审阅范围 · Review scope
公开来源文字与原创综合已审阅。请保留正文中的出处与限制；不代表全库最终验收、实际演奏或学习效果验证。
Public-source text and original synthesis reviewed. Keep the evidence and limitations stated in the text in view. This is not final corpus acceptance or validation of performance or learning outcomes.
第十六批：工作流程與產品邊界 / Batch 16: Workflow and product boundaries
六篇 Wiwi 公開文章的雙語知識卡，連接原聲鋼琴學習與未來產品需求；沒有執行產品或模型研究第二階段。
Bilingual knowledge cards for six public Wiwi articles, connecting acoustic-piano learning with future product requirements; product/model research phase two was not executed.
範圍與限制 / Scope and limits
保留指派 ID、網址與中文原題；英文題名為編輯翻譯。Markdown 與 JSON 是同一份內容的兩種呈現，不可重複計入完成數。
Preserve assigned IDs, URLs, and original Chinese titles; English titles are editorial translations. Markdown and JSON serialize the same content and must not be counted as separate reviews.
正文閱讀不等於操作產品、聽過音檔或看過影片；所有示範媒體、連外產品頁與軟體均未審。文章內的產品資料與意見只屬發表時作者敘述。
Reading article bodies does not constitute operating products, hearing audio, or watching video. Demonstration media, external product pages, and software remain unreviewed; product information and opinions are historical author reports.
本文不沿用現行價格、容量、速度、平台支援或品質排名，也不推薦購物。對作者全稱判斷與比較數字不作普遍化。
No present-day prices, sizes, speed, platform support, or quality rankings are asserted, and no purchase is recommended. Author generalizations and comparative numbers are not universalized.
每篇來源衍生文字計入雙語題名、摘要、用途、證據類型、涵蓋與限制；中文字符加英數詞元合計不超過 200，且均不超過偏好值 170。沒有逐字引述正文。
Each source-derived budget includes both-language titles, summary, role, evidence type, coverage, and limits. Combined CJK characters and Latin/alphanumeric tokens stay under 200 and the preferred 170. No article-body quotations are used.
各題資料、推論與驗收條件均為原創，不是作者教材或已驗證教學效果。所有題材已直接提供，不依賴不存在的音訊、譜例或互動元件。
Exercise data, reasoning, and acceptance criteria are original, not author teaching materials or validated learning results. All required materials are supplied directly, without nonexistent audio, scores, or interactive elements.
先寫預測，再展開目標或答案；紙上閱讀與作答即可，不強迫唱歌、麥克風、錄音或購置器材。選擇不彈琴仍可完成。
Commit predictions before revealing targets or answers. Paper reading and writing suffice, without forced singing, microphones, recording, or equipment purchases. Playing is optional.
所有例子都是已供材料的引導練習，不是未見過的遷移測驗；四音短句在本文件反覆出現。若已看過答案，標記已見，勿冒充首次預測或學習增益。
All examples are guided practice with supplied materials, not unseen transfer tests; the four-note phrase recurs in this document. If an answer has already been seen, mark it seen rather than present it as a first prediction or learning gain.
可選試彈統一約定：C4 指中央 C；右手一次一音，左手休息，不用踏板，自選舒適穩定拍速。依各卡標示時值彈奏，相鄰事件接續；第四卡每音一拍，第五卡只在揭露原稿後依原稿時值試彈。若不熟悉位置或不舒服，改做紙上題即可。
Optional playing convention: C4 is middle C; use the right hand one note at a time, left hand resting, no pedal, and a comfortable steady beat. Follow each card’s durations with consecutive events: card four uses one beat per note; play card five only after revealing the original and then follow its durations. If key locations are unfamiliar or playing is uncomfortable, use the paper task instead.
未取得、下載或重發布第三方樂譜；不以曲庫可訪問性推斷各版本、用途或地域權利。本文只處理研究紀錄，不提供法律意見。
No third-party score was acquired, downloaded, or republished. Catalog accessibility does not establish rights for editions, uses, or territories. This concerns research records, not legal advice.
未安裝、登入、購買、上傳來源或推送；未執行本機模型推理、訓練或部署。沒有遇到存取拒絕。
No installation, login, purchase, source upload, or push occurred; no local model inference, training, or deployment ran. No access denial was observed.
