Verified cards
35
Named house voice, quoted, with the permalink and the capture on file.
Observatory
The Observatory is where we look at the book as a book. Counts are descriptive. SAMPLE rows are excluded. Gaps are not a next-reset estimate.
Verified cards
35
Named house voice, quoted, with the permalink and the capture on file.
Houses on the masthead
5
Codex, Claude, Grok, Cursor, Gemini — unequal bureaus.
Bureaus modelled
3
A hazard model with its calibration published. Confidence is Low today.
The Observatory counts. It does not estimate. The estimating happens on one page and one page only, where the hazard model prints its parameters, its hour and weekday profiles, and the Brier score it earned against a flat base rate. If the model has no edge, that page says so in the same size type as the percentage.
Read the forecast and its method →
Mar
0
Apr
0
May
3
Jun
5
Jul
9
Aug
10
Sep
8
Codex
Named hard resets, banked grants, milestone flushes — Irregular. Historically quiet 01:00–08:00 PT.
24 cards
Claude Code
Weekly Max resets, policy memos, temporary multipliers — Sparser than Codex. Often a named plan, not 'everyone'.
8 cards
Grok Bot
Global token flushes announced on X — Tied to model drops. Account weekly pool is separate.
3 cards
Cursor
Confirms when a partner flush hits Grok Bot meters — No independent 'Cursor reset' ritual. Partner-led.
0 cards
Gemini
Quota incidents more than goodwill flushes — No public Tibo equivalent. Treat whispers as SAMPLE.
0 cards