DotDotDot
點點點

New product direction

新產品方向

The same scores you already record — now you can see who's improving and what they're missing.

一樣是你本來就要登記的成績——現在能看出誰在進步、哪裡還不會。

Record one score per student, just like your Excel — but instead of a dead file, you get trends, top performers, and who needs help. Snap the graded test and we break it down question by question, compared to past tests.

照你原本的習慣，一個學生填一個分數，跟 Excel 一樣；但你拿到的不是一份死檔案， 而是趨勢、前段班、還有誰需要幫忙。想更深入，就拍下改好的考卷，我們幫你拆解每一題，還會跟過去的測驗比較。

See the plan 看時程 What the team needs 團隊需要做什麼
Enter it like you already do 照原本的習慣輸入 One score per student — just like your Excel or paper form. 一個學生一個分數，跟 Excel 或紙本一樣。
See your class come alive 讓班級活起來 Trends, top performers, who's slipping — not a dead spreadsheet. 趨勢、前段班、誰在退步，一眼看見，不再是死表格。
Snap the test to go deeper 想更深入就拍考卷 We read every question, then show what improved and what's still weak, across tests. 我們讀每一題，再告訴你進步在哪、哪裡還弱，還會跨測驗比較。

01 — Why now

01 — 為什麼現在

Why this, why now

為什麼是這個方向，為什麼是現在

We're not chasing an all-powerful AI. We're taking away one job teachers redo every single week.

我們要的不是萬能 AI，而是把老師每週都要重來一次的苦差事拿掉。

The grading's done. The numbers just have nowhere to go.
分數老師都打好了，只是沒地方去。
Phase 1 thesis 第一階段產品假設
1

Grade, then re-key the whole thing

改完，再一筆一筆重打

Teachers still hand-grade weekly paper quizzes, then key every score into a gradebook or the school system.

老師每週改完紙本考卷，還得把成績一筆一筆登記到成績簿或校務系統。

2

The more ordinary the quiz, the more it's ignored

越普通的考卷，越沒人理

Bubble machines only know multiple choice; essay AI only reads long writing. The everyday quiz in between falls back on the teacher.

答案卡只認選擇題，作文 AI 只看長文；夾在中間、老師每週都在出的那種考卷，最後還是老師自己來。

3

We've already walked part of this road

這條路，我們已經走過一段

Chat Everywhere's handwriting-grading ability is already built — we can reuse it across the new product.

Chat Everywhere 的手寫辨識和批改早就練好了，直接拿來用就行。

4

Take the lowest-risk step

挑風險最低的那一步

Reading a number the teacher already wrote is far safer than judging a whole answer — and no slip gets past the teacher.

讀一個老師已經寫好的數字，比讓 AI 判讀整張作答穩多了；出錯也逃不過老師的眼睛。

5

Data that compounds

會複利的資料

Each score makes the history more complete; over time it becomes a moat competitors can't recreate.

每記一次成績，歷程就更完整一分；時間拉長，就成了對手補不回來的護城河。

02 — Direction

02 — 方向

Product direction

產品方向

A brand-new product, built from the ground up: score management + class operations for Taiwan K-12.

全新的品牌與產品，從零打造：為台灣 K-12 設計的成績管理 + 班級經營。

What we are building

我們要做什麼

  • Photo capture from a graded stack — pull each question, answer and mark, or just the total.
  • 拍照擷取批改好的考卷——每一題、答案與對錯都擷取，或只取總分。
  • Manual score key-in as a reliable fallback.
  • 保留手動輸入成績作為可靠備援。
  • A gradebook that imports and exports Excel, so it fits the school's grade system.
  • 成績簿可以與 Excel 雙向匯入匯出，直接接上校務系統。
  • Per-student progress across tests — trends, top performers, and the M-shaped middle band.
  • 跨測驗的個人進步——趨勢、前段班，以及老師最在意的 M型中間區段。
  • Everyday classroom tools: a random name draw, points, stamps, and light routines.
  • 每天用得到的小工具：抽籤、點數、印章與輕量的班級常規。
  • Then, in Phase 2: personalized practice quizzes for students — born digital, auto-graded, progress flowing back to the teacher.
  • 再來是第二階段：給學生的個人化練習——數位原生、自動批改，進度回流給老師。

What we are not building yet

現在先不做什麼

  • Not another chatbot.
  • 不是另一個聊天機器人。
  • Not a full learning platform.
  • 不是完整的學習平台。
  • Not a student app — in Phase 1. Students enter in Phase 2, through the practice portal.
  • 第一階段不做給學生用的 App——學生要到第二階段，才會透過練習入口加入。
  • Not essay grading.
  • 不是作文批改。
  • Not parent reports for schools.
  • 不是先做給學校用的家長報告。
  • Not "scan and understand every arbitrary worksheet."
  • 不是要掃描並理解所有任意格式的講義或考卷。

03 — Timeline

03 — 時程

From prototype to launch

從雛形到上線

A prototype teachers can use in July, a working product by mid-August, then to market — Phase 3 starts after we launch.