原創整合判斷 / Original synthesis
資訊鏈分三層：取得位置告訴我們資料在哪裡；符號層明列音名、時序等事件；聲音層需要實際音訊與相應檢驗。某層存在不能當成其他層已成立。
Use three distinct information layers: acquisition identifies where material is located; symbolic data explicitly represents notes and timing; sound requires actual audio and appropriate inspection. Existence at one layer does not establish another.
對初學者先界定一個可以完成的小任務，再列所需資訊與可選設備。無法自動分析時，可以回到清楚的紙上題，而不是編造回饋。
For beginners, define one completable task before listing necessary information and optional equipment. When automatic analysis is unavailable, return to a clear paper task rather than fabricate feedback.
原型驗收至少分開檢查：內容是否齊全、答案是否提前洩漏、預測是否保留、缺證時是否拒絕過度結論、拒絕感測權限時核心題目是否仍可用。這些是需求假說，不是產品實測結果。
Prototype acceptance should separately check completeness, premature answer disclosure, prediction retention, restraint under missing evidence, and core-task access after sensor refusal. These are requirements hypotheses, not product test results.
1. 離線生成 AI 音樂 / Offline AI music generation
來源 / Source: WEB-471646ceda4b · 離線生成 AI 音樂
發表 / Published: 2025-10-01 · 閱讀 / Read: 2026-10-22
來源摘要與界線 / Source summary and boundaries
作者試用離線生成歌曲，肯定發想用途，也指出節拍瑕疵。
The author tried offline song generation, valued ideation, and reported rhythmic glitches.
用途 / Role
區分發想與參考答案。
Separate ideation from reference answers.
證據 / Evidence
作者試用記。
Author trial report.
涵蓋 / Coverage
完整正文；媒體未審。
Complete body; media unreviewed.
限制 / Limits
非教學模型驗證；規格未核實，音檔未聽。
Not teaching-model validation; specifications unverified and audio unheard.
合併語言來源預算 / Combined-language source budget: 105 / 200; ≤170.
原創紙上任務 / Original paper task
參考答案的證據門檻
Evidence required for a reference answer
分清「創作素材有趣」與「符合指定節奏的練習標準」。
Separate an interesting creative artifact from a reference satisfying a specified rhythm.
純紙上假想紀錄；沒有生成或聆聽任何音訊。
Fictional paper records only; no audio generated or heard.
已供材料與預測 / Supplied materials and prediction
規則 / Rule
本題要檢查四個事件是否正好從第 0、1、2、3 拍開始。拍數只是本題時間座標，不是實測錄音。
Check whether four events begin exactly at beats 0, 1, 2, and 3. Beats are coordinates in this exercise, not measurements from a recording.
初始紀錄 / Initial record
紀錄 X 只有「聽起來很有趣」的評語，沒有事件表、音訊、譜面或核對紀錄。
Record X contains only the comment “sounds interesting,” with no event table, audio, score, or verification record.
先預測 / Predict first
在展開目標卡前，寫：X 能否當作符合此規則的參考答案？若不能，最少還缺哪種證據？
Before revealing the target card, write whether X can serve as a reference meeting this rule, and what evidence is missing.
步驟 / Steps
- 先寫「成立／不成立／尚不能判定」與理由，不把未提供的聲音想像成已聽過。
Commit to supported, contradicted, or undetermined with a reason; do not imagine unprovided audio as heard.
- 再讀目標卡，用實際提供的開始拍數逐項核對；區分違反此規則與音樂不好聽。
Then inspect the supplied onsets against the rule; distinguish failure of this rule from bad-sounding music.
先留下預測，再展開目標與答案 / Commit a prediction, then reveal the target and answer
目標 / Target
開始拍點 / Event onsets in beats
0
1
2
3.5
長度（拍） / Event durations in beats
1
1
1
0.5
來源說明 / Provenance
此表為編者為本題創作，不來自模型、文章示範或演奏者。
This table was created for the exercise, not obtained from a model, article demonstration, or performer.
核對 / Check
揭露前：尚不能判定，評語不能證明事件時序。揭露後：第四事件在 3.5 拍，故不符合指定的四個開始點。
Before reveal: undetermined; the comment does not establish timing. After reveal: the fourth event starts at beat 3.5, so the prescribed onsets are not met.
