Today in AI: agents get a scoreboard, skills get a neutral home

Oct 23, 2026

BREAKING: Claude Opus 5.5 in Claude Code tops Artificial Analysis Coding Agent Index v2.0 (61.7)

Artificial Analysis shipped v2.0 of its Coding Agent Index on Oct 23, swapping in Cognition's FrontierCode v1.1 and re-grading repo Q&A; Claude Code + Opus 5.5 leads at 61.7, and 20 of 21 configs scored lower than on v1.5. For AI coders: the harness+model pair is what gets ranked now, and cost per task spans from about $0.39 to $11.42, so pick your agent setup by budget, not just by leaderboard rank.

Source: https://x.com/ArtificialAnlys/status/2113424551396934134

Oct 22, 2026

NEW: Anthropic donates Agent Skills to the Agentic AI Foundation

Anthropic handed the Agent Skills standard (the SKILL.md folder format) to the Linux Foundation's Agentic AI Foundation on Oct 22. It now sits alongside MCP, with maintainers from several companies and a roadmap covering versioned releases and security guidance. Anthropic says it has 1M skills published and about 280M installs across 70+ platforms, including Cursor, Copilot and ChatGPT. Skills you write today should keep working across agents.

Source: https://www.anthropic.com/news/donating-agent-skills-to-the-agentic-ai-foundation

Oct 22, 2026

NEW: Claude apps switch to auto mode by default next week

Starting next week, Claude web, desktop and mobile sessions on Pro, Max and Team will run in auto mode, where an automated check blocks risky actions instead of asking you to approve every step. You can switch back to manual at any time. Expect fewer approval prompts during long agent runs. Keep manual mode for anything that touches prod or money.

Source: https://x.com/claudeai/status/2113360845409583307

Oct 21, 2026

BREAKING: OpenAI models and Codex are coming to Google Cloud

OpenAI is bringing its frontier models and Codex agentic workflows to Google Cloud's Model Garden, with private previews underway. Teams on GCP can soon buy OpenAI and Codex on the same invoice as Gemini and Claude. That makes it easier for teams to use more than one vendor and harder for any one vendor to lock you in.

Source: https://x.com/OpenAI/status/2112988118471295323

Oct 23, 2026

TEASED: Kimi K3.1 shows up in Kimi Code, building Three.js 3D scenes

Moonshot AI posted a 24-second retro teaser on Oct 23 that shows "Model: Kimi K3.1" in the Kimi Code terminal while it writes a complex Three.js 3D scene. There is no release date, pricing or benchmarks yet. For indie game devs it is the open-weights model to watch for cheap 3D and web-game prototyping, since K3 was already on Copilot and AWS Kiro.

Source: https://www.ithome.com/1/017/098.htm

Oct 21, 2026

THIS WEEK: Anthropic previews Claude Duet, a proactive agent app

Claude Duet (research preview, announced Oct 21) is a standalone iOS, Android and web app that watches your connected apps, flags what needs attention, and works on its own cloud computer, asking for approval before sensitive actions. Access is by waitlist for US Pro and Max subscribers. It is another sign that the agent fight is moving from the IDE to an always-on assistant.

Source: https://x.com/claudeai/status/2112984664256241837

Ship faster this weekend

Grab the free Agentdex starter kit with prompts and agent configs for Cursor, Claude Code and Codex.