Reader-focused coverage of AI models, APIs, infrastructure, developer tools and changes that matter in practice. Client service announcements remain inside the authenticated account area.
Anthropic has retired Claude Opus 4.1 from its API and related Anthropic-operated platforms, meaning requests to the old model ID now fail unless teams migrate routing to Claude Opus 4.8 or another supported model.
Read news →Amazon Bedrock Agents is now Bedrock Agents Classic and is closed to new customers as of July 30, 2026. Existing users can continue, but new agent infrastructure is being steered toward Bedrock AgentCore.
Read news →Article 50 transparency obligations under the EU AI Act started applying on August 2, requiring covered AI systems to make certain AI interactions and AI-generated or manipulated content recognizable to people.
Read news →GitHub Models has been fully retired, removing its catalog, playground, inference API and BYOK endpoints. Teams that used it as a lightweight model gateway now need a replacement plan.
Read news →OpenAI has sharply reduced GPT-5.6 Luna and Terra API prices and introduced Sol Fast mode, changing cost and latency calculations for teams routing production AI workloads.
Read news →Microsoft has introduced MAI-Cyber-1-Flash, a cybersecurity model integrated into its MDASH defender workflow, with access limited to verified defenders and a design aimed at handling most security-agent tasks without always invoking larger models.
Read news →GitHub has made agent skills and MCP server context generally available in Copilot code review, giving teams a supported way to inject repository rules and read-only external context into AI-assisted reviews.
Read news →The Model Context Protocol is moving to a stateless core in a breaking 2026-07-28 revision, a shift that should make remote MCP deployments easier to scale but will require compatibility work across agent tooling.
Read news →