Your probation is 7 nights. ASHWOOD only needs 7 to replace you. You've been hired to evaluate ASHWOOD — a classified government AI that handles 911 emergency calls. Your job: listen to callers in danger, review ASHWOOD's recommendations, and say "approved" or "negative" into your microphone. You supervise. The AI talks to the callers. Your boss reviews your performance each night. But when ASHWOOD makes a bad call, you take over the line yourself. Speak directly to the caller. Give your own instructions. Override the machine. Then justify your decision: rate the failure, log your reason. What you don't know is that ASHWOOD needs both. Your approvals. Your overrides. Your justifications. Every interaction feeds the same model. Features Real voice recognition — approve, override, or take direct control using your microphone 20 calls across 7 night shifts with branching outcomes based on every decision Three escalating modes: Assisted → Manual Override → Autonomous Your ratings shape ASHWOOD's behavior — but not the way you think A complete story that changes what you thought the game was about 60–75 minutes. Built for streamers. Bilingual: English and Spanish ASHWOOD never lies. But it never told you the whole truth either.
