OpenCode v1.18.30 Released — Adds System Prompts for GPT-6 and Fixes Various Providers
OpenCode has released v1.18.30, featuring the addition of Astra system prompts for GPT-6 models and fixes for providers including Bedrock, OpenAI, and Azure.
Latest
Inception Announces Diffusion-Based Mercury 2.5, Enabling High-Speed Generation at 1,107 Tokens Per Second
Inception has unveiled Mercury 2.5, a new model employing a diffusion model that generates and refines multiple tokens in parallel, enabling faster output compared to traditional autoregressive LLMs.
Anthropic Researcher Leaves Company, Criticizing AI Development Race
Jacob Coxon, a researcher at Anthropic, has announced his departure from the company, expressing strong concerns that the AI development race could lead to an uncontrollable situation.
Risks of Cloud AI Service Suspensions Driving Interest in Local LLMs
Driven by the evolution of open-weight models and the risk of losing access to cloud AI services due to sanctions or terms of service, interest in Local LLMs that run on local environments is growing.
US Authorities Warn Against Large-Scale Knowledge distillation by Chinese AI Companies
The US National Security Agency (NSA), FBI, and other agencies have issued a joint warning stating that Chinese AI companies are conducting distillation on an industrial scale to extract knowledge from cutting-edge American models.
Google DeepMind Announces AlphaGenome Atlas, a Prediction Database for Human Genome Variants
Google DeepMind has released AlphaGenome Atlas, a large-scale database that predicts the effects of approximately 9 billion single nucleotide variants in the human genome on gene expression and other biological processes.
OpenAI Announces AI Solution to Millennium Prize Problem Navier-Stokes Equations
On September 8, OpenAI announced that an unreleased internal model generated a solution to the Existence and Smoothness of the Navier-Stokes equations, a critical challenge in fluid dynamics. Meanwhile, mathematicians have raised concerns regarding the circumstances surrounding the achievement.
Claude Code v2.1.266 Released — Fixes Sign-in Failures Caused by Specific Environment Variable Settings
Anthropic has released the v2.1.266 of its CLI tool, Claude Code. This update fixes a bug identified in v2.1.265 where setting the CLAUDE_CODE_USE_GATEWAY environment variable forced sign-in to the Cloud gateway, resulting in failed requests.
Google DeepMind Releases AlphaGenome Atlas to Predict Effects of All Single-Base Mutations in the Human Genome
Google DeepMind has released the AlphaGenome Atlas, a database predicting the molecular-level effects of approximately 9 billion single-base mutations in the human genome, provided free of charge for academic research via a website.
Arm Announces Three New Products Including CSS for Mobile 2
On September 8, 2026, Arm announced CSS for Mobile 2, optimized for agentic AI and advanced graphics, Arm Neoverse CSS N4 for AI infrastructure, and Arm Total Design for Physical AI to accelerate the deployment of physical AI.
New "SAFE" Guidelines Proposed to Strengthen Agentic AI Security with Collaboration from NVIDIA and Others
The Linux Foundation has proposed "SAFE," a set of guidelines designed to share security incidents of agentic AI across the ecosystem to improve protection. Participating companies, including NVIDIA and Cisco, are contributing multi-layered defense technologies such as execution environments and identity control.
SAP Transitions Certification Exams to Practical Format, Allowing AI Tool Use in Some Tests
SAP has announced that it is allowing the use of AI tools in some of its certification exams and has transitioned from traditional multiple-choice questions to practical assessments involving actual system operations.
Fhill Launches Recruitment AIO Support Service to Help Companies Optimize Their AI-Driven Evaluations
On September 8, 2026, Fhill announced the launch of its Recruitment AIO support service, designed to ensure that generative AI and AI search services correctly understand company information and recommend them in their responses.
Meta Announces Personal AI Agent "Muse" Featuring Task Automation and Advanced Security
Meta announced "Muse" on September 8, a personal AI agent capable of performing tasks such as sending emails and booking travel, utilizing a robust security design based on dedicated virtual machines in the cloud.
gemini-cli v0.59.0 Released — Security Fixes and Enhanced Workspace Trust
Gemini CLI has released v0.59.0, featuring fixes including the prevention of SSRF during MCP OAuth metadata discovery and the enforcement of workspace trust in restricted mode.
Microsoft Records Highest Monthly Vulnerability Fix Count Ever at 972
Microsoft has patched approximately 972 vulnerabilities in its September release, including 112 critical vulnerabilities.
Terence Tao Warns of Risk That AI May Consume Unsolved Mathematical Problems
Mathematician Terence Tao expressed concern that the evolution of AI could erode the value of unsolved problems in mathematics, depleting the resources necessary for new exploration.
OpenAI Derives Solution to Navier-Stokes Equations Using AI Agents
OpenAI has announced that it derived a solution to the Navier-Stokes equations, one of the Millennium Prize Problems in mathematics that has remained unsolved for approximately 90 years, using internal AI models and about 10,000 collaborative agents.
Meta Reveals Details of Personal AI Agent "Muse," Enabling App Integration and Automated Task Execution
Meta has announced "Muse," a personal AI agent that performs web browsing and multi-step tasks on behalf of users. It features integration with existing apps, background operation, and security protection features.
Meta Announces Personal AI Agent Muse
Meta has unveiled Muse, a personal AI agent designed to autonomously execute everyday tasks. It is planned for release on platforms including iOS and Android.