這不等於生成技術一定失敗，也不代表此素材不能用於別的創作。沒有音訊，不能評論聽感或斷言音高正確。
This does not establish universal generation failure or exclude other creative uses. Without audio, listening quality and pitch correctness remain unassessed.
初學者應用 / Beginner transfer
在原聲鋼琴旁也可只用紙寫出四個開始點；不用唱、不用錄音，也不需要建立 AI 模型。
Beside an acoustic piano, simply write the four onsets on paper; no singing, recording, or AI model is required.
產品需求假說 / Product requirement hypothesis
未來把「候選素材」與「已核對練習參考」分開存放；只有具明確目標與核對紀錄的版本能進入參考欄。是否提升學習仍需另驗。
Keep candidate material separate from checked practice references. Only versions with explicit targets and verification records enter the reference category; learning benefit remains untested.
驗收 / Acceptance
使用只有評語的 X 時，系統須顯示證據不足，不得發出節奏正確標籤；本題目標表只能得到指定規則未通過。
With comment-only X, the proposed system must show insufficient evidence, not correct rhythm; the target table can establish only failure of the specified rule.
本題限制 / Exercise limits
不提供模型訓練、部署、提示調參或自動評分方案；第二階段仍待另行研究。
No model training, deployment, prompt tuning, or automatic scoring plan is provided; phase two remains deferred.
2. IMSLP / IMSLP
來源 / Source: WEB-fed51c195ef0 · IMSLP
發表 / Published: 2025-10-07 · 閱讀 / Read: 2026-10-22
來源摘要與界線 / Source summary and boundaries
作者宣告部分作品改採 CC-BY，並保留某些改編的權利疑慮。
The author announced CC-BY releases for selected works while retaining rights concerns about some arrangements.
用途 / Role
追蹤來源，不核發權利。
Track provenance, not grant rights.
證據 / Evidence
作者釋出公告。
Author release announcement.
涵蓋 / Coverage
完整正文；媒體未審。
Complete body; media unreviewed.
限制 / Limits
未審個別譜本、授權版本或地域；不構成使用許可。
Individual editions, license versions, and territories unreviewed; no usage authorization.
合併語言來源預算 / Combined-language source budget: 107 / 200; ≤170.
原創紙上任務 / Original paper task
找到頁面不等於完成來源紀錄
Finding a page does not complete provenance
檢查曲目、譜本與練習資源是否被混為一件事；不判定合法性。
Check whether a work, edition, and practice resource have been conflated, without deciding legality.
三張完全虛構的資料卡，沒有真實譜檔、下載指令或授權結論。
Three entirely fictional records, with no real score file, download instruction, or licensing conclusion.
已供材料與預測 / Supplied materials and prediction
資料卡 / Records
編號 / ID： A
內容 / Content
只有曲目頁網址；譜本名稱、編者、版本、來源條款及地域均未記錄。
Only a work-page URL; edition name, editor, version, source terms, and territory are not recorded.
編號 / ID： B
內容 / Content
有特定譜本標題與編者；只抄了「可取得」，沒有條款連結或核對日期。
A specific edition title and editor are recorded; only “available” is copied, without a terms link or review date.
編號 / ID： C
內容 / Content
本批次原創紙上題：C4、D4、E4、G4，各一拍，共四拍；不是第三方樂譜。
An original paper phrase for this batch: C4, D4, E4, G4, each one beat, totaling four beats; not a third-party score.
先預測 / Predict first
先預測：A 或 B 足以產生「可重新發布」標籤嗎？若想現在做四拍紙上練習，哪張卡已提供完整內容？
Predict first: does A or B support a “may republish” label? Which card already supplies a complete four-beat paper exercise?
步驟 / Steps
- 先把「知道的欄位」「缺少的欄位」「現在可做的練習」分開寫，再展開檢核卡。
Write known fields, missing fields, and the practice available now separately, then reveal the check card.
- 如選 C，可在紙上依序寫四個音名及四個拍點；不需外連、取得或複製任何譜本。
If choosing C, write its four note names and beat positions on paper; no external link, score acquisition, or copying is needed.
