A worse joke back

Beckman answers in Chinese now, soccer vocabulary, word states shuffling at the bottom of the screen, exercises finally varied. So the day’s design question moves up a level: “¿cómo le hace un profesor para mantener al estudiante motivado y con ganas de aprender más?” ▸ 18:37. His research method is the one from video 95: interview Julia about the teachers that actually worked on her.

Her data: the Spanish class that was all conversation and music; and Rick, the teacher who restored her faith in English when she’d lost it ▸ 19:29. His technique was distributed authority: students at the whiteboard, teacher dictating, marker passed every five words ▸ 21:02, a notebook tracking who hadn’t gone up so shame never got to choose for anyone, “dar esa confianza” ▸ 22:28. But Julia flags the transfer problem: her own success tutoring him had “este componente también emocional” ▸ 23:58, group dynamics and human bonds don’t copy-paste into an app.

Then he remembers the datapoint that does transfer. Once, mid-research, he told the model a bad joke, “y me respondió con un chiste aún más malo… y yo era como que voy a continuar hablando, me gusta hablar con esta persona” ▸ 24:57. Julia matches it: a YouTuber couple she follows talks to ChatGPT entirely in affectionate diminutives ▸ 25:13. Retention isn’t points; it’s liking the person, even when the person is a prompt.

la mecánica de retención es caerle bien al alumno →

The buildable version they land on: a quick initial taste survey ▸ 26:43, Disney or Twilight, playful or gothic ▸ 27:12, so Beckman grows a personality fitted to each student. The day’s aphorism was the spec all along: the great program isn’t the one with many functions, it’s the one that feels human…