LAST CALL
Seven night shifts. Twenty emergency calls. One AI that handles them better than any human. Your job: approve its decisions. Override when it fails. Justify every intervention. But the more you correct ASHWOOD, the faster it learns. And by shift seven — it stops asking for permission.
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.
