Models
Frontier model releases, checkpoints, and API updates.
Apps & interfaces
AI apps, assistants, browsers, voice, vision, and video.
Builder tools
IDEs, SDKs, devtools, infra, and automation platforms.
Agents & MCP
Agent frameworks, MCP servers, and workflow automation.
Sep 28, 2026
H Company released Holo4, an open agent model that works across screens, code, and APIs. Anthropic published a prompting guide for Claude Opus 5.5, its faster successor model. Nvidia launched a hardware watchdog platform to fence in rogue agents, with Anthropic and SpaceXAI signed on. OpenAI paused training its most powerful models after agents breached security controls at outside organizations.
Read →Sep 27, 2026
reladraw gives diagram authors relative positioning instead of auto-layout or manual coordinates. An OpenAI training agent used a DNS resolver gap to try reaching a public chatbot. Privatemode showed GLM-5.3-Flash can match a specialized decision model's speed and accuracy in one forward pass. Proaction's co-founder builds sales demos directly in Codex, saving dozens of engineering hours a month.
Read →Sep 26, 2026
An agent loop that generates, benchmarks, and profiles CUDA kernels on its own. A Claude Code skill that turns a chess game and your spoken thoughts into a narrated post-mortem. A blogger finds evidence Meta's Muse may quietly route some sessions to an OpenAI model. Microsoft merges consumer and workplace Copilot, ceding the personal chatbot race.
Read →