The clenching in my jaw was relentless, a dull ache that wouldn't abate. The news anchor’s voice droned in the background, but I wasn't hearing the words. The wind, a constant, biting presence, was rattling the windows, and the rhythmic *thump* of the branches against the siding set my teeth on end. I had been attempting to read, but the words blurred, swimming before my eyes.

The crash, a deafening explosion, sent a jolt of pure adrenaline through me. I slammed the book shut, my heart leaping into my throat. Across the street, Mrs. Gable's ancient, twisted apple tree had succumbed.

The immediate impulse was to run, to escape the suffocating feeling that had been brewing all day. I grabbed my keys, my hands slick with sweat. I imagined the arguments, the accusations, the inevitable blame game. I found myself staring at the closed front door, as if at a barrier that could not be breached.

I wanted to scream, to lash out, to rip something apart. I forced myself to sit, to focus on the feeling of the worn fabric of the chair. My breath came in shallow, ragged gasps. The silence after the fall was nearly as unsettling as the crash.

Emotion: on edge

Cluster: Fear / Anxiety
PC1 (Valence): -1.77 Negative
PC2 (Disposition): -0.35

Role in Research

This story is one of 1,000 stories generated for the emotion on edge. During extraction, it was fed through Gemma4-31B and its hidden state activations were captured at 11 layers.

The mean activation across all 1,000 on edge stories, after denoising with neutral dialogue baselines, produces the on edge emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the on edge vector is projected through the unembedding matrix.

Promoted:
😖0.228
😣0.224
worse0.223
urges0.216
느껴0.214
Suppressed:
de-0.514
la-0.391
a-0.366
B-0.332
L-0.283