Gemini Reset TimeAntigravity public resets History 中文 Alerts

How this works

Method and sources

What we record, how records are grouped into events, how posts are verified, how the reset frequency is calculated, and where the archive runs out.

Scope

Public extra resets of the Gemini quota inside Google Antigravity, as announced by the Antigravity team. Posts by named team members count when they speak about the product. The Gemini app, the Gemini API and your plan's regular refresh are out of scope.

Record types

  • Confirmed: the team says the reset has happened.
  • Teaser / signal: the team says a reset is coming, possibly after a condition, without a time.
  • Official schedule: the team gives a time or a window.
  • Rollout started: the team says the reset is in progress.

Each record keeps the scope as stated: product, model group, plan or users.

Times

Every time is when the post was published, stored in UTC. Pages show your local time by default and UTC on request. We don't know when the reset reached any account.

An official schedule keeps the announced start and, for a window, the end. A future time or window gets a countdown. When it passes without a confirmation, the record shows as awaiting confirmation. Nothing is marked complete automatically.

Open and historical

A teaser or rollout note is shown as waiting for 7 days after the event's latest post, judged against the current time. A passed time or ended window shows as awaiting confirmation for 7 days. After that the record stays in the history but is no longer shown as waiting. An announced future time or window is always shown, even if an unrelated reset was confirmed in between.

Grouping teasers and confirmations

Records share an event ID when the posts refer to each other, or when the condition and scope match. Being close in time makes two posts a candidate, not a match. Example: on 20 Nov 2026 at 18:38 UTC, post 2123848516581892181 said a reset would follow Neon cache work; at 23:57 UTC a follow-up said a cluster migration was taking time; on 21 Nov at 09:18 UTC, post 2124070033534161407 said the Gemini quota had been reset. All three carry event ID neon-cache, teaser to confirmation took 14h 40m, and the event counts once.

Sources and verification

  • Primary: posts by named Antigravity team members; official Google Antigravity documentation.
  • Verification: X returns HTTP 403 to direct requests, so the editor checks each post's text and time through X's embedded post view and against public archives that link the original, such as resetnow.ai, AGY Resets and OpenTheRank, or a second team post. Some older records cite the AGY Resets archive index rather than a page for that event; it's used to confirm the date.
  • Not enough on their own: screenshots without a link, unattributed reposts, personal reports of quota changes, service status pages.

Summaries are short paraphrases with a link to the source, not copies.

Reset frequency

  1. Take confirmed events in the archive up to the last check, one per event ID, at the first confirmation.
  2. Leave out confirmations marked countInBaseline: false. The 21 Jul 2026 weekly reset is one: its post didn't name a model group. These stay in the history with a note.
  3. A recorded gap is the time between two consecutive baseline resets. The running gap is excluded.
  4. For H = 24 hours, 48 hours or 7 days: share = recorded gaps ≤ H ÷ all recorded gaps. A percentage is shown only with at least 5 recorded gaps.

At the 21 Nov 2026 check: 14 confirmed resets in the archive, 13 in the baseline, 12 recorded gaps. 2 of 12 were 24 hours or shorter, 2 of 12 were 48 hours or shorter, and 6 of 12 were 7 days or shorter.

The historical baseline describes how long previous gaps lasted.

Reset chance

The homepage gauge is an editorial guess for fun. We read recent posts from the monitored team accounts and link to the post behind a score. The quiet default is 15% for the next 24 hours and 35% for seven days. A hint, an explicit plan, or rollout can raise the score. We use a signal for up to 48 hours. A posted schedule takes over the display and a review older than 36 hours pauses the gauge.

Freshness

An editor checks sources and publishes by hand. The last check time is on every page. If it's more than 36 hours old, pages say the data may be stale. If the last check was before your current day began, the home page says what is on record instead of answering "not today". Feeds, the JSON endpoint and MCP change only when we publish.

Known gaps

  • The archive starts 21 May 2026 and was assembled from public posts and archives after the fact. It has recorded gaps rather than continuous coverage, and earlier resets aren't counted.
  • Teasers that never led to a reset are easy to miss, so the signal list may overstate how reliable teasers are.
  • Twelve gaps is a small sample. One unusual month moves the shares noticeably.

Corrections

Send corrections through the moderated form. Accepted corrections are listed there with their date.

Last reviewed 21 Nov 2026