Morning brief

Sunday, June 7

1 people · 0 companies · 3 topics

TL;DR A quiet Sunday overnight. Claude Code shipped three patch releases on June 6 adding fallback models and harder session security; steipete’s OpenClaw dropped two fresh betas with Anthropic extended-thinking recovery and SQLite auth.

Tools & dev

  • Claude Code v2.1.166 (June 6, ~01:00 UTC) adds a fallbackModel setting so agents recover when the primary model is unavailable, glob pattern support in deny-rule tool names, hardened cross-session message security, and a MAX_THINKING_TOKENS=0 flag to disable thinking on models that expose it. GitHub releases
  • v2.1.167 and v2.1.168 followed the same day (01:33 and 23:41 UTC) with reliability and bug fixes. GitHub releases
  • The Claude Agent SDK billing split lands June 15: headless claude -p calls, GitHub Actions, and third-party Agent SDK apps move off subscription quota onto a separate monthly credit pool ($20/$100/$200 for Pro/Max 5x/Max 20x). Unused credits do not roll over and there is no automatic fallback once the pool is empty. TechTimes

OSS

  • steipete OpenClaw v2026.6.5-beta.1 (June 6) and v2026.6.5-beta.2 (June 7): QQBot now strips model reasoning before delivery, MCP tool results handle non-text content, Anthropic extended-thinking sessions recover properly on reconnect, a parallel web-search provider is bundled, Google Vertex ADC issues are fixed, Matrix voice-message preflight added, and auth profiles moved to SQLite for durability. GitHub releases