Anthropic released Mods with Claude Code 2.1.287 on October 1, 2026. Small TypeScript functions that hook into Claude Code's runtime β rewriting prompts, blocking or rewriting tool calls, auto-approving permissions, and redacting secrets before the model reads them. The built-in /diff feature now ships as a mod too.
Anthropic added the claude plugin eval command to Claude Code v2.1.269 (September 11, 2026). Every test case runs twice β once with the plugin loaded (with-arm) and once without (without-arm) β and the difference, Ξ, is the plugin's measured contribution. Six grader types, JSON + HTML reports, and CI threshold flags complete a plugin quality-assurance pipeline.
Claude Code v2.1.232 onward brought a wave of August updates: collaborate across sessions via @ mention, automatically resume work when usage limits reset, and get cleaner responses with Concise output mode.
Anthropic added webhooks to Claude Managed Agents. Long-running AI agents now proactively notify your app on key state changes β idle waiting for approval, outcome evaluation complete, errors. With HMAC signing, replay protection, and SDK helpers in 7 languages, this is the practical guide for developers running AI agents in production.
At Google I/O 2026, Gemini 2.5 Pro Deep Think mode went generally available. It scored impressively on the 2025 USAMO math competition and added MCP support to connect with the open-source tool ecosystem. Gemini 2.5 Pro now leads both the WebDev Arena and LM Arena leaderboards.