Google's Gemini 4 launch marred by reports of hour-long hangs in Antigravity
op7418 · x · 2026-10-14
op7418 tried Gemini 4 Argon in Antigravity on launch night and hit severe reliability issues: thinking steps hung for thousands of seconds (screenshot shows "Thought for 2821s"), auto-approval failed with every step requiring manual confirmation, and his own agent built on the API hit repeated problems too. His verdict: Google's infra is squandering its models. Gemini 4 Argon is available via the Gemini API, the app for Pro/Ultra users, and Antigravity/AI Studio for AI Ultra subscribers.
More from coding & agent
- Vercel designer adds multiplayer to his AI-built truck app, asks people to race him — tomjohndesign · 2026-10-14
- GitHub Copilot App v1.1.29 ships 52 updates, drops clone requirement for PR links — DanWahlin · 2026-10-14
- JPMorgan Blocks External AI Agents Like Muse and Instinct From Accessing Accounts — geoffwolfe · 2026-10-14
- Open-source ntm orchestrates Claude, Codex and Gemini agents across tmux panes — doodlestein · 2026-10-14
- Local inference tip: explicitly tell models to spend tokens freely or they'll hold back — natesiggard · 2026-10-14
- Why three frontier models with different training beats one: doodlestein's case for Argon as the third mind — doodlestein · 2026-10-14