La nona never carried the dog

The tutorial opened with a joke, “no los hemos utilizado para hacer deepfakes” ▸ 0:30, and the day supplies the case study. In the family WhatsApp, a cousin posts a photo: the late nona, smiling, carrying the late family dog. It moves everyone; “qué fotaza”, someone writes ▸ 35:20. He asks when it was taken. December 2023, says the cousin, and the uncle confirms ▸ 36:51.

The debugging instinct that catches lying error messages catches this too, because the timeline won’t close: his nona died around 2019, “ella no alcanzó ni siquiera a vivir la pandemia” ▸ 38:19, while the dog wears the face of his final year. Even the physics is wrong: a dog that size, held effortlessly by an elderly woman with a perfectly straight back ▸ 41:02. He asks again the next morning, and the family confesses: the photo is AI ▸ 40:22. No malice, a nostalgic gift joining two beloved dead. And still: “uno se sentía como medio traicionado” ▸ 42:05, because the image was already installing a memory of a grandmother who lifted and hugged that dog, and “nunca pasó eso”, she was the quítenme esa vaina de encima type ▸ 42:09.

el deepfake familiar no miente sobre el presente; edita el pasado →

The generalization is the Mandela effect with a manufacturing arm: “en un futuro vamos a tener memorias falsas que vamos a creer que son reales y ni siquiera vamos a saber” ▸ 43:08. Julia, who has almost no baby photos, realizes a well-meaning relative could simply make her childhood ▸ 44:21. And they feel the temptation from inside: her father visited Disney World around 2000; his grandmother toured the US in that era. One prompt could have them meet. “Si alguien lo hace, yo me la creo”, he admits, and there’s the whole problem, followed by the epitaph for the genre: if it’s fake and you find out, “se pierde la magia” ▸ 47:02…