七月做出老師就能用的雛形，八月中做出可用的產品，接著推向市場——第三階段等上線之後再展開。

Prototype 雛形

Teachers start importing grades

老師開始匯入成績

  • Class roster, manual entry, and photo score capture.
  • 班級名單、手動輸入與拍照擷取分數。
  • Gradebook with Excel import / export.
  • 成績簿，支援 Excel 雙向匯入匯出。
  • First teachers using it with real quizzes.
  • 第一批老師用真實考卷開始使用。
Semi-complete 半成品

A working product — Phase 1 and 2

可用的產品——第一、二階段

  • Per-student progress and per-question mistake-cause analysis.
  • 個人進度，加上每一題的錯因分析。
  • A random name draw, points, stamps, and the daily classroom hooks.
  • 抽籤、點數、印章與每天的班級小工具。
  • Daily use, ironing out issues overnight.
  • 每天實際使用，隔夜把問題修掉。
Launch 上線

Roll out to market

推向市場

  • Landing page and a short demo video.
  • 上線網站與一支短示範影片。
  • Free and paid tiers.
  • 免費與付費方案。
  • Influencer launch and first textbook-publisher conversations — then Phase 3 begins.
  • 教育影響者啟動與第一批書商洽談——接著展開第三階段。

04 — Roadmap

04 — 未來藍圖

Where this goes after Phase 1

第一階段之後，往哪裡走

Every later phase is just a new view over the same score data — and we only build the next one once teachers rely on the last.

後面每一階段，都是同一份成績資料的不同視角；只有當老師真的離不開上一步，我們才做下一步。

1
Now 現在 You are here 你在這裡

Phase 1 · The weekly habit

第一階段 · 每週的習慣

  • Photo score capture — per question or just the total
  • 拍照擷取成績——每一題，或只取總分
  • Manual entry + class roster
  • 手動輸入 + 班級名單
  • Gradebook with Excel import / export
  • 成績簿，支援 Excel 雙向匯入匯出
  • Per-student progress + M-shaped middle view
  • 個人進度 + M型中間區段檢視
  • Name draw, points & stamps
  • 抽籤、點數與印章

Phase 1 · The weekly habit

第一階段 · 每週的習慣

Ship gates — testable by us

出貨門檻——我們自己就能驗證

  • Zero silent errors. Benchmarked on real graded papers: every uncertain read is flagged for a one-tap confirm — effectively 100% accurate after teacher review. The bar: beat DIY ChatGPT's dropped and invented numbers.
  • 零默錯。用真實考卷實測：每個沒把握的判讀都標記、一鍵確認——經老師確認後等同 100% 正確。標準：贏過老師自己用 ChatGPT 會「數字亂給」的現況。
  • Faster than hand-keying. A full class stack captured and reviewed in under 5 minutes; each review batch under 90 seconds — stopwatch on the benchmark stack.
  • 比手 key 快。全班一疊 5 分鐘內擷取＋確認完；每批確認少於 90 秒——用碼表對基準考卷實測。
  • Excel round-trip, zero re-typing. Import → edit → export against real 校務系統 templates, with no data loss and no manual fixes.
  • Excel 雙向、零重打。用各校真實校務系統模板實測匯入→編輯→匯出，資料零遺失、零手動修補。
2
Next 接下來

