Anigravity 1.23.2 will not start

[Jay_Cee]

title- just blank ui says it is hung up. restarted muliple times. any ideas?

[huoxin]

same issue here, happened when I switch the model to gemini 3.6 flash and it stop responding, cant even reload or do anything now

[Jay_Cee]

any input from the google team would be great. this is the worst day for this to happen.

[Roberto_Garcia]

same here, after select 3.6 flash, pum, stop working and can´t go back, app freeze everytime i´ve opened.
I tried all, restart win, remove cache folders… nothing.

[chetanpatel_LNU]

Same issue. Stuck for me too… Nothing worked.

[Jay_Cee]

worst possible day for this to happen. just tried antigravity ide 2.1.1, it wont register/import my past conversations/brain. have tried everything on that end. what a disaster. now the agent is locking up in antigravity ide…

[fatihlimitless]

yes, I also switched to 3.6 and everything was messed up

[Taniv_Ashraf]

Same here in MacOS latest version. Anyone got the solution ?

[Roberto_Garcia]

well, the only solution for me, remove 1.23 app, and install IDE 3.11… i´ve lost all my history, brain, extensions… you know. But at least i can work again.
The problema is i´m working now with fear something goes wrong again.
Just in case, i keep on working with 3.5 versions of model.

This is insane, having fear of choosing a model on the chat selector for a paid subsription i use to work, this is not for fun, it´s for work, to make money so i can pay de subscription!!!

Well, that ´s all for me today.

[mehechiger]

Update: further isolated testing found a second runtime failure mode affecting new windows after sign-in; details are in my follow-up reply below.

Sharing detailed findings and a non-destructive workaround in case this helps the Antigravity team investigate.

Reproducible root cause and non-destructive workaround for Antigravity 1.23.2

I reproduced this issue on macOS Monterey 12.7.6 after selecting Gemini 3.6 Flash High.

The sequence is:

1. Gemini 3.6 is added to the remotely fetched model catalog.

2. Selecting 3.6 freezes the renderer.

3. After force quitting, the new model state remains persisted.

4. Subsequent cold starts freeze before the workbench finishes loading.

I compared a known-good profile with the affected profile.

Key findings:

• The main SQLite databases pass PRAGMA quick_check; this is not physical database corruption.

• Normal settings files are unchanged.

• The critical persisted state is in:

User/globalStorage/state.vscdb

• The main affected key is:

antigravityUnifiedStateSync.userStatus

The known-good value is 8,252 bytes. After the Gemini 3.6 rollout it becomes 12,396 bytes and contains new entries for:

Gemini 3.6 Flash (High)
Gemini 3.6 Flash (Medium)
Gemini 3.6 Flash (Low)

• Once the new userStatus has been persisted, Antigravity 1.23.2 freezes on the next cold start even when modelPreferences is reset.

• The unresponsive renderer samples repeatedly include:

CodeWindow: detected unresponsive
at Array.sort (<anonymous>)

• Switching to Gemini 3.5 works in the current process.

• Switching back to Gemini 3.6 reproduces the freeze before the 3.6 preference is written.

I built and tested a non-destructive recovery launcher. Before every start it restores only these two keys from a pre-3.6 known-good state:

antigravityUnifiedStateSync.userStatus
antigravityUnifiedStateSync.modelPreferences

It then verifies the SQLite database and launches the same user profile. Workspaces, settings, extensions, and history remain intact. The workaround also recovers the application after accidentally selecting 3.6 again.

This strongly suggests that Antigravity 1.23.2 is receiving a Gemini 3.6 tiered model catalog that its cold-start model restoration or sorting path cannot safely process.

A server-side mitigation may be sufficient:

• do not return Gemini 3.6 tiered entries to client version 1.23.2;

• return the legacy model catalog to V1 clients;

• or provide compatibility-safe sorting/default fields.

I can provide a sanitized forensic report, hashes, protobuf summaries, and the recovery script. I will not upload the full database because it contains private account and project state.

Could someone from the Antigravity team confirm whether a server-side compatibility change is possible for 1.23.2?

[mehechiger]

Further investigation update: the issue is not limited to persisted cold-start state.

