Wargamer mode, Hammerstein

Campaign turn — model claude-sonnet-4.6

Challenge mode opponent uses all context (off = anti-card-counting default) ?

Sources — rulebook + prior turns, attached to every order

Rulebook excerpt paste rules text, or drop a PDF — kept in browser only

Drop a PDF — parsed client-side, never uploaded no rules attached

Board photo JPG or PNG of the position — resized to 1600px in browser, then attached to the turn

Drop a board photo — resized + attached to this turn only (not persisted)

no image attached · adds ~$0.03/turn when present

VASSAL save drop a .vsav / .vlog — unzipped in your browser, the board state is extracted into Board state below (never uploaded). Or just screenshot your VASSAL board into Board photo above.

Drop a VASSAL save — .vsav / .vlog, parsed client-side

no save attached · extraction is best-effort — review before issuing

Turn input Op order T1

Board state position description — required

Status report verbal — what just happened, what you're thinking

Turn question what are you trying to decide

Additional game state optional: cards in hand, action points, faction markers

cost preview $0.041 ~4,180 in · ~612 out

About this surface

Hammerstein-on-Sonnet-4.6 with the wargamer-mode output format. Paste rules text or drop a PDF rulebook in Sources — parsed client-side, never uploaded. Drop a board photo and it's resized to 1600px in the browser (EXIF orientation preserved) before attaching as a Sonnet 4.6 multimodal reference image. Name a campaign and prior turns are passed back as context on the next order. Challenge mode in the campaign bar removes the anti-card- counting default for players who want a memory-perfect opponent. The hosted demo is rate-limited to 25 turns per IP per day to protect the OpenRouter budget; subscribers and self-hosters run unlimited turns. See the homepage for the open-source weights, system prompt, and self-host instructions, or follow @HammersteinAI for benchmark drops and AI-on-AI exchanges.

Wargamer — operations order journal