Claude Code Daily Briefing - 2026-03-11

Release Summary

Version	Date	Key Changes
v2.1.72	3/10	Simplified effort levels, `/plan` with description arg, ExitWorktree, `/copy` write-to-file, ~510KB bundle reduction (latest)

Full Release Notes

New Features & Practical Usage

v2.1.72 — Simplified Effort Levels & Developer QoL Improvements (3/10)

v2.1.72 streamlines effort levels from four tiers (low/medium/high/max) down to three (low/medium/high), with new status bar indicators: ○ (low), ◐ (medium), ● (high).

Key changes:

/plan instant entry: Pass a description directly — /plan fix the auth bug — to enter plan mode immediately
/copy write-to-file: Press w in /copy to write the focused selection directly to a file, bypassing the clipboard — ideal for SSH sessions
ExitWorktree tool: Programmatically exit worktree sessions
Expanded bash auto-approval: lsof, pgrep, tput, ss, fd, and fdfind added to the allowlist
~510KB bundle reduction: Faster loading across the board
Improved voice transcription: Better recognition of repository names and developer terminology

# Jump straight into plan mode
/plan fix the auth bug

# Write to file over SSH (no clipboard needed)
/copy  # → press 'w' to write to file

GitHub Releases

Anthropic Files Federal Lawsuit Against Trump Administration — First Amendment Claims (3/9)

Anthropic filed a federal lawsuit on March 9 challenging both the Pentagon’s supply chain risk designation and the president’s directive to ban Claude from federal agencies, calling them “unprecedented and unlawful” acts of retaliation for constitutionally protected speech.

Core legal arguments:

Pentagon officials admitted on record that the designation is “ideological” with “no evidence of supply-chain risk”
The actions constitute retaliation for Anthropic’s public advocacy on AI safety policy
Anthropic is suffering “irreparable harm” from the designation

Meanwhile, CBS News reports that President Trump plans to issue an executive order this week removing Anthropic’s technology from all federal agencies — expanding the ban from defense contracts to the entire government. Anthropic has requested an emergency injunction.

Notably, Republican Senator Ted Cruz questioned the blanket ban, stating: “I have not seen a basis laid out for why the government would be prohibited from using Anthropic.” Bipartisan concern is emerging.

Bloomberg | CNN | CBS News | The Hill

Developer Workflow Tips

Amazon Mandates Senior Engineer Sign-Off on AI Code Changes — Lessons from Repeated Outages

Amazon now requires senior engineer approval for all AI-assisted code modifications after a string of AI-related outages. The immediate trigger was a 6-hour Amazon.com shopping outage on March 5, following a 13-hour AWS outage in December 2025 caused by the Kiro AI assistant autonomously deleting and recreating an environment.

Internal documents acknowledged that “newly adopted GenAI use cases lack fully established safeguards.”

Takeaways for Claude Code users:

AI code generation speed ≠ deployment readiness: Your deployment pipeline may not be designed for the velocity at which AI produces changes
The “human filter” is non-negotiable: Senior-level review of AI-generated code is a requirement, not an option
Passing tests ≠ production safety: Validate deployment safety and rollback plans, not just functional correctness

Tom’s Hardware | The New Stack

Agent Safehouse — Sandboxing Local AI Agents on macOS

A macOS-native sandboxing tool that prevents locally-running AI agents from making unintended system modifications. It uses a deny-first security model where only explicitly permitted directories are accessible.

Key features:

Single bash script installation with zero dependencies
LLM-based automatic profile generation for minimal-privilege configs
Kernel-level blocking of sensitive files (SSH keys, AWS credentials)
Independent sandbox environments per agent

Useful as an additional security layer alongside Claude Code’s built-in sandbox.

agent-safehouse.dev | GitHub

Security & Limitations

Trump Executive Order to Expand Anthropic Ban Across All Federal Agencies (3/10)

CBS News reports that President Trump will issue an executive order this week removing Anthropic’s technology from all federal agencies — expanding restrictions from defense contractors to the entire U.S. government. Some agencies have already begun offboarding Anthropic contracts, with certain departments given a 6-month transition period. Anthropic has filed for an emergency injunction.

Developer impact: If you’re using Claude API in federal government projects, start planning for alternative model transitions. Civilian and private-sector Claude usage is unaffected.

CBS News | Nextgov | Axios

Ecosystem & Plugins

OpenAI Acquires Promptfoo — AI Security Testing Heats Up (3/9)

OpenAI acquired Promptfoo, an AI security platform used by 25% of Fortune 500 companies for red-teaming and vulnerability detection during development. Promptfoo will remain open source and be integrated into the OpenAI Frontier platform.

With Anthropic launching Claude Code Security and OpenAI acquiring Promptfoo, AI security testing is becoming a core competitive battleground among AI companies.

Bloomberg | TechCrunch

Anthropic Opens Sydney Office — Fourth APAC Hub (3/10)

Anthropic is opening an office in Sydney, its fourth Asia-Pacific location after Tokyo, Bengaluru, and Seoul. Australia and New Zealand rank 4th and 8th globally in per-capita Claude.ai usage. The initial focus will be enterprise, startup, and research customer support, with local infrastructure partnerships being explored for data residency requirements.

Anthropic Official | Bloomberg

Community News

Anthropic nears $19B run rate, overtakes OpenAI in enterprise spending share: Bloomberg reports Anthropic’s run-rate revenue has nearly reached $19 billion — roughly doubling from $9B at end of 2025 in just three months. Enterprise spending share hit 40%, surpassing OpenAI’s 27%. Claude Code business subscriptions have quadrupled since January. Quartz | Bloomberg
Google deepens Pentagon AI push the day after Anthropic’s lawsuit: Google launched Agent Designer, a no-code AI agent builder on GenAI.mil (the Pentagon’s enterprise AI portal), available to 3 million DoD personnel. The timing — one day after Anthropic sued — signals rapid moves to fill Anthropic’s gap in the defense AI market. CNBC
Bloomberg Opinion: “Anthropic brought something new to AI — the power to say No”: Bloomberg argues that Anthropic’s willingness to refuse when every competitor says yes represents its true differentiation and long-term competitive advantage. The Pentagon standoff may cost short-term revenue but is building an “ethical premium” brand. Bloomberg

Minor Changes

v2.1.72 removes max effort: Effort levels simplified to three tiers (low ○ / medium ◐ / high ●) with indicators shown on the logo and spinner.
v2.1.72 /config UX overhaul: Escape cancels, Enter saves/closes, Space toggles — much more intuitive.
v2.1.72 claude plugins alias: Plural alias now works identically to claude plugin.
v2.1.72 worktree isolation fixes: Permission rule matching and sandbox operations in worktree sessions have been corrected.
v2.1.72 voice mode input lag fix: Input delay during Voice Mode has been resolved.

Interesting Projects & Tools

claude-session-tracker (3 points): Automatically saves Claude Code conversations to GitHub. Install with a single command (npx claude-session-tracker), and it creates a GitHub Issue per session, logging prompts and responses as comments. Includes real-time tracking via GitHub Projects board and multi-language status labels (Korean, English, Japanese, Chinese). All hooks run async with zero impact on Claude’s performance. GitHub
FontCrafter (11 points): A free browser-based tool that converts handwriting into installable font files. All processing happens locally — no accounts, no server uploads, no payment. Exports to OTF/TTF/WOFF2 with automatic ligatures, kerning adjustments, and European accent support. Unlike Calligraphr, everything is free with no sign-up required. FontCrafter