An app for almost-right answers

The mathematical-intuition conviction does what his convictions do: it turns into an app idea within days. Not an app for exact mathematics, an app for the intuition itself ▸ 3:37. The faculty he wants to train is the one that knows the neighborhood of an answer before computing it: ask 10 × 50 and something says it’s under a thousand ▸ 4:14; take your change at a counter and something says whether it’s right without arithmetic ▸ 4:41. The mechanics are already specified: questions you can’t compute in time, what’s 2 to the 10th, what’s the root of 300, answer inside ten seconds, and score by closeness to the true value ▸ 5:20. Estimation as a game loop, the vueltos-sense with a leaderboard.

la intuición se entrena con respuestas casi correctas, no exactas →

It slots perfectly into the week’s reading: the memory-paradox article argues intuition lives in the slow system that only practice writes to, and this is a practice machine for numbers. Whether it gets built lands on the familiar pile of ideas awaiting their turn, but the design constraint, ten seconds, no calculators possible, is exactly the anti-offloading rule the article prescribes.

Meanwhile the last soldier advances one bureaucratic step per day: Severo now runs on a paid API key with the card attached, because a bug traced to the free tier is unacceptable in something people will pay for ▸ 2:23, and the Play Console’s counter reads five consecutive tester-days of the fourteen Google demands, pointing at a public launch in early January ▸ 3:01. The morning also held a toy experiment, a small network guessing the gaps between prime numbers ▸ 0:56, nothing novel, he says, but the hands never stopped…