What's new in Claude Code

A twenty-minute summary of what's new in Claude Code: what shipped, why we built it, and how to get started.

May 5, 202624mWatch on YouTube ↗

WHAT IT’S REALLY ABOUT

Claude Code updates focus on smoother workflow and greater autonomy

Claude Code’s updates cluster into two themes: improving day-to-day developer experience and increasing agent autonomy for long-running work.
Remote Control lets you start a Claude Code session on your machine and continue it from the web or mobile with the same terminal context.
A new flicker-free full-screen Terminal UI virtualizes scrollback to prevent repaint artifacts while enabling clickable UI elements and stable memory usage.
The desktop/web GUI now supports multi-session management with split views, plans/diffs/files panes, inline comments, and an experimental “pin as chapter” transcript table-of-contents.
Autonomy features include automode permission decisions, safer parallel work via Git worktrees, persistent project auto-memory, multi-agent code review (/ultrareview), and scheduled automation through routines and /loop.

IDEAS WORTH REMEMBERING

5 ideas

Remote Control turns a single session into a portable dev environment.

You can start a session locally, enable remote control, and then interact with the same terminal context from a link on the web (including mobile), making it practical to kick off long tasks and check in while away.

Full-screen TUI eliminates terminal flicker by changing the rendering model.

Instead of appending to traditional scrollback (prone to repaint misalignment), Claude Code virtualizes scrollback in full-screen mode for smooth output, clickable elements, and flat memory usage even in long sessions.

The GUI is shifting from “chat transcript” to a multi-surface workspace.

You can manage multiple sessions via pinned/grouped sidebar and split views, while optional panes (plan, diff, files) and line-level comments help you review and steer work without leaving Claude Code.

Transcript chapters make long sessions navigable and reviewable.

The experimental “Pin as chapter” feature labels key assistant messages, creating a table of contents for jumping across long histories—useful when a session spans multiple debugging threads or PRs.

Automode reduces babysitting by making permission decisions automatically.

A classifier evaluates tool calls for destructiveness and prompt-injection risk; safe actions proceed without prompts, while risky ones are blocked so the agent must adapt rather than waiting on you.

WORDS WORTH SAVING

5 quotes

Our new features fall into two themes. First, developer experience, how we're making Claude Code nicer to live in as you spend more of your day in, uh, with it. Second, autonomy, how we're enabling Claude to do more while you're away from your keyboard.

— Dixon Tsai

Remote control lets you start a session on your machine and pick it up on the go. From your phone, you can access the same session, same dev setup. Why not kick off a long-running task and run some errands?

— Dixon Tsai

We've heard from you loud and clear that flickering ruins the terminal experience.

— Dixon Tsai

Automode is a new permissions mode in Claude Code where Claude makes permission decisions on your behalf using a classifier.

— Dixon Tsai

Your time is valuable, and we should leverage Claude's ability to make the most of your time.

— Dixon Tsai

Remote Control (mobile/web session continuation)Terminal UI full-screen mode and virtualized scrollbackVoice mode in Claude Code (/voice)Revamped desktop/web GUI: split view, plan/diff/files, commentsTranscript “Pin as chapter” table of contents (experimental)Autonomy: automode permissions, worktrees, auto memory, code reviewRoutines and /loop scheduling; connectors (GitHub/Slack); tool search

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.