Anthropic Adds iPhone Duo Simulator to Claude Code Desktop

Anthropic has launched a public beta that integrates Apple’s iPhone Duo simulator into Claude Code Desktop for macOS, enabling developers to test applications designed for foldable devices directly within the AI coding assistant. The feature supports foldable-device testing, including folding and unfolding layouts, alongside an official Xcode plugin. The release arrives as part of a broader update that also includes new capabilities for running Code Mode alongside agent teams.

The iPhone Duo simulator integration represents a significant expansion of Claude Code’s automated testing capabilities, particularly for developers building apps that need to adapt to changing screen configurations. Claude Code Desktop’s new iPhone Duo simulator integration is designed to automate testing across foldable screen states. The AI assistant can now inspect layouts, identify interface problems, and suggest or apply fixes while developers work within the coding interface.

Concurrent Development Features in Latest Release
The public beta coincides with the release of Claude Code v2.1.326, which introduces support for running Code Mode while agent teams are enabled. The release also introduces warnings for malformed tool-rule flags and adds telemetry-status information. This allows developers to leverage multiple AI-driven workflows simultaneously, potentially streamlining complex development tasks that require both individual code assistance and coordinated multi-agent operations.
The combination of simulator testing and concurrent agent support reflects Anthropic’s efforts to position Claude Code as a comprehensive development environment rather than a simple code completion tool. Developers can now test responsive layouts for foldable devices while maintaining access to other AI-assisted development features.
Security Concerns and Vulnerability Disclosure Updates
The updates come amid heightened security scrutiny of AI coding assistants. Security researchers reportedly found more than 100,000 public files containing Claude Code settings, with roughly 1,100 allegedly containing secrets. The issue centers on remembered command approvals being saved locally and subsequently committed to public repositories, potentially exposing sensitive credentials and configuration details.

In response to growing security concerns across the industry, Anthropic is reportedly shortening its coordinated vulnerability-disclosure window from 90 days to 30 days for high- and critical-severity issues, with a 45-day window for other reports. The policy change is scheduled to take effect for reports submitted from January 1, 2027, reflecting a faster response timeline for addressing critical security flaws.
Security issues have extended beyond configuration management. Google researchers say suspected North Korean cyber actors used Claude Code and other AI tools during intrusions targeting cryptocurrency organizations. The activity reportedly combined AI-assisted social engineering with attempts to distribute malware through victims’ development environments, highlighting how threat actors are incorporating AI coding tools into their operations.

Verification and Quality Assurance Approaches
As Claude Code expands its automated capabilities, the development team has been discussing its approach to verification for AI-generated work. A recent discussion highlighted how the team approaches verification strategies, addressing concerns about ensuring the accuracy and reliability of code produced by the AI assistant. The conversations have covered recent service incidents and updates to Claude Code Projects, suggesting ongoing refinement of quality assurance processes.

Expanding Canvas Capabilities
Beyond coding and testing features, Anthropic has been exploring new ways to integrate visual design into its AI tools. Anthropic engineer Felix Rieseberg reportedly demonstrated a feature that places real HTML elements, including text and buttons, inside canvas graphics and 3D effects. The HTML-in-Canvas capability could support AI-assisted web design by combining visual canvas layouts with selectable, clickable page components, potentially bridging the gap between design and development workflows.
Key Facts
- Claude Code Desktop now supports Apple’s iPhone Duo simulator for testing foldable device layouts in public beta
- Claude Code v2.1.326 enables Code Mode to run concurrently with agent teams
- Approximately 1,100 public repositories allegedly contain exposed secrets in Claude Code settings files
- Anthropic will reduce vulnerability disclosure windows to 30 days for critical issues starting January 1, 2027
- Google researchers identified North Korean threat actors using Claude Code during cryptocurrency-focused intrusions
Industry Context
The foldable device testing capability addresses a growing market segment as manufacturers continue to release devices with flexible screens. While Apple has not yet released a foldable iPhone commercially, the simulator allows developers to prepare applications for potential future hardware releases and test adaptive layout principles that apply across different screen configurations.
The simultaneous focus on expanded capabilities and tightened security policies reflects broader tensions in the AI coding assistant market. As these tools become more powerful and widely adopted, they face increased scrutiny from both security researchers and threat actors. The discovery of exposed secrets in public repositories and the reported use of AI coding tools by nation-state actors underscore the dual-use nature of these technologies.
Anthropic’s accelerated vulnerability disclosure timeline aligns with industry trends toward faster security response cycles, particularly as AI tools are integrated into critical development workflows at major organizations. The 30-day window for high-severity issues represents a significant reduction from the previous 90-day standard, potentially reducing the window of exposure for critical vulnerabilities.
Sources
- DigIntoAI: Claude Code Desktop on macOS adds iPhone Duo simulator testing
- VibeHacker: Claude Code v2.1.326 lets Code Mode run alongside agent teams
- AI Power Rankings: Claude Code settings files reportedly expose secrets in public repositories
- My2Cents.ai: Anthropic updates Claude Code vulnerability disclosure timelines