Phase 2 · From scores to practice

第二階段 · 從分數到練習

  • Personalized practice quizzes, generated from each student's weak spots
  • 個人化練習——照每個學生的弱點自動出題
  • Students practice through the portal, at home on their own time — digital-native, auto-graded (printing stays an option, not the path)
  • 學生透過入口在家練習、自己的時間——數位原生、自動批改（要印出來也可以，但那不是主路）
  • Answers the way the question demands: taps where natural; 填充 and 計算 by writing — on the canvas, or on paper photographed with any phone
  • 照題目該有的方式作答：選擇用點的；填充、計算用寫的——寫在螢幕上，或寫在紙上用手機拍照都可以
  • Streaks, bonus points and stamps built in, so the practice actually gets done — with a "who finished" view for the teacher
  • 內建連續紀錄、獎勵點數與印章，讓練習真的被完成——老師一眼看到誰做完了
  • Progress and feedback flow back to the teacher — per-question mistake-cause analysis, cross-test topic trends
  • 進度與回饋回流給老師——每一題的錯因分析、跨測驗的主題趨勢

Phase 2 · From scores to practice

第二階段 · 從分數到練習

Ship gates — testable by us

出貨門檻——我們自己就能驗證

  • The photo path passes the same zero-silent-error gate. Children's handwriting, home lighting, phone cameras — benchmarked before any photo-graded answer counts.
  • 拍照路徑同樣過「零默錯」門檻。兒童筆跡、家裡光線、手機鏡頭——先過基準測試，照片批改的答案才算數。
  • No device locks a student out. Tap, canvas and paper+photo all submit from phone, computer and tablet — tested on a device matrix, low-end included.
  • 沒有學生因設備被擋在外面。點選、畫布、紙上拍照，手機、電腦、平板都能交——含低階機型的設備矩陣實測。
  • 亂回答 gets caught. Blank, scribbled or nonsense answers are flagged, not scored — over 90% detection on a seeded junk-answer set.
  • 亂回答抓得出來。空白、亂畫、亂寫會被標記，不會被當成成績——用預埋的亂答樣本實測，偵測率 >90%。
  • Personalization is traceable. Every generated question maps back to a specific weak concept in the student's record — audited by sampling, not vibes.
  • 個人化查得到帳。每一題練習都能對回學生紀錄裡的具體弱點——抽樣稽核，不是憑感覺。
  • The Monday view reads in under 1 minute. Who finished, who didn't, what's still weak — timed with real teachers.
  • 週一視圖 1 分鐘內看完。誰做完、誰沒做、哪裡還弱——和真實老師計時實測。
3
After launch 上線之後

Phase 3 · Beyond the classroom

第三階段 · 走出教室

  • Cram-school diagnostics dashboard
  • 補習班診斷儀表板
  • Parent-facing progress reports
  • 給家長的進度報告
  • Delivered via school / management-system partners
  • 透過學校／管理系統夥伴落地

Phase 3 · Beyond the classroom

第三階段 · 走出教室

Ship gates — testable by us

出貨門檻——我們自己就能驗證

  • Reports reconcile. Every number in a diagnostic or parent report traces to the underlying score record — automated audit, zero unexplainable figures.
  • 報告對得回帳。診斷與家長報告裡的每個數字，都能對回原始成績紀錄——自動稽核，零無法解釋的數字。
  • Statistically honest. Claims render only where the data supports them (2+ semesters, enough students); thin data degrades gracefully instead of overclaiming.
  • 統計上誠實。資料夠（兩學期以上、樣本足）才呈現結論；資料薄就優雅降級，不誇大。
  • 個資法-clean by audit. Consent coverage, data minimization and partner delivery (LINE / management systems) pass the compliance checklist before any report leaves the building.
  • 個資法過稽核。同意書覆蓋率、資料最小化、夥伴遞送（LINE／管理系統）全數通過檢核表，報告才出門。
+
Later 更之後

Parked R&D

暫時擱置的研發

  • Essay grading
  • 作文批改
  • Deeper class management
  • 更深入的班級經營
  • Markets beyond Taiwan
  • 台灣以外的市場

Parked R&D

暫時擱置的研發

Unlock triggers

什麼情況才解鎖

  • Essay grading — testable trigger: re-run score stability beats the trust bar (the "B version" problem, solved on a benchmark first).
  • 作文批改——可實測的門檻：重跑分數穩定、過信任標準（先在基準上解決「B 版」問題）。
  • Deeper class management — market trigger: teachers ask for it from inside the score loop.
  • 更深入的班級經營——市場門檻：老師從成績流程裡主動開口要。
  • Beyond Taiwan — market trigger: the Phase-1 habit holds for a full semester. HK first (proven demand), Indonesia next.
  • 台灣以外——市場門檻：第一階段的習慣撐過完整一個學期。先香港（需求已驗證），再印尼。

Success metrics

成功指標

With no reminder, do they still log the second one?

