Anthropic just pushed Claude Code past the terminal.

Computer use is now available in research preview inside Claude Code, which means Claude can open files, use dev tools, point, click, and navigate what is on your screen while working through a task. Anthropic’s changelog shows the current Claude Code version at 2.1.88 as of March 30, 2026, and recent Claude app release notes say computer use rolled out in research preview for Pro and Max users on macOS.

That is the real shift.

Claude Code is no longer just something that writes code and runs commands. It is starting to actually interact with the machine the way a human would. For builders, that means the gap between “Claude suggested it” and “Claude tested it” just got a lot smaller.

This is a bigger update than it sounds like at first.

Anthropic says Claude can now open files, navigate apps and browsers, and run development tools directly on macOS in this early preview. The company also notes that the system asks for explicit permission before acting, and it is sharing the feature in an early stage to gather feedback and understand limitations.

That matters because Claude Code has been steadily moving from assistant to operator. The product was already useful when it could inspect a repo, write code, and run commands. But once it can also click through interfaces and validate what it built in the actual environment, it stops feeling like a fancy terminal add-on and starts feeling like a real execution layer.

The simple version is this:

Claude can now act on the machine, not just talk about what to do next.

That means it can move through the UI, launch apps, inspect outputs, and help complete tasks that used to require the human to bridge the last mile manually. Anthropic’s public documentation frames this as part of Claude’s broader computer-use capability, and The Verge reports it now extends into both Claude Code and Cowork.

That changes the loop.

Before, Claude Code could help write, refactor, and run. Now it can also help verify, navigate, and interact. That may sound like a small product extension, but it is really a change in posture. Claude is getting closer to handling the whole workflow end to end.

For most developers, the bottleneck is not always writing the next line of code.

It is the annoying space between steps.

Open the app.
Click through the UI.
See if the change worked.
Go back.
Fix it.
Run it again.

Computer use starts eating into that friction.

Anthropic is effectively pushing Claude Code from a coding assistant into something that can both build and verify. That is where these products start becoming much more useful in real workflows. It is also where trust starts to matter more, because once the model is acting on the machine, mistakes stop being theoretical. Anthropic says the preview still has limitations and should be treated as early-stage.

The vibe is pretty clear.

Anthropic is trying to turn Claude Code into the thing that does the work, not just the thing that helps you think about the work.

And the pace has been fast. Claude Code’s changelog continues to update rapidly, and Anthropic is now layering in more autonomy with things like auto mode, which lets Claude decide some permissions on its own in research preview for Team users. Taken together, the direction is obvious. This is no longer one isolated feature drop. It is a broader product shift.

The AI coding race is turning into an AI execution race.

Qwen 3.6 Plus Preview is now live on OpenRouter, and Alibaba Qwen’s official X account says it is available free for a limited time there. OpenRouter describes it as the next evolution of the Qwen Plus series with a hybrid architecture, stronger reasoning, and more reliable agentic behavior than the 3.5 series, especially in agentic coding, front-end work, and complex problem-solving.

That matters because Qwen is not just trying to be cheap or available. It is pushing harder into reasoning and agent workflows too. So today’s picture is pretty simple: Anthropic is expanding what the agent can do on your computer, while Qwen is upgrading the model layer underneath the broader agent race.

Check Out Our Tweet To Learn More! - Tweet

Anthropic just gave Claude Code computer use in research preview, letting it move beyond the terminal and actually interact with your Mac. That is a meaningful jump because it compresses the distance between writing code and testing what got built. Claude is getting closer to being an operator, not just an assistant.

At the same time, Qwen 3.6 Plus Preview just landed, adding more heat to the reasoning and agent side of the model race. So today is not just about one feature. It is about the next layer of the AI stack becoming more real, more useful, and a lot more hands-on.

Stay building. 🤖