Claude grew "hands" late at night, took over the computer and ran the code at full speed. The quota was used up at the speed of light, and the whole network wailed.

Claude is really taking off! Today, Claude Code officially launched the "Computer Usage" feature, allowing direct control of the CLI to write code, operate the UI, and fix bugs. With a single click, you can activate the "Autopilot" mode, completely freeing the hands of workers.

Claude's delivery speed is simply insane!

Early this morning, Anthropic dropped another bombshell -

The "Computer Usage" capability is officially integrated into Claude Code.

This means that Claude is no longer just a chat AI hiding behind a dialog box.

It now has "hands" and can directly reside in the CLI, taking over the computer with a single click to get the job done.

Now, Claude Code can, just like a "real - life programmer", independently complete the closed - loop of development, debugging, and testing.

Moreover, with just a single prompt, Claude can complete the entire process from writing code, compiling, launching the application to automatic selection and testing.

If the program crashes, it can search for bugs on its own, fix them, and complete the verification.

This has completely panicked netizens. Some people say that this moment marks the official replacement of software engineers by Claude.

Currently, this feature is available in a "Research Preview" for Pro and Max users, only on macOS.

Claude Code finally grows "hands and eyes"

Humans can only watch

Actually, Claude Code is already very powerful: it can understand your entire codebase, write code, modify files, and run commands.

However, its capabilities are ultimately limited to the terminal and text world.

Once the workflow goes beyond the terminal and enters the browser, desktop applications, or system UI, humans have to take over.

From now on, Claude can directly take over the Mac, controlling the browser, mouse, keyboard, and screen to complete tasks.

Just enter /mcp in the terminal to activate Claude's "Autopilot" mode.

After integrating the computer usage capability, Claude Code can perform the following operations -

Cross - application interaction: Open various apps installed on the computer and perform clicks and swipes on the UI interface.

End - to - end closed - loop: With a single command, it can complete the entire process task:

Write code -> Compile -> Launch App -> UI automation click -> Discover Bug -> Fix code -> Verify again.

Ignore tool boundaries: Whether it's a locally compiled SwiftUI application, an Electron project, or a graphical tool without a CLI, it can directly start operating.

The most core breakthrough in this update lies in the "autonomous debugging" ability in complex environments.

In the past, when there was a problem with the code running, people needed to manually take screenshots and provide feedback to the AI, or copy the error information, which was time - consuming and labor - intensive.

Now, Claude can directly see the interface of the built program and find visual or logical vulnerabilities by simulating user operations.

This "what you see is what you get" debugging method greatly reduces the cost for developers to switch between different tools.

This feature update truly realizes a complete closed - loop for developers, allowing them to complete development without manual intervention throughout the process.

Although Claude Code has become more powerful, without sufficient quotas, developers can only sigh in frustration.

Quotas are used up "at the speed of light", and the whole network is wailing

Because today, Claude has run into big trouble...

Just one day into this week, global developers have collectively hit the "quota wall" of Claude Code.

Even those "Max premium users" who spent $200 have become "suckers", receiving quota limit warnings before they've even used much.

This sudden quota limit has directly triggered a wave of complaints across the network, and posts about the "Claude quota shortage" are flooding the screen.

What's most heart - breaking is that before Monday is even over, Claude has already "clocked out".

Swipe up and down to view

Facing the overwhelming wailing, CC engineers urgently responded that an in - depth internal investigation is underway.

However, as of now, the cause of this "problem" remains a mystery.

Exposed serious bugs, Token cost soars 20 times

A guru on Reddit couldn't sit still.

He conducted a reverse - engineering of Claude's binary files through a man - in - the - middle (MITM) attack and discovered a shocking truth:

There are two serious bugs in the system's underlying layer, which directly lead to cache invalidation.

You know, once the cache is invalidated, the Token consumption cost will soar 10 - 20 times instantly, which is simply "murdering" users' quotas.

If you're using it in an API call, the situation will be even worse. His in - depth research found that these two bugs are extremely hidden.

The first one is a string replacement bug in the Bun runtime environment. Since Claude's independent CLI comes with a customized binary file, it causes the cache to be frequently invalidated.

The currently known temporary solution is to use npx @anthropic - ai/claude - code to run it.

The second bug is even more troublesome. When using the -- resume command to resume a session, the cache will crash 100% of the time.

Currently, there is almost no solution other than "rolling back" to an older version that sacrifices many functions.