Historical baseline
How often Antigravity Gemini resets have come
A count of how long the recorded gaps between confirmed public resets have been. It's the baseline any forecast would have to beat, not a forecast itself.
Reset frequency by gap length
Recorded gaps between baseline resets that lasted 24 hours or less
2 of 12 recorded gaps
17%
2 of 12 recorded gaps between baseline resets were 24 hours or shorter.
Historical baseline from our archive. It isn't calibrated or backtested, and it doesn't predict the next reset date or anything about your account.
Archive: 21 May 2026 – 21 Nov 2026 (UTC), assembled from public posts and archives rather than continuous monitoring.
| Gap length | Recorded gaps ≤ length | Share |
|---|---|---|
| 24 hours | 2 of 12 | 17% |
| 48 hours | 2 of 12 | 17% |
| 7 days | 6 of 12 | 50% |
- Confirmed resets in the archive
- 14
- Counted in the baseline
- 13
- Excluded for unclear scope
- 1
- Recorded gaps (n)
- 12
- Since last baseline reset
- 7h 49m
- Median gap
- 7d 13h 33m
- Shortest gap
- 20h 43m
- Longest gap
- 75d 22h 11m
Formula
Take confirmed events in the archive, up to the last check. Each event ID counts once, at its first public confirmation.
Leave out confirmations marked as outside the baseline, such as the 21 Jul 2026 weekly reset whose post didn't name a model group. They stay in the history with a note.
A recorded gap is the time between two consecutive baseline resets. The gap that is still running is left out.
For a gap length H of 24 hours, 48 hours or 7 days:
share = recorded gaps of length ≤ H ÷ all recorded gaps
A percentage appears only once there are at least 5 recorded gaps.
What is left out
- Teasers, official schedules and rollout notes. They describe the same events, so counting them would double up.
- Extra confirmation posts for an event already confirmed.
- Service status, community chatter and personal reports. They are context only.
How to read it
The archive starts 21 May 2026 and was assembled from public posts and archives, so it has recorded gaps rather than continuous coverage. The shares describe how long past gaps were. They aren't calibrated or backtested, they ignore the time already elapsed since the last reset, and they don't name a date. Resets are decisions made by people, so rely on announcements for timing: an announced time or window appears on the home page with a countdown.
Full definitions are in Method and sources.