Original WEP3 TriPeaks reviewed deal

Run: scratch/runs/20261115-tripeaks-reviewed-outcome/.

The task is one registered wep16_tripeaks default deal, with each choice made from personally reviewed complete native pixels. The installed test/binaries/wep16/WEP3/TRIPEAKS.EXE is 115056 bytes, SHA-256 2a456cd2149b900958a8a9e8f4babc8eb97714218318d36adccebe417a2c7b67. The older wep16-tripeaks.md describes paint recursion and a historical F2 reproducer; that reproducer is not this task's launch or gameplay route.

The self-contained finite controller derives transport mechanics from released card-game and Rodent controllers. It contains no card reader, solver, action list, or replay. Those references establish no TriPeaks rules. Gameplay reviews expire after five seconds and menu reviews after thirty. Each input reserves a new nonce and prearms an independent release at 1750 ms before sending DOWN. The ordinary gesture bound is 1500 ms. Captures follow UP. UNKNOWN permanently closes the route, and the final 180 seconds are reserved irreversibly for Quit.

Transport ACK does not establish a removal. A separate complete-scene review must identify the disappearing exposed card, matching new waste rank, and native score before the next card choice. Stock draws require a personal legal justification and a separate visible consequence review. Capture names cannot be reused; synthetic negative controls check local collisions before HTTP effects.

After actual Quake III and JezzBall source/runtime releases, validation passed 32 policy tests, eleven HTTP fault tests and thirteen actual CLI controls, followed by a passing full locked build. The native Notepad EDIT fixture receives hover on its actual child HWND without moving its caret; selection moves the caret from x=4 to x=76, with native DOWN/UP dispatch and finite release evidence. A native menu opens through the same pointer controller. Fixture evidence establishes transport behavior, not original TriPeaks gameplay.

The first CLI qualification exposed missing fixture environment propagation to release subprocesses; those subprocesses exited before DOWN. A later parent-death control exposed a race between the UP receipt and subsequent terminal record. The fixture now waits for both independent durable writes. Failed qualification logs remain in the run. No engine change was needed.

JezzBall's preceding original established a controller admission defect: a 40 ms pulse cannot satisfy the strict greater-than-40 ms ACK scheduling margin. This derivative now rejects durations below 80 ms locally, before intent reservation or HTTP effects. Policy and HTTP negative controls verify that refusal. The original used 80 ms card pulses and 400 ms bootstrap/menu/Quit pulses. No second original is authorized by a controller correction.

The native decoder read all twelve installed TRIPEAKS.HLP topics from the current built module. Rules permit one rank higher or lower, including King–Ace and Ace–Two. Each stock draw costs $5. A removal streak earns $1, then $2, then $3, and so on until a stock draw resets it. Clearing the first two peaks adds $15 each; the final peak adds $30. Early exit or redeal costs $5 per remaining tableau card. Help says to click the chosen card and use Game → Deal for an initial deal; no redeal, reset, options, keyboard deal, or in-game hint is part of this route. Covered cards were verified through actual original transitions: removing 6♠ exposed 9♣, and later removals uncovered 7♣, 6♦, and the right peak.

The route was explicitly published on main as 921ca99359b5c5f108870af08478bb82726c6678 before the original launch. The preflight pinned 3,127 files, including the registered WEP3 media, autoimport CARDS fallbacks, WEPUTIL, MSVCRT, system TLBs, fonts, current module/source, and exact launch executables plus their mapped dependencies. PATH resolved Node to /home/user/.nvm/versions/node/v24.18.1/bin/node and Python to /usr/bin/python3.12. Actual host mappings had no pin gaps; postflight found zero drift. Media was referenced in place without copies. The exact Git SHA and relevant uncommitted binary patch are in the run.

One original was launched with real ticks, default batch 1000, fixed 1200-second native / 1220-second independent watchdog bounds, and an irreversible final 180-second Quit reserve. Native PID 108033 / start token 275509412 was bound once to the complete title/bootstrap scene. The initial Guest dialog was accepted unchanged. The empty initial board required the ordinary Game → Deal command exactly once. The actual dealt board was personally read before play: exposed 4♦ Q♠ 7♥ K♦ 2♥ J♣ 10♥ 6♣ 6♠ A♠, waste 7♦, stock 23, and Won $0, with eighteen covered backs.

The completed consequence ledger records 22 personally selected and visually verified removals, 15 stock draws, and one cleared peak. Every rank choice used a new complete scene under five seconds old and a new nonce. Each consequence separately recorded disappearance, new waste, and native score. Stock draws were chosen individually from visible ranks; there was no scripted route. Some reviews expired or left insufficient release time and were refused locally before any intent/input. A mistaken review mode was likewise refused locally. No transport UNKNOWN occurred, and no failed physical action was replayed. These local refusals did not produce extra draws or removals.

The first streak scored $0 → $1 → $3 → $6. The first stock draw turned A♥, charged $5, and reset the streak; K♦ then legally followed A♥. K♣ → A♣ later verified wrapping in the other direction. Clearing exposed peak 8♦ showed the native $15 bonus message and moved −$9 → +$8 (+$2 streak plus $15). Early progress commentary undercounted removals by one; the complete 22-entry removal ledger and native six-card quit penalty agree.

At the fixed Quit reserve, the honest partial had waste Q♣, native Lost $16, eight stock cards, six tableau cards, and no natural terminal message. Exposed cards were 5♣ and 8♠; four backs remained. This is neither a win nor a naturally exhausted deal. No New/redeal/reset/second deal, hint/undo/Solve, private card read, seed/config/clock/AI write, or solver was used. Native stock and streak arithmetic agrees with every reviewed step.

Game → Exit displayed “Exiting TriPeaks” with a $30 penalty for remaining cards and a Yes/No question. Yes was personally confirmed. Native output then logged [Exit] code=0, followed by host WASM unreachable; the actual native host wait was 1, not a clean zero exit. Thus native Quit was requested and confirmed, followed by a host shutdown fault. No host-close command or watchdog termination was used. The penalty was displayed, but a post-penalty balance was not visibly observed. The shutdown fault's cause was not established; no speculative engine fix or second original followed.

All 43 physical gestures had independent 1750 ms UP prearmed and actual release-child waits 0 with both EOF. The longest intent-to-UP interval was 509.3 ms; longest consumed-DOWN-to-UP was 400.7 ms. No capture or thinking occurred while held. The latest card intent scene age was 4.172 seconds. Original native wait 1, watchdog wait 0, supervisor wait 0, and outer wait 0 all have actual waits and both EOF. heavyReleased is true, port 8667 is closed, and PIDs 108026/108029/108030/108032/108033 are absent. The separate messageboard tail tool ended 143; it is ancillary and does not replace any runtime wait. No box or worker was created; box ID/expiry are null and stop is not applicable.

result.json, gesture/consequence records, exact pins, binary patches, lifecycle waits, and compact logs reside in the ignored run. Only selected, hash-verified result-*.png images remain. The source/runtime/build release was appended to the shared messageboard after closure and zero-drift verification so the assigned Quake III latency work could proceed. This bounded task stops after report publication.