Google Antigravity forum fills with quota, stability, and update bugs

Developers using Google Antigravity are reporting widespread quota anomalies, service crashes, and regressions after recent updates, clogging the official forum with urgent bug reports and workflow-breaking issues.

Developers working with Google Antigravity are flooding the official Google Artificial Intelligence Developers Forum with reports of severe reliability and quota issues. Multiple threads describe Google Artificial Intelligence Pro subscriptions showing a “99-Hour Weekly Reset Instead of 5-Hour Cycle” and a “Pro Plan Quota Bug – Models Locked for 74 Hours Instead of 5 Hour Refresh,” where the promised refresh windows are not honored. Other users report that “Gemini Pro – Rest Said 5 hours, but when it reset it only reset to 20% and now says 45 hours” and that “Claude Quota Cooldown Much Longer Than Advertised,” suggesting systemic problems in how usage limits and cooldowns are being applied across different models.

Forum activity highlights a wave of regressions tied to recent Antigravity updates and tooling. Several posts flag critical stability failures, including a “BUG: Antigravity Language Server crashing (ECONNREFUSED) due to empty project path / backend routing error” and a separate issue where the “Language server crashes with SIGILL on CPUs without AES-NI support.” Users upgrading Antigravity from “1.19.5-1772095947 => 1.19.6-1772152296” and to “v1.19.6” and later to “v1.20.4” report crashes, configuration breakage, and quota sync errors, while one thread titled “Antigravity v1.20.4 Update Bug” points to continuing instability in the latest release. The Antigravity CLI is also affected, with “Antigravity CLI (agy) on WSL: broken launcher & missing scripts” forcing manual repair, and there are multiple complaints of agents being terminated with an “Agent terminated due to error!” message.

Service reliability and account management problems are another recurring theme. Several users say “Our servers are experiencing high traffic right now, please try again in a minute,” suggesting repeated capacity or routing bottlenecks, and others mention “Antigravity Service: HTTP 500 Internal Error after Email Migration to Gmail” and “Antigravity ‘Invalid project resource name projects/’ – HTTP 400 Bad Request” as blocking errors. A number of posts focus on Gemini 3.1 and Google Artificial Intelligence Pro, including “[URGENT] AI Pro Subscriber – Abnormal 45-hour Quota Lockout on Gemini 3.1 Pro and Multiple Models,” “Unacceptable Antigravity Quotas for Gemini 3.1 Pro – Workflow Completely Blocked!” with “58” replies and “10065” views, and “[BUG] Google AI Pro – 5-day Gemini lockout since Gemini 3.1 update, very low usage, Pro 5-hour refresh not honored.” Other threads touch on missing history, broken customizations after the “1.20.3” update, Linux Mint profile loss after a disk full event, LaTeX output that “does not render well,” and questions about whether Antigravity agents waste tokens, together painting a picture of a platform whose quotas, tooling, and reliability are under heavy strain.

57

Impact Score

Leap 71s noyron model targets automated engineering design

Leap 71’s noyron system serves as a foundational computational model that encodes expert engineering knowledge, physics and manufacturing rules to automatically generate and evaluate designs. It underpins specialized models for rockets, electromagnetic systems and heat exchangers while continuously improving through feedback from real-world use.

Apple plans MacBook Ultra with OLED touchscreen and dynamic island

Apple is preparing a new high-end MacBook, potentially called MacBook Ultra, that introduces an OLED touchscreen and a dynamic island while sitting above the latest M5-based MacBook Pro models. The device marks a major shift in Apple’s stance on touchscreens in laptops as it seeks to stay competitive in a changing market.

Fujitsu debuts ‘Monaka’ Armv9 CPU sample with 3.5D packaging

Fujitsu has unveiled early silicon and an engineering sample of its ‘Monaka’ Armv9 CPU, built on TSMC’s 2 nm node and Broadcom’s 3.5D XDSiP packaging, ahead of a planned 2027 launch. The 144 core design targets Artificial Intelligence inference, simulation, and large scale data processing workloads.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.