Claude Code has evolved overnight!
Just today, Anthropic officially released an epic-level update for the Claude Code desktop version!
It can now launch multiple conversations simultaneously, running tasks multi-threaded, and whoever finishes first throws the result back to the main session.

Furthermore, the new interface has completely revamped session management.
It not only supports environment reuse and one-click launching of subsequent tasks but also allows you to type /btw to instantly switch to a new tab and start running tasks separately.
Almost instantly, the global developer community was flooded with the same news.
Anthropic engineer Thariq could hardly contain his excitement, exclaiming, "This is almost a v2 version"!

Claude Auto-Splits Itself, Multiple Sessions Work Simultaneously
First, let's look at the most eye-catching feature this time: automatic creation of nested sessions.
Previously, if you wanted the AI to do three things at once, you had to manually open three windows and then check the results one by one yourself.
Worse, with too many tasks, the main thread would turn into a disaster zone: front-end code mixed with back-end error messages, and to trace back a specific modification step, you had to frantically scroll your mouse in front of the screen.
Now, this logic has completely changed.
In the new Claude Code desktop version, you only need to hand over all tasks at once in the main dialog.
When Claude recognizes that these tasks can be processed in parallel, it instantly performs a "Shadow Clone Jutsu"—automatically creating an independent nested session for each subtask!

Even better is its information flow mechanism:
Underlying Concurrent Execution: Several sub-sessions start working simultaneously in the background, each reading context, modifying code, running tests, without interfering with each other.
Extremely Clean Main Thread: Their tedious intermediate trial-and-error and hundreds of lines of terminal output won't pollute the main conversation interface at all.
Automatic Summary and Reporting: When each sub-session finishes its work, it automatically reports the final execution results back to the main session neatly, just like a subordinate reporting to a director.
This way, the main thread maintains extremely high readability from start to finish.
Humans don't need to worry about the process of debugging those hundreds of lines of error messages in between. They can just sit in the main session, reviewing the final answers submitted by each AI like approving official documents.
If a Claude clone gets stuck or you want to see the details, you can click through to the underlying session at any time.
This round of upgrades makes it easier for the AI to work separately. But engineering acceptance still requires human oversight.
How Can a Person Keep Track of a Pile of AI Conversations?
When several Clauses start working together, you immediately run into another problem: who should you look at?
One is still analyzing code, one is running commands, and another might have stopped waiting for your decision. If you can only click into each one to check the records, the waiting time you saved gets eaten up by switching windows.
The new version moves these statuses to more visible positions.
The sidebar allows users to clearly see which sessions are working and which require personal intervention.
After entering a session, a real-time to-do list follows Claude's plan as it progresses; when it needs a user response, a review card will summarize the situation of this round.

The side information bar displays the task's running location, context, and code changes.
Second-Level Entry into the "Battlefield," One-Click Environment Reuse
Besides concurrency during task execution, Anthropic has also taken "pre-work preparation" to the extreme this time.
Many full-stack architects' daily development environments are very complex: the front-end is in one folder, backend microservices are in another, and some long-running tasks even need to be run in the cloud.
Every time you restart the software or switch projects, it takes a long time just to configure these directories and connections properly.
The new Claude Code has drastically restructured the "new session interface."
The core goal is four words: seamless continuation.
One-Click Access to Commonly Used Directories
When you open the interface, "Recently used folders" are directly placed in the most prominent position. No need to search through the system directory tree—just one click and you can start working right away.
Custom "Environment" Packaged and Saved
Now, you can package and save an entire complex development configuration directly as a named "Environment."
For example, a combination of several local folders, a specific SSH remote host, or a cloud runtime environment. The next time you need to use it, you can instantly launch the entire multi-directory, cross-end workbench with one click.

Native-Level Git Worktree Isolation
Since there are multiple concurrent sessions, wouldn't it be a problem if two AIs modify code in the same local folder simultaneously?
Anthropic has already thought of this for you!
When launching multiple sessions, the new CC directly provides the option for "Independent Worktree for each folder."
Just tick the box, and Claude will automatically use Git Worktree to pull up independent worktree branches for different parallel sessions.
Session A refactors the underlying architecture, Session B fixes an urgent online bug—both are physically isolated at the file level. No more worrying about code overwriting each other or the local environment crashing.
Detail-Obsessed, /btw to Chime in Anytime, Actively Recommends Follow-up Tasks
Why did Thariq marvel that the new Claude Code desktop version has "an extremely strong sense of self-awareness of its own capabilities"?
The answer lies in two clever interactive details: proactive recommendations and the /btw side conversation.
Previous AI tools were like "abacus beads"—you had to push them to move.
The new CC, while working, actively recommends the next steps based on the current changes.
No need to type for confirmation; just click once, and the follow-up suggestions will directly start running in the background as independent sessions! Additionally, the empty input box will intelligently display grey prompt suggestions, which can be instantly adopted by pressing the Tab key.

Another design that makes programmers applaud is the newly added /btw command (by the way).
During the main task progression, you inevitably have temporary questions, like "By the way, what's the default expiration time for this authentication token?"
Now, in CC, just type /btw (or use the shortcut Cmd + ; / Ctrl + ;):
The system immediately opens a side conversation in an independent tab, fully inheriting all the context from the main session up to that point, with no need to re-explain the background.
The main session continues at full speed, completely undisturbed! The side question will never pollute the main thread record. Once you resolve your doubt and close the tab, the main work remains clean.
This feature seems simple, but it solves a very fundamental pain point: The "flow state" should not be interrupted.
Programmers fear the "context-switching tax" the most. Every time they are interrupted and try to return to a state of deep work, it might take over ten minutes.
/btw allows people to resolve temporary questions without interrupting the main workflow. Often, efficiency is saved bit by bit in this way.

Moreover, Claude Code has added an "Idle Compaction" feature.
When a session remains idle for a long time, CC will automatically compress the session context before the prompt cache is about to expire (around 1 hour).
This retains critical context while reducing token consumption, saving on usage quotas.

The Second Half of AI Programming: Competing on "Team Leadership"
To put it bluntly, this major overhaul of the Claude Code desktop version directly pushes AI programming into its second half.
It's evolving from a single "AI Programming Assistant" into a full-fledged "Multi-Agent Collaboration Workbench."
Over the past two years, the AI coding competition focused on single-point capabilities—code completion, single-generation accuracy, scores on benchmarks like SWE-bench.
But today, the bottleneck has become "How can the AI simultaneously advance the development of ten modules without messing up the engineering?"
As Thariq said, Claude Code is highly aware of its capabilities and knows how to cooperate with humans by spawning sub-sessions.

Most importantly, under this new paradigm, the role of human developers has fundamentally shifted.
The era of single-threaded work is ending. The "Multi-Agent Parallel Engineering Army" for each individual is officially launched today.
References:
https://x.com/ClaudeDevs/status/2110445849516101743
https://code.claude.com/docs/en/desktop-quickstart
https://x.com/trq212/status/2110463962877989276
This article is from the WeChat public account "Xin Zhiyuan," author: Peach (Taozi).








