Anthropic's Claude Code Introduces NO_FLICKER Mode to Eliminate Terminal Flickering
Anthropic has released an experimental new renderer for their Claude Code AI assistant that eliminates the terminal flickering issue by using a virtual viewport to only update changed characters, rather than redrawing the entire screen.
Why it matters
The new NO_FLICKER renderer in Claude Code significantly improves the user experience by eliminating the distracting terminal flickering, making it a more seamless and productive tool for developers and AI enthusiasts.
Key Points
- 1The old renderer caused flickering by redrawing the entire screen on every update
- 2The new NO_FLICKER mode uses a virtual viewport to only update changed characters
- 3It also adds mouse support, cleaner copy-paste, and more stable CPU/memory usage
- 4The new renderer is experimental but preferred by most Anthropic internal users
Details
Anthropic has developed a new experimental renderer for their Claude Code AI assistant that addresses the common issue of terminal flickering. The old renderer would clear the entire screen and repaint everything from scratch on every update, leading to visible flickering, especially during long code generation sessions. The new NO_FLICKER mode uses a virtual viewport approach, where the renderer maintains an internal copy of the screen and only updates the specific characters and lines that have changed. This diff-based rendering eliminates the need for full screen redraws, resulting in a much smoother and more stable experience. Additionally, the new renderer adds support for mouse events, allowing users to click, scroll, and interact with the Claude Code interface using a mouse or trackpad. It also provides cleaner text selection and copy-paste functionality, as well as more consistent CPU and memory usage over long sessions. While still experimental, Anthropic has stated that most of their internal users already prefer the new NO_FLICKER mode.
No comments yet
Be the first to comment