Pressure that isn't points

Gemini, asked to organize his September, prescribes focus: one project, the Shipaton, a mobile Sanfanson ▸ 20:16. But the session that follows is deliberately not programming: “antes de ponerme a programar quiero pensar en este principio que realmente sea el vencedor” ▸ 30:07.

Three principles surface. First, teacher memory: when Julia taught him Portuguese, she remembered exactly when she’d introduced a word and which example she’d used, “acuérdate, hace dos días vimos esta palabra” ▸ 20:48. An AI tutor can replicate that: keep a history of the words that still cost you, drop the ones you’ve mastered, and give feedback that cites your own past attempts ▸ 21:34. Second, the origin story restated: Sanfanson was born from noticing that repetition plus fast feedback is literally how neural networks learn, predict, err, adjust parameters, human brains included ▸ 25:02.

Third, the sharpest cut, aimed at Duolingo. Julia’s audit of it: no speaking, barely any listening ▸ 23:45. His: the leaderboard does create pressure, but “tienes presión por conseguir más puntos, que no significa aprender”, you farm easy exercises and climb ▸ 26:55. The pressure that works is the one that finished his Portuguese: needing to communicate in Brazil. Julia’s mechanism to fake that need, borrowed from Boardy’s phone calls: complete the week and the app calls you, a live voice conversation as the week’s exam ▸ 27:57.

diseña la presión correcta antes de diseñar los puntos →

His verdict is properly scientific: “hay que probar esa hipótesis” ▸ 30:03. But the design bar is set: gamification that rewards the metric is the disease it claims to cure; the winning principle must make the learner need the language, even for five scheduled minutes a week…