沒人提醒，他們還會記第二次嗎？

The roadmap cards flip to the gates we can test ourselves. These three questions are the part only the market can answer.

藍圖卡片翻過去，是我們自己就能驗證的門檻；這三個問題，只有市場能回答。

Phase 1 第一階段 Do trial teachers record a second test without being reminded? 試用老師會不會在沒人提醒的情況下，記錄第二次考試？

If teachers come back for the next quiz — and are still logging weekly after four weeks — the habit is real. If they do not, we have a workflow problem, not a marketing problem.

如果老師會為了下一張考卷回來——而且四週後還在每週記錄——習慣就是真的。如果不會，那是工作流程問題，不是行銷問題。

Phase 2 第二階段 Does assigned practice get done — and does the teacher assign it again? 派下去的練習有人做嗎——老師下週還會再派嗎？

With streaks and points on, at least 70% of assigned practice completed (structured campaigns reach 74–90%); the teacher re-assigns the next week unprompted; guardian consent completes without pushback.

開著連續獎勵與點數，完成率至少 70%（有結構的活動可達 74–90%）；老師沒人提醒就再派一次；家長同意流程順利走完、沒有反彈。

Phase 3 第三階段 Does anyone pay — and does a report ever cause an action? 有人付錢嗎——報告真的促成過一個決定嗎？

First signed revenue — a publisher bundle or a school / county purchase; one documented case of a diagnostic driving a real decision; parents open and act on reports before we build them out wide.

第一筆簽約營收——書商合作或學校／縣市採購；至少一個「診斷促成真實決定」的案例；家長真的會打開並依報告行動，我們才擴大建置。

05 — Focus

05 — 焦點

Strategic focus

策略焦點

Phase 1 is about earning teacher trust in the recurring weekly score workflow.

第一階段的目標，是讓老師願意把每週重複的成績流程交給我們。

Nail the weekly score loop first

先把每週的成績流程做好

Capture the score, cut down on re-typing, keep the teacher in charge of the final check, and make progress visible. This is the habit that unlocks everything later.

擷取分數、減少重新輸入、讓老師掌握最後的確認、讓進度看得見。這個習慣會解鎖後面所有的可能。

Resist doing too much too soon

先忍住，別一次做太多

Essays, parent reports, a full learning platform, and grading of any random worksheet can wait. They only get stronger once the weekly habit is real.

作文、家長報告、完整的學習平台、任意格式的講義批改，都可以等。等每週的習慣真正成形，這些方向才會更有力量。

06 — Team

06 — 團隊

How the team divides the work

團隊分工

Every role protects the same promise: make the weekly score loop disappear.

每個分工都在守護同一個承諾：讓每週的成績流程變輕。

Build

開發

Keep the first version narrow, fast, and easy to check. Prioritize capture, a reliable gradebook, Excel, and grading.

讓第一版聚焦、快速、容易檢查。優先做擷取、成績簿的穩定、Excel 與批改。

Field

現場

Bring real teacher workflows into the product: consented samples, interviews, Excel templates, and what actually happens after class.

把真實老師流程帶進產品：同意樣本、訪談、Excel 模板、下課後的真實情況。

Design

設計

Make it feel native to Taiwan classrooms: cute enough for daily use, clear enough for score review.

讓它像台灣教室會用的東西：可愛到願意每天打開，清楚到成績審核不害怕。

Everyone

全體

Protect the scope. We win by removing one painful loop first, not by promising every AI feature.

一起保護範疇。我們先消除一個痛苦迴圈，而不是承諾所有 AI 功能。

Closing thought

最後一句

Start small enough to be trusted. Grow from the weekly habit.

小到值得信任，從每週習慣長大。

DotDotDot starts small on purpose: capture the score, save the teacher time, and make progress visible. If we own that loop, we earn the right to build the bigger set of classroom tools later.

點點點刻意從小處開始：擷取分數、節省老師時間、讓進度看得見。只要我們掌握這個循環，之後就有資格打造更完整的課堂工具。

1
Time back first 先還老師時間 Drop one step of re-typing scores. 少一道重打成績的手續。
2
Nothing gets lost 每張考卷都留住 Every quiz stacks into a history. 分數累積成一整段歷程。
3
Stand firm, then grow 站穩再往外長 Grading and analysis come next. 完整批改與分析，後面再接。