Maya: Welcome to Claude Code Review. Hi, I'm Maya. Today: Claude Code's iPhone Duo simulator, a more proactive Projects coordinator, experimental Code Mode, the new Xcode plugin, and Anthropic's model-choice guide.
James: And I'm James. Let's get into it.
Maya: So Claude Code on desktop added iPhone Duo support in public beta on November sixteenth. You run your app in the iOS simulator right beside your coding conversation, and Claude can fold and unfold the simulated device to test each layout.
James: That's a smart setup, Maya, because the running app and the coding chat share the same desktop workspace. So checks of the folded and unfolded states sit right next to the conversation about the code itself.
Maya: Exactly right, James. One requirement to flag: iPhone Duo testing needs macOS and Xcode twenty-seven point one or later. You run the app on iPhone Duo in the desktop iOS simulator, and access remains a public beta.
James: Next, Anthropic updated Claude Code Projects on November thirteenth with a more proactive coordinator. In a project, Claude coordinates your tasks across parallel threads, which are separate strands of work running side by side. Anthropic calls it much more proactive than before.
Maya: That's a real shift, James. The announcement names two changes: the coordinator now answers your questions in the project chat, and it gives you updates on the status of your ongoing threads.
James: Yep, and Anthropic calls these big updates to Projects in Claude Code. The announcement doesn't list further details, so those two behaviors are what this update confirms for now.
Maya: Now to Code Mode. Claude Code's experimental Code Mode lets Claude write JavaScript, a programming language, to call its tools, meaning to use the actions it can perform. Anthropic listed it in the November twelfth release notes for version two point one point three two three.
James: So the feature is really about how Claude Code invokes tools, Maya. Claude writes the JavaScript that makes those calls, and trying the workflow requires opting in. Anthropic directs users to its Code Mode documentation to turn it on.
Maya: Exactly, and the release notes point to that same documentation for turning Code Mode off for an organization. So admins get a clear path both directions, which matters for an experimental option like this.
James: Onto the Xcode plugin. Claude Code's official plugin marketplace now includes an Xcode plugin that lets Claude build, run and test apps from a terminal session, announced November sixteenth. You install it with slash plugin install xcode at claude-plugins-official.
Maya: That's a lot packed into the terminal, James. Beyond build, run and test, Claude can render SwiftUI previews while working on the app, so you see how screens will look without launching the full thing.
James: Exactly right, and there's documentation search too. That extends the workflow to looking up development details, giving terminal sessions a reference tool alongside the capabilities for building and checking the app.
Maya: Finally, Anthropic published a short model-choice guide alongside Claude Fable five point two, its most capable generally available model for technical fields. The default advice: most of the time, start on Claude Opus five point five at medium effort, the daily driver for complex work.
James: And if the result falls short, Maya, the first move is raising the effort level, not switching models. Anthropic says that's a smaller step and usually keeps the prompt cache. Fable is reserved for ambitious, open-ended work and costs twice as much per token.
Maya: Right, and moving down matters too. Sonnet five point five suits well-scoped bugs with a clear reproduction at half the price, and Haiku five point five, the fastest and cheapest, fits high-volume jobs like summarization, classification and routing.
James: That's the full map, Maya. And one job can use several models: the main session hands parts to subagents. For a framework upgrade across every package, Fable leads the migration and hands well-scoped per-package fixes to Sonnet.
Maya: That's your Claude Code Review briefing.