Claude Code Daily Briefing - 2026-05-08

Release Summary

VersionDateKey Changes
v2.1.1335/7worktree.baseRef setting, effort level in hooks, memory pressure worker release, parallel session 401 fix
v2.1.1325/6CLAUDE_CODE_SESSION_ID, fullscreen disable option, SIGINT graceful shutdown

Full Release Notes

New Features & Practical Usage

v2.1.133 — worktree.baseRef, Hook Effort Level, Memory Optimization (5/7)

Released on the same day as Code with Claude Extended SF, v2.1.133 adds significant worktree and enterprise management features.

worktree.baseRef setting: Choose between fresh (branch from origin/<default>) or head (branch from local HEAD) when creating worktrees. This makes the v2.1.128 HEAD-based fix configurable per project.

Effort level in hooks: Hooks now receive the active effort level via effort.level JSON field and $CLAUDE_EFFORT environment variable. This enables conditional hook behavior—e.g., running full lint only in high-effort mode.

parentSettingsBehavior admin key: Controls policy merge strategy (first-wins or merge) for SDK managedSettings. Useful for resolving admin vs. project settings conflicts in enterprise deployments.

Memory optimization: Warm-spare background workers are now released under memory pressure, reducing overall memory footprint.

{
  "worktree": {
    "baseRef": "head"
  }
}

GitHub v2.1.133

Claude Security Public Beta — Opus 4.7 Vulnerability Scanning (5/4+)

Anthropic launched Claude Security in public beta for Claude Enterprise customers, with Team and Max access coming soon.

Key capabilities:

Partner ecosystem: CrowdStrike, Microsoft Security, Palo Alto Networks, SentinelOne, and Wiz are integrating Opus 4.7 into their existing security platforms. Service firms including Accenture, Deloitte, and PwC are deploying Claude-integrated security solutions.

Accessible directly at claude.ai/security with no API integration required—designed to lower the adoption barrier for security teams.

Help Net Security | DevOps.com | IT Pro

Anthropic Acquires Vercept — Computer Vision & UI Automation

Announced at Code with Claude SF, Anthropic acquired Vercept, a company specializing in AI perception and UI interaction. The acquisition strengthens Anthropic’s desktop and mobile automation capabilities beyond the existing Computer Use feature.

This signals a next-generation agent roadmap combining visual understanding with autonomous manipulation—potentially impacting GUI testing, browser automation, and mobile app verification in Claude Code.

Blake Crosley

Developer Workflow Tips

Using Effort Level in Hooks

With v2.1.133 passing $CLAUDE_EFFORT to hooks, you can scale validation intensity based on effort level. Light checks for quick exploration, full lint and tests for serious implementation:

{
  "hooks": {
    "PostToolUse": [{
      "matcher": "Write",
      "hooks": [{
        "type": "command",
        "command": "if [ \"$CLAUDE_EFFORT\" = \"high\" ]; then npm run lint -- --fix $CLAUDE_FILE_PATH; fi"
      }]
    }]
  }
}

“Claude Code Isn’t Improving Your Product” — The K-Shaped Productivity Gap

A thought-provoking analysis gaining traction in the community argues that code velocity and product quality are separate concerns.

Key arguments:

The practical lesson: Claude Code delivers real value when used to validate better designs with less code, not to produce more code faster.

Substack | GeekNews

Security & Limitations

Claude Security Public Beta — See New Features Section Above

This is Anthropic’s practical response to Dario Amodei’s “moment of danger” warning. Following Mythos discovering tens of thousands of vulnerabilities, enterprise customers can now scan their own codebases at a similar level. Currently Enterprise-only; Team/Max timeline TBD.

Ecosystem & Plugins

Code with Claude London (5/19) & Tokyo (6/10) — Registration Open

Following the successful SF main event and Extended sessions, the next stops are London (May 19) and Tokyo (June 10). Both events will be livestreamed. Tokyo also has an Extended session on June 11.

The SF Extended focused on indie developers and early-stage founders sharing their journey from prototype to first revenue with Claude. Session recordings will be published later.

Code with Claude | Tokyo Extended

Plugin Ecosystem Scale — 4,200+ Skills, 770+ MCP Servers

As of Code with Claude SF, the ecosystem has grown to 4,200+ Skills, 770+ MCP servers, and 2,500+ marketplace listings. Notable additions include Meta’s Facebook/Instagram Ads MCP (launched 4/29), Google Ads MCP, Shopify AI Toolkit, and Klaviyo MCP for email/SMS automation.

Community News

Minor Changes Worth Knowing

Interesting Projects & Tools

By subscribing, you agree to our Privacy Policy.