When running pi.dev in the default terminal and Tilix, I occasionally hit severe screen flickering and text jumping. It happens most often while the model is waiting for user input.

According to Gemini AI, the default terminal struggles with rapid frame redraws inside the Crostini Linux container. It is not a deal-breaker, but it is annoying enough for my old eyes.

I installed Ghostty — no flickering or text jumps. However, typing had an obvious lag at first. After enabling GPU support and rebooting, Ghostty runs smoothly.

The following guide is generated by Gemini AI.

Step 1: Enable ChromeOS GPU Support

Before installing Ghostty, you must ensure your Chromebook’s Linux environment has direct access to your graphics hardware.

  1. Open a new Chrome browser tab and navigate to: chrome://flags
  2. Search for Crostini GPU Support.
  3. Change the setting from Default to Enabled.
  4. Click the blue Restart prompt at the bottom right to apply changes.

Step 2: Install Ghostty on Linux

Because Ghostty is not available in standard Debian stable channels, you can install it using the trusted community setup script. Open your default terminal app and execute:

1/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/mkasberg/ghostty-ubuntu/HEAD/install.sh)"

Follow the on-screen prompts to complete the installation. Once finished, you can launch Ghostty directly from your ChromeOS application drawer (under Linux Apps).

Step 3: Verify Hardware Acceleration with glxinfo

To confirm that Ghostty is running natively on your hardware graphics card rather than falling back to an unaccelerated, laggy software emulator, you can check the OpenGL reporting metrics inside the container.

  1. Install the baseline Mesa graphics utilities:
    1sudo apt update && sudo apt install -y mesa-utils
  2. Query your active OpenGL graphics pipeline renderer details:
    1glxinfo -B

Reading Your Results

  • GPU Accelerated (Correct): If the output lists Accelerated: yes and the Device: line shows virgl alongside your graphics card framework (e.g., Mesa Intel(R) HD Graphics 630), your GPU is handling the text rendering pipeline.
  • Software Rendered (Slow): If the device reads llvmpipe, hardware virtualization failed. You will need to explicitly grant user permissions by running sudo usermod -a -G video,render $USER and restarting your Linux subsystem.