Anthropic released Claude Code v2.1.266 to address a regression introduced in version 2.1.265. The issue affected users configuring LLM-gateway or proxy setups via the CLAUDE_CODE_USE_GATEWAY environment variable, which previously caused requests to fail with a 'Not signed in to the Cloud gateway' error when used alongside API keys or custom auth headers. This patch restores the original behavior where the variable is ignored unless ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN are both set, allowing existing configurations to function correctly without modification.
Read originalEarlier coverage that leads up to this article, and what followed. Lines connect each piece to the closest one after it, converging here.
MIT Technology Review AI · May 21, 2026 · Related
Matt Wolfe · May 22, 2026 · Related
AI News · May 29, 2026 · Related
Duncan Rogoff · June 19, 2026 · Related
Matt Wolfe · July 29, 2026 · Related
Lev Selector · July 31, 2026 · Related
TechCrunch AI · August 9, 2026 · Related
The AI Daily Brief · August 10, 2026 · Related
The Verge AI · September 17, 2026 · Related
Claude Code Releases · September 18, 2026 · Same story
Duncan Rogoff · September 23, 2026 · Related
Claude Code Releases · October 22, 2026 · Same story
Anthropic is tightening control over Claude Code's autonomous behavior with new environment variables that let you disable 'ultrathink' or cap subagent effort levels. For teams, the introduction of pooled group budgets on the gateway allows shared monthly spending limits across members, a practical fix for enterprise cost management. The release also patches critical bugs where shell commands would hang or crash on exit, stabilizing the developer experience significantly.
This release tightens security boundaries by blocking writes outside working directories across all permission modes, a critical step for enterprise adoption where data leakage is the primary concern. The addition of idle compaction for conversations exceeding 200K tokens addresses the persistent context window bloat that degrades performance in long-running coding sessions. Bug fixes resolve frustrating authentication loops with MCP servers and improve resilience against gateway timeouts, making the tool significantly more stable for complex, multi-step workflows.
This release quietly cements llama.cpp as the universal inference runtime by finally bringing ROCm 10.0 to Linux and Windows alongside CUDA 13.4, effectively closing the hardware gap for AMD users who were previously left behind. The inclusion of native Snapdragon support on Linux marks a significant step toward running local LLMs on ARM-based mobile chips without emulation overhead. While KleidiAI on Apple Silicon is disabled in this build, the broader expansion to diverse accelerators ensures that developers aren't locked into NVIDIA ecosystems anymore.
This release addresses a specific inference artifact where decision prefixes were incorrectly bleeding into output embeddings. By capping shared prefixes at the first output position, llama.cpp ensures that option tokens remain distinct for parent and child tasks. This is a targeted fix for multi-step reasoning or voting workflows rather than a performance boost. It stabilizes complex generation patterns without changing the underlying model architecture.
This release quietly expands llama.cpp's hardware support to include Qualcomm Snapdragon devices with CPU, Adreno GPU, and Hexagon NPU acceleration. It also updates ROCm to version 10.0 for both Linux and Windows, keeping AMD users in step with the latest driver stack. CUDA builds now default to 13.4 libraries alongside 12.x options, reflecting the industry's gradual shift toward newer NVIDIA toolkits. While no new model architectures are introduced, the binary matrix grows significantly, ensuring local inference remains viable on a wider array of edge and server hardware.