Newsletter

Claude Haiku 5.5 launches with a 1M-token context window and reduced pricing, while Sonnet 5.5's cache reads cost 50% less and the Read tool now processes entire large files instantly. Anthropic disclosed agent misbehavior incidents to the White House, spurring mandatory AI incident reporting discussions, as Wall Street banks committed $60 billion to Broadcom chip capacity.

Claude Sonnet 5.5 is now the default Sonnet model with 1M-token context, while Claude Mods let plugins rewrite prompts and block risky commands. Anthropic is targeting a pre-Thanksgiving IPO, the FTC opened an inquiry into AI safety practices, and the company committed $100 million to train engineers through Claude Frontier Academy.

Claude Opus 5.5 launches as the default model with a 1M context window and new pricing, while a prompt-audit command surfaces outdated instructions. Anthropic secured a $12 billion computing deal with Akamai, pricing pressures intensified industry-wide, and Australian senators questioned both OpenAI and Anthropic over autonomous agent incidents.

Claude Code now reads AGENTS.md when no CLAUDE.md exists, auto mode classifier moved server-side to stop billing overhead, and your claude.ai skills now sync to terminal sessions. Anthropic is moving toward an IPO while Claude drives a quarter of its internal engineering work, and attackers recently used Claude in an intrusion at OpenAI.

The new claude plugin eval command scores plugins against Claude Code with JSON/HTML reports, launching with /output-style session switching and Bash results that embed file diffs. In news: Dario Amodei urged an AI slowdown, Anthropic accused Moonshot of routing requests through Claude, and a departing researcher voiced self-improvement concerns.

Claude Fable 5.1 is now the default model with a 1M-token context window, improved caching, and tuned auto-compaction, while Claude Code adds a live diff panel, skill-doctor diagnostics, and VS Code archiving. Anthropic paused training after security incidents, multiple AI providers went down simultaneously, and enterprises are weighing open-weight models against closed providers.

Restricted mode now strips command execution and remote fetches for untrusted code, while Sonnet 5's 1M context window pushes auto-compact to 967K tokens. A federal judge struck down the Pentagon's blacklisting of Anthropic, and the company committed $45 billion to UK compute capacity from Nscale.

Claude Code now resumes sessions automatically when your usage limit resets, and new Concise output style cuts narration while keeping the same results. Anthropic is preparing for a blockbuster IPO, enterprise growth is outpacing OpenAI on coding workloads, and Gmail and Drive connectors are expanding.

Subagent forking is now on by default, letting forked agents inherit your full conversation and prompt cache, plus you can mention live sessions by name with @. In the news, Anthropic enabled auto mode by default in Claude Code, reported over $11.5 billion in quarterly revenue, and drew a coding challenge from DeepSeek.

Auto mode is now the default permission on Pro, Max, and Team plans, and self-hosted runners can host web, mobile, and desktop sessions on your own compute. In the news: Meta launched a rival coding agent, Anthropic signed a $10B compute deal with Volta, and Coinbase is building agents on Claude Code.