I continued testing in an isolated external-drive profile; the production profile was not modified.

New findings:

1. A blank File → New Window can freeze before any folder or workspace is selected.

2. The new workspaceStorage database contains only the normal new-storage marker. No workspace database contains the model catalog, model preference, Gemini 3.6 strings, or the model sentinel. This is not a per-workspace database problem.

3. I installed a temporary SQLite trigger in the lab profile that prevented the userStatus key from being updated. Both state.vscdb and state.vscdb.backup remained on the known-good values and passed PRAGMA quick_check, but a new window still froze after the account had loaded the Gemini 3.6 catalog.

This shows that protecting the persisted database is not enough: the remotely fetched model catalog is also held in runtime/shared process state and is consumed by subsequently initialized windows.

The controlled sequence was:

• known-good userStatus and zero modelPreferences;

• offline or online but signed out: multiple windows work, including opening folders;

• sign in manually;

• Gemini 3.6 entries appear in the model selector;

• the three already initialized windows remain usable;

• create a fourth blank window;

• the fourth window freezes immediately.

I also found a separate follow-on problem. A frozen blank window is saved into storage.json as a restorable empty window. On the next launch this can cause a no-window/exit or recovery loop. Restoring only the pre-failure window state in a cloned profile makes offline startup work again.

The updated failure model is therefore:

A. Persisted catalog problem: the Gemini 3.6-era userStatus can make Antigravity 1.23.2 freeze on a later cold start. The recovery launcher fixes this part.

B. Runtime catalog problem: after sign-in loads the Gemini 3.6 catalog into the running process, any subsequently created or reinitialized editor window can freeze even if the SQLite values remain known-good.

C. Window recovery problem: the failed window may then be persisted as an empty window and retried on the next start.

The current non-destructive operating workaround is:

• disconnect the network;

• run the safe launcher;

• create/open every required window while signed out;

• reconnect and sign in;

• do not create additional windows or use Focus Editor afterward;

• do not select Gemini 3.6.

A server-side mitigation still appears possible and would be the safest immediate fix: do not return the Gemini 3.6 tiered catalog to client version 1.23.2, or return a compatibility-safe legacy catalog. The client should also avoid persisting a window as recoverable until its workbench initialization has completed.

I have preserved sanitized before/after SQLite snapshots, storage.json diffs, runtime-condition tests, hashes, and the recovery scripts. I can provide them without uploading private account or project state.

[anthony_galvez_chinc]

Same issue on Ubuntu x64 with v1.23.2.
UnresponsiveSampleError on launch in workbench.desktop.main.js (PJr function).

[abdifahadi]

Opening any project folder causes the Agent panel to crash into “Drag a view here to display” and freezes the editor with a “Window is not responding” popup.

Steps Already Tried (Tried Solutions):

1. Clean Reinstall: Fully uninstalled (using Revo Uninstaller) and cleared all %AppData%, %LocalAppData%, and .antigravity cache folders.

2. Version Switching: Tested multiple stable/older versions (v1.19.4, v1.22.2, v1.23.2) after fresh installs.

3. Auto-Update Lock: Set "update.mode": "none" in settings.json and locked the pending folder.

4. Cache & Layout Reset: Deleted workspaceStorage, cleared project-level .vscode cache, and attempted resetting view locations via Command Palette.

Expected Fix Needed:

Please resolve the Agent Sidebar View initialization conflict when mounting workspace state on project load.

[krunal_panchal]

same issue in ubuntu 24,

[hoangkimquoc]

I tried installing the latest version, Agent Antigravity 2.0. Then, clicking ‘Open IDE’ from this interface fixed the issue

[fogwildernessdrumlin]

On Ubuntu 22.04, I had v1.23.2 (apt), removed the app and all its data, installed the standalone 2.1.1 IDE instead, and now it works fine.

[Donovan]

Same issue on Windows 26200.8875
and reported to me first by another user on Windows as well
Clearing App Data temporarily resolves it

[Zoze]

ANY FIXxxxxxxxxxxx ???

[Axio_SmpProtect]

I have the same problem. Is there a solution, please?

[Sim_Zar]

Anyone was able to find a solution?