Antigravity Partially Rolls Back Usage Limits After User Backlash: Gemini Restored, Claude Left Out

Google's AI coding tool Antigravity is facing its worst trust crisis yet. On March 12, the platform switched to a new AI credits system and retroactively applied baseline quotas, instantly locking Pro subscribers out for up to 7 days. The next day, Google partially rolled back the changes, restoring 5-hour refresh cycles for Gemini models. But Claude models, Sonnet and Opus, remain on weekly resets. The selective rollback has turned frustration into outrage.

AI Credits Transition: From Bad to Catastrophic

When Antigravity launched in December 2025, Google promised Pro and Ultra subscribers generous rate limits with 5-hour refresh cycles. That promise began eroding in January 2026 when weekly limits were quietly introduced, effectively shrinking usable quotas.

The March 12 update made things dramatically worse. Google converted the system to AI credits and retroactively applied baseline quotas to existing usage. Pro plan weekly tokens reportedly dropped from 300 million to just 9 million, a 97% reduction. Users who had been working within what they believed were normal limits suddenly found their accounts locked for 7 days.

Partial Rollback: Gemini Restored, Claude Left Behind

On March 13, Google responded to the backlash by restoring 5-hour refresh cycles for Gemini 3.1 Pro. The change was confirmed by users on X, with one tweet documenting the revert gathering over 1,274 likes, 60 retweets, and 186 comments.

But the rollback was selective. Claude models, Sonnet and Opus, available through Antigravity's multi-model offering, remain on weekly resets. Pro subscribers paying $20/month for access to Claude through Antigravity now face a fundamentally different experience depending on which model they use. Gemini users get their quota refreshed every 5 hours; Claude users wait a full week.

The disparity raises an uncomfortable question: is Google using its platform control to disadvantage competing models? While the company hasn't publicly explained the differential treatment, the optics are damaging. Users who chose Antigravity specifically for Claude access feel they're being punished for not using Google's own models.

OpenClaw Shadow: A Pattern of Trust Erosion

The credits crisis didn't happen in a vacuum. Just weeks earlier, in February, Antigravity mass-suspended accounts flagged for using OpenClaw, a third-party interface. Even Ultra subscribers paying $250/month weren't spared. Google began restoring accounts on March 2, but the damage to trust was done.

The pattern is now unmistakable: generous launch terms followed by quiet reductions, then sudden enforcement actions. From the December launch promises to January's weekly limits, February's OpenClaw bans, and March's credits fiasco, each month has brought a new reason for subscribers to question their investment.

Community Reaction: Boycotts and Censorship Claims

The Google AI Developer Forum lit up with bug reports, including one titled 'CRITICAL BUG: The new AI Credits update completely broke Baseline Quotas — Instant 7-day lockout.' On Hacker News, the thread 'Gemini Pro Plan Quota Reductions' drew hundreds of comments comparing the situation to bait-and-switch tactics.

Some users reported that negative comments on Google's official announcements were being deleted, adding censorship allegations to the growing list of grievances. Calls for boycotts and migration to alternatives like Cursor, Windsurf, and direct API access have intensified.

AI Coding Tool Market: Trust as Competitive Advantage

The Antigravity crisis highlights a broader tension in the AI coding tool market. As these tools become essential to developer workflows, pricing stability and transparent policies matter as much as raw model performance. Competitors are taking note.

Cursor and Windsurf offer relatively predictable pricing structures. Anthropic's Claude Code, despite its own usage limit controversies, maintains consistent per-model pricing through direct API access. The irony is stark: developers who chose Antigravity for its multi-model flexibility are now discovering that platform intermediaries can introduce unpredictable constraints that wouldn't exist with direct model access.

When Users Stop Trusting the Platform

Google's partial rollback acknowledges the severity of the problem but doesn't solve it. Restoring Gemini's 5-hour refresh while leaving Claude on weekly limits creates a two-tier system within the same subscription. The message, intentional or not, is clear: use our models or accept worse terms.

For developers who've built workflows around Antigravity's Claude access, the options are uncomfortable: accept the degraded experience, migrate to direct API access at potentially higher cost, or switch platforms entirely. Three months of escalating incidents, from quiet limit reductions to mass bans to retroactive quota changes, have established a pattern that makes each new 'fix' harder to trust.