive Credit’s $40 Million Raise Makes Data Provenance an AI Product Feature

An AI-generated credit analysis is only useful if an analyst can establish where its numbers…