最後一行是重點。GPU 買得到，「一個六歲小孩覺得數字很有趣」買不到——那要有人坐在旁邊，陪他把 7 跟 5 撥出來。

That last row is the point. You can buy a GPU. You cannot buy a six-year-old finding numbers interesting — that takes someone sitting beside them, moving 7 and 5 into place.

大人容易把「算得快」當成目標。但四五歲的孩子卡住的地方，其實在更前面：數字對他來說還只是一個念出來的音。「七」是聲音，不是一個量。所以他只能用手指——手指是他唯一看得見的量。

問題是手指只有十根，而且數到一半會忘。這不是笨，是工作記憶被數數本身吃光了，沒有餘力去想加法。

Adults tend to aim at "calculate faster." But where four- and five-year-olds get stuck is earlier than that: a number is still just a sound they say out loud. "Seven" is a noise, not a quantity. So they use fingers — the only quantity they can see.

Trouble is, there are ten fingers and they lose count halfway. That isn't slowness; it's working memory entirely consumed by the counting itself, leaving nothing for the addition.

算盤在做的事：把「量」變成看得見的東西

What an abacus does: it makes quantity visible

一四珠算盤上，上面那顆珠子是 5，下面四顆各是 1。「七」不再是一個音，而是一個看得到的形狀：上面撥下來一顆、下面撥上去兩顆。孩子不用數，他用看的。

更關鍵的是「湊十」。8＋5 為什麼難？因為要跨過十。在算盤上，這件事變成一個具體動作：5 不夠位置放，就跟 8 借 2 湊成 10，剩下 3 放隔壁。補數不再是背起來的規則，是手上做過的事。

On a soroban the upper bead is 5 and each of the four lower beads is 1. "Seven" stops being a sound and becomes a shape you can see: one bead down from the top, two up from the bottom. The child doesn't count — they look.

More important is crossing ten. Why is 8+5 hard? Because it crosses. On the abacus that becomes a physical move: 5 doesn't fit, so borrow 2 from the 8 to make 10, and the remaining 3 goes next door. Complements stop being a memorised rule and become something the hands have done.

珠心算的宣傳很容易寫成「開發右腦」「提升專注力」「變聰明」。我不會這樣寫，因為我自己也是家長，看到那種文案只會警覺。

比較誠實的說法是三件事：孩子會把數字拆開再組合（7 是 5＋2，13 是 10＋3）、會對「大概多少」有感覺（看到一盤珠子知道比另一盤多）、以及練到後來，腦中真的會浮現那個算盤的畫面。這些是可觀察的，我在自己小孩身上看得到。

但「珠心算會讓數學變好」我不敢保證——那牽涉到能不能遷移到應用題、幾何、邏輯，那是另一回事。也不是每個孩子都吃這一套：有的孩子就是喜歡數手指、喜歡畫圖，那也完全沒問題。工具不對就換工具，別換孩子。

Abacus marketing slides easily into "develops the right brain," "improves focus," "makes kids smarter." I won't write that — I'm a parent too, and that kind of copy only makes me suspicious.

The honest version is three things: children start decomposing and recombining numbers (7 is 5+2; 13 is 10+3), they develop a feel for roughly how much (one tray of beads is visibly more than another), and with practice the image of the abacus really does appear in their head. Those are observable — I can see them in my own child.

But I won't promise it "makes them good at maths" — that depends on transfer to word problems, geometry and reasoning, which is a separate question. Nor does it suit every child: some just prefer fingers, or drawing. That's completely fine. Change the tool, not the child.

一、先玩量，不要急著寫數字

1. Play with quantity before writing numerals

吃飯時問「你這盤比較多還是我這盤？」、分餅乾時問「要怎麼分才一樣多？」。這些不用紙筆，但正在蓋的是後面所有計算的地基。

At dinner: "does your plate have more, or mine?" Sharing biscuits: "how do we split them so they're the same?" No paper needed — this is the foundation everything later sits on.

二、不要比速度，尤其不要跟別人比

2. Don't race — especially not against other kids

算得慢不代表不會，往往只代表他還在腦中把珠子排好。你一催，他就會改用「猜」的來換取速度——那才是真的學壞了。

💡 把「算對幾題」換成「今天有沒有自己想出來」。前者會逼他猜，後者會逼他想。

Slow doesn't mean unable; usually it means the beads are still being arranged in their head. Rush them and they'll switch to guessing to buy speed — that's the habit that actually damages learning.

💡 Swap "how many did you get right" for "did you work one out by yourself today." The first rewards guessing; the second rewards thinking.

三、每天五分鐘，勝過週末一小時

3. Five minutes daily beats an hour at the weekend

這件事跟訓練模型意外地像：小批次、高頻率，比一次灌一大堆有效。五分鐘的重點不是練完幾題，是讓「碰數字」變成一件不需要下決心的日常。

This is oddly like training a model: small batches at high frequency beat one giant dump. The point of five minutes isn't how many problems get done — it's making "touching numbers" an everyday thing that needs no decision.

實體算盤很好，但家裡不一定有，而且沒有人示範的時候，孩子撥完不知道對不對。我想要的是一個「隨時能撥、撥錯會告訴你、而且不會逼你」的東西。

所以 LetAbacus 是這樣的：珠子可以多點觸控真的撥、有「啪噠」聲和觸覺回饋；課程幾乎沒有文字，用發光的珠子示範，孩子看完自己做；答錯不會出現紅色叉叉，只會說「再試一次」。還有數感島先玩量、看算聽算練心算、9 款珠子皮膚用練習解鎖而不是用錢。

它完全免費、零廣告、不收集任何資料，學習內容完全離線——不是為了行銷才這樣寫，是因為使用者是小孩，我不想在裡面放任何需要跟家長解釋的東西。

A real abacus is great, but not every home has one — and without someone demonstrating, a child can't tell whether what they just did was right. I wanted something they could pick up any time, that would tell them when a move was off, and that wouldn't push.

So LetAbacus works like this: multi-touch beads that really slide, with a click and haptics; lessons with almost no text, demonstrated by glowing beads for the child to copy; and no red X on a wrong answer — just "try again." Plus Number Sense Island for quantity first, visual and listening drills for mental arithmetic, and nine bead skins unlocked by practising, not by paying.

It is completely free, has zero ads, collects no data, and the learning content works fully offline — not as marketing, but because the users are children and I didn't want anything in there that would need explaining to a parent.