[Bug] Antigravity v2.0.6 crashes due to Electron GPU process failure / grey window on Windows 11

Google Antigravity
,
1

Environment

  • Antigravity v2.0.6

  • Windows 11 Pro 64-bit (Build 26200)

  • AMD Ryzen 5 3600

  • NVIDIA RTX 2060

  • NVIDIA Driver: 32.0.15.9186

  • 16GB RAM

Issue

Antigravity fails to launch correctly on my system due to repeated Electron GPU process crashes.

By default, the application exits with:

[ERROR:content\browser\gpu\gpu_process_host.cc:996]
GPU process exited unexpectedly: exit_code=-2147483645

[FATAL:content\browser\gpu\gpu_data_manager_impl_private.cc:417]
GPU process isn't usable. Goodbye.

The language server starts successfully, but the Electron renderer crashes before UI initialization.

Logs

Language server starts correctly:

Starting language server process
Language server listening on random port

Electron repeatedly restarts renderer ports before crashing.

Workaround Found

The application only launches with:

Antigravity.exe --disable-gpu-sandbox --disable-gpu --use-angle=swiftshader

This prevents the fatal GPU crash.

However, after launching successfully, the app only displays a blank grey window with no UI rendered.

Additional Errors

DevTools console shows:

Error: No native storage bridge found during initialization

Also seeing injected scripts such as:

chext_driver.js

Attempted Fixes

Tried all of the following without success:

GPU / Electron flags

Antigravity.exe --disable-gpu

Antigravity.exe --disable-gpu --disable-gpu-compositing

Antigravity.exe --disable-gpu-sandbox --disable-gpu --use-angle=swiftshader

Environment variables

ELECTRON_DISABLE_GPU=1
ANGLE_DEFAULT_PLATFORM=swiftshader

Cleanup

Deleted:

  • GPUCache

  • Cache

  • Local State

Other troubleshooting

  • Reinstalled Antigravity

  • Verified DirectX/D3D status with DxDiag

  • Updated NVIDIA drivers

  • Confirmed no DirectX issues reported by DxDiag

image
image1507×440 39.9 KB

Logs:

Let me know if anything else is needed.

2

In my case, this wasn’t a gpu related issue, rather something like this. It might be a related problem, so sharing this here:

On Windows, the Antigravity desktop applications spawn a backend Go-based language server (language_server.exe) to manage tasks, code analysis, and run commands. Because the Node/Electron host processes and the spawned Go language server share the same console group, any console interrupt signal (such as hitting Ctrl+C in an active terminal or terminating a reloading dev server) propagates to all processes in that console group.

Go’s runtime contains a default signal handler that intercepts CTRL_C_EVENT / SIGINT. When it catches this signal, language_server.exe terminates immediately (usually exiting with code 0).

When the language server crashes or exits:

  1. The Electron frontend loses its IPC connection to the backend.

  2. The UI automatically triggers a page refresh to recover the session.

  3. Your current task execution is abruptly halted, and the screen resets to a new conversation.

To fix this, I used a Python-based wrapper using Windows Job Objects and ctypes to isolate the console signals and stabilize the environment.

If you are using Antigravity on Windows and your screen keeps refreshing unexpectedly, I’ve packaged the wrapper and an automated patcher into an open-source repository to save your sanity.

Check out the fix here: https://github.com/erdometo/antigravity-windows-crash-patch

3

Hi @Milind, @Erdem,

Thank you for reporting this issue. It has now been resolved. Please try updating to the latest version and let us know if you run into any other issues.