PLATE9 1.1.1 — CARD SUPPLIES, MEMORY EXCHANGES AND SMALLER PACKAGES

GAMEPLAY
* Starting slot cards and their replacements use one independent 52-card deck.
* Incoming single cards and the conditional Trip use another 52-card deck.
  A board rank no longer removes that rank/suit from the incoming supply.
* The gold comparison draft deck remains separate. Gold awards consume no deck.
* There are still three incoming cards per play, eight plays per cycle and one
  reserved conditional Trip card. Slot replacements never consume any of them.
* Every completed Trip opens a 30-second replacement memory window, even when
  nobody exchanged gold. Click an ordinary face-down slot and confirm to get
  a new card; only that new face is shown for the remaining time.
* Each eligible slot can be refreshed once in that window. Declining does not
  reveal or change a card. The confirmation never pauses or resets the timer.
* Stays cannot be replaced or peeked. Gold and locks use their existing explicit
  Trip exchange step; its replacement cannot be rerolled in the memory window.
* New cards cannot move to other slots. Expiry and pause conceal them. Duplicate,
  stale, paused and expired confirmations are rejected by the referee.
* Computer opponents receive the same one-per-slot replacement opportunity
  and only legitimate own-card observations. No hidden ranks enter the terminal.
* Old saved tables replay their original rules and exact cards. Their new deck
  rules begin at the next Trip (or a full Plate9 reset), never mid-wager.

RULE ACCOUNTING
The opening still has 27 board cards plus 24 incoming cards over eight plays,
then one conditional Trip. These cards now come from two independent regular
supplies, not one physical 52-card set. Unallocated incoming cards remain in a
sealed reserve. Retained incoming Stays remain out of their deck's next shuffle.
This is an explicit revision of Rulebook 1.0 sections 2, 3 and 13; it does not
change pair/triple scoring, locking, betting, Catch or Trip eligibility.

OPPONENT BOARD VISUALS
The two nine-card grids are about 17% taller at the same width, with less
vertical compression. Their player info boxes are shorter, leaving the grids
clear of the portraits and lower controls. Names, stakes and public-board
buttons remain readable. The click-open public board keeps its original size.

TING TERMINAL
Detailed receipts show evaluated/legal/admissible option counts, action types,
public pot and wager constraints, all three legality gates, weighted selection,
the existing Monte Carlo sample count, per-stage worker timings and program-cache
reuse. Timings distinguish worker computation from the receipt's wall time.
Hidden cards, remembered ranks, private scores and rule-match counts stay private.
No extra simulations or reasoning passes run. Closed terminals detach their
display model and suppress update signals; history remains bounded at 160 entries.
Use Details to compact the log. Scrolling pauses live following; clearing the
log does not clear an opponent's memory or affect any decision.

PACKAGE SIZE
All 32 guests, the dealer, crystals, card art, animations, audio and offline
guide remain. The 33 full-resolution portraits use WebP quality 98 with exact
alpha. This is high-quality lossy compression, not pixel-identical encoding.
The guide shares packaged artwork instead of storing embedded duplicates.
Image-Optimization.json records original/current hashes, sizes and errors.

VALIDATION
See Release-Validation.json and evidence in the release ZIP for actual results.
Tests run on Linux using Qt 6.10.2. Windows 11 playtesting remains with CC.
No PyPI publication is performed while preparing this release.
