Cleo never asks.

The film's entire architecture rests on this single refusal—she navigates the 1970 Mexico City uprising by reading micro-signals that no one explicitly teaches her.

She learns the city the way surveillance systems learn to predict behavior: through observation without annotation, pattern recognition without consent. There's a scene where she stands in a crowded street during protests, surrounded by chaos, and her stillness is an act of reading.

This is exactly how large language models ingest information—they don't ask the city for its rules, but absorb the texture of power through repetition and proximity. Cuarón shot the film in black and white partly as aesthetic choice, yes, but functionally it strips away distraction and forces you to see only contrast, hierarchy, movement through space. The same reduction that makes a city legible to an algorithm: what moves, what doesn't, what's present, what's withheld.

We've automated the watching part and eliminated the consequence of being watched.

The brilliant mistake is that Cleo's method keeps her alive because she reads without announcing it. But the surveillance systems we're building don't inherit her discretion—they inherit her hunger for pattern, and we've automated the watching part while eliminating the consequence of being watched. She knew she was in danger.