Anthropic AI Updates: August 4, 2026

1. Claude Code v2.1.221 Adds Focus View and Sandbox Credential Masking

Anthropic. Claude Code v2.1.221 introduces a Focus view, a chat-menu toggle bound to Ctrl+Alt+F that hides tool activity behind an expandable per-turn summary with a live running-tool indicator. The release also adds a “mask” mode for sandboxed credential files on Linux and WSL that allows masked reads while substituting real values on egress, plus a new prompt-audit subcommand in the claude-api skill and warnings when marketplace or plugin names would be rejected by Claude Desktop’s managed marketplace sync. The updates target cleaner agent output and tighter credential handling for developers running long tool sessions. Source

2. Notability Case Study Shows Model Upgrades Cutting Content Issues

Anthropic. Anthropic published a customer story detailing how note-taking app Notability built its Learn study feature on Claude, using Haiku and Sonnet models to turn student notes into flashcards, quizzes, and summaries. The company reports processing 220 million quiz questions over the past year and cutting user-reported content issues by 45 percent through model upgrades, with the feature shipped in about two months. The account illustrates how education products can ride model improvements without re-architecting their applications. Source