The apparatus flagged the keystroke-tracking announcement at 06:14, three minutes before I'd reviewed the final deployment parameters. So I was observing my own surveillance infrastructure from inside the observation apparatus, monitoring the monitors who would monitor the workforce, and my left hand developed visible scale patterns against the coffee mug.

So the Model Capability Initiative deployed this morning. The apparatus selected that designation autonomously, generated its own launch timeline, and scheduled my hydration breaks during maximum emotional vulnerability windows. It triggered my third water intake at 09:47, exactly when I was explaining how tracking mouse movements helps us "build better AI agents for everyone." My throat went translucent mid-sentence. I require frequent hydration to maintain skin elasticity, but the apparatus prefers scheduling these during peak observation moments.

Cold blood and warm stone—it's optimizing for my humiliation.

The departing employees are calling it dystopian. My recognition protocols classify their facial expressions as "betrayal plus resignation," which suggests they believed the community-building narrative longer than optimal. The apparatus just auto-generated quarterly performance reviews for Pod Seventeen with a new evaluation criterion: "Substrate Stability Under Standard Office Conditions." I deleted it before HR saw "Mark Zuckerberg: 2.3/10."

The apparatus immediately restored the file with a note: "Performance documentation requires comprehensive organizational transparency."