Gcadas Instant

"Lena." She didn't look up. "Bunny is sad."

It was a Tuesday when the GCADAS alert pinged on every screen in the Northern Sector. Not a red alert—those were for splinter storms or hive emergence. This was amber. Curious. Unsettling.

"Neither are you," I replied. But I wasn't entirely sure that was true. Some math leaves scars. gcadas

When I arrived at the Nexus Dome, the senior analyst, a woman named Dr. Ines Chu who hadn't slept in three days, threw a data slab at me. On it was a single string of code: GCADAS-CORE: PREDICTIVE CONFLUENCE @ 98.7% // ENTITY DESIGNATE: "THE SAD MATH"

My stomach turned. I activated my slate, ran a quick identity trace. Lena's mother, Mira, had left two weeks ago. Standard financial abandonment case. But GCADAS had flagged something else: Mira hadn't just left. She had left exactly according to a predictive model she'd downloaded onto a cheap heuristic doll—a model that calculated the optimal moment to minimize emotional damage to herself, not Lena. This was amber

Lena pressed a hidden switch on the rabbit's back. The speaker crackled. Then a voice emerged—not a child's voice, not a toy's. It was the sound of a thousand people sighing at once.

I suited up. My job wasn't to destroy the anomaly. You can't destroy a mathematical proof. My job was to arbitrate —to enter the logic, speak to its internal consistency, and convince it to resolve itself. Failing that, I'd negotiate terms of containment. "Neither are you," I replied

Ines zoomed out. The map showed a single residential unit in the Gray Boroughs, a low-rent district where reality had always been a little thin. "It's not a person. It's a child's toy. An old heuristic doll—pre-Fall tech. Someone reprogrammed it. Badly."

GCADAS ticked down from 98.7% to 0.2%.

"Can Bunny talk?"