While the source code leak discussion is still reverberating, Anthropic has quietly shipped two substantial Claude Code updates. Versions 2.1.89 (April 1) and 2.1.90 (April 2) deliver a mix of new features and important fixes.
/powerup — Interactive lessons
The most visible new feature: the /powerup command launches interactive tutorials with animated demos that teach you Claude Code features right in the terminal. Instead of reading docs, you learn by doing. A smart idea, especially for newcomers.
Performance and stability
The most impactful fix under the hood: a bug was resolved where --resume caused a complete prompt cache miss when deferred tools, MCP servers, or custom agents were active. That means faster session resumption for anyone working with complex setups.
There’s also flicker-free rendering via the new CLAUDE_CODE_NO_FLICKER=1 environment variable — helpful in terminals that struggled with standard rendering.
Other highlights
- New ‘defer’ permission option for headless sessions
.huskydirectory now protected in acceptEdits mode- Fix for an infinite loop where the rate-limit dialog kept reopening after hitting usage limits
- Hook system extended with
PermissionDeniedevents - Better Windows support (CRLF handling, PowerShell permissions)
What it means
These updates won’t grab headlines, but they’re exactly the kind of work that makes Claude Code more reliable day to day. The /powerup command lowers the entry barrier, the performance fixes save time in practice. Solid engineering work.
Sources: