The temperature of feedback

The tester recruitment goes exactly as human nature predicts: he writes to friends asking for Severo feedback ▸ 27:21 and “básicamente todos me sacaron el culo” ▸ 27:28. One is in Malaysia where the app apparently won’t work ▸ 27:31; one broke his phone, and his backup phone, and spent all of December phoneless ▸ 27:54; one only has an iPhone ▸ 28:09. Exactly one friend actually installed it ▸ 28:21, and his surface-level pass still paid: English strings leaking into Spanish mode, which he’ll fix with emojis rather than maintain translations ▸ 28:51; voice feedback sometimes blessing wrong answers, a model problem he’s betting time will solve ▸ 29:16.

The feature requests get triaged by first principles. Replaying your audio before sending, rejected: that instinct is for human listeners, and an AI “no lo va a juzgar” ▸ 30:24. Canceling a recording midway, WhatsApp style, accepted ▸ 30:32. And the friend studying Chinese finding it beginner-hostile surfaces the real gap: “eso sí es algo 100% que le hace falta a Severo”, a mini tutorial ▸ 31:52.

“está interesante” es una temperatura, no una opinión →

Then the Roughly prototype faces its first outside eyes and returns “está interesante” ▸ 41:00 plus the colder “no sé a quién le gustaría esto” ▸ 41:05. What’s notable is his refusal to launder it: “fue muy tibio” ▸ 41:22, he didn’t say he disliked it only because he’s my friend ▸ 41:27, “no le disgustó, pero tampoco le gustó” ▸ 41:34. His remaining hope is scoped correctly too: maybe the real MVP reads better than a bare Canvas demo ▸ 41:48. Reading lukewarm as lukewarm is the cheapest market research there is…