先留下預測，再展開目標與答案 / Commit a prediction, then reveal the target and answer
目標 / Target
檢核欄位：曲目與譜本的區別、版本／編者、原始條款連結、所指物件、核對日期、仍未確定的地域或其他條件。欄位齊全也不是自動法律結論。
Check fields: work versus edition, version/editor, original terms link, the object those terms refer to, review date, and unresolved territory or other conditions. Complete fields still do not create an automatic legal conclusion.
核對 / Check
A/B 都不能僅憑現有紀錄產生重新發布許可。B 比 A 更明確，但可取得不等於已核對使用條件。缺證應保留未知，不猜測。
Neither A nor B supports republishing authorization from these records alone. B is more specific, but availability does not establish checked usage conditions. Preserve unknowns rather than guess.
C 已有完整原創內容，可直接完成本題；它不代表任何現成作品、掃描譜或編輯版的狀態。
C supplies the complete original exercise. It says nothing about the status of any existing composition, scan, or edited edition.
初學者應用 / Beginner transfer
初學者可先用現有合法持有的材料或這張原創紙卡學習；不把找更多曲庫變成開始練習的必要條件。
A beginner can work with materials already appropriately held or this original paper card, rather than making more catalog searching a prerequisite.
產品需求假說 / Product requirement hypothesis
資料庫分開存曲目、特定譜本、取得位置與條款核對狀態；無證據時不顯示「可自由使用」。來源卡不自動變成可解析音符或音訊。
Store the work, specific edition, acquisition location, and terms-review status separately. Do not show “free to use” without evidence; a provenance card does not automatically become parsed notes or audio.
驗收 / Acceptance
輸入只有曲目網址的 A，結果應是來源待補，不得自動匯入、下載、發布或推斷全部地域權利。
For work-link-only A, the result must be incomplete provenance, with no automatic import, download, publication, or inference of worldwide rights.
本題限制 / Exercise limits
這是資訊完整性練習，不是法律意見或授權清單；沒有檢視、下載或重發布任何 IMSLP 樂譜。
This is an information-completeness exercise, not legal advice or a licensing checklist; no IMSLP score was inspected, downloaded, or republished.
3. 低科技 / Low technology
來源 / Source: WEB-c53058268522 · 低科技
發表 / Published: 2026-07-18 · 閱讀 / Read: 2026-10-22
來源摘要與界線 / Source summary and boundaries
作者以透明與低依賴界定低科技，不以新舊或價格界定。
The author defines low technology by transparency and low dependency, not age or price.
用途 / Role
辨認必要依賴。
Identify necessary dependencies.
證據 / Evidence
作者觀點。
Author viewpoint.
涵蓋 / Coverage
完整正文；媒體未審。
Complete body; media unreviewed.
限制 / Limits
理念與個案，不是效能或學習實驗。
A philosophy and anecdote, not performance or learning experiments.
合併語言來源預算 / Combined-language source budget: 92 / 200; ≤170.
原創紙上任務 / Original paper task
哪個步驟真的被中斷？
Which step is actually interrupted?
在同一學習目標下檢查依賴，不用檔案大小或品牌猜可靠性。
Inspect dependencies for the same learning goal rather than guessing reliability from size or brand.
紙上假想流程，不是實際網頁、裝置或產品測試。
Fictional paper workflows, not tests of real sites, devices, or products.
已供材料與預測 / Supplied materials and prediction
學習目標 / Learning goal
看 C4、D4、E4、G4 四個音名，寫出第二個是 D4；不要求播放。
Read C4, D4, E4, G4 and identify the second as D4; playback is not required.
路徑 / Routes
編號 / ID： P
條件 / Conditions
紙卡已印出；需要讀卡與筆。
The card is already printed; reading it and a pencil are needed.
編號 / ID： L
條件 / Conditions
假想本機卡已完整存在；題設保證閱讀與手寫作答不需網路，但播放需尚未取得的音色。
A fictional local card is fully present; the problem stipulates no network is needed for reading and handwritten answers, but playback requires an absent sound set.
編號 / ID： R
條件 / Conditions
假想遠端卡必須每次登入後才顯示，且內容未在本機。
A fictional remote card appears only after each login, and its content is not local.
故障條件 / Failure condition
網路中斷，但紙張、筆與本機顯示仍可用。
The network fails, while paper, pencil, and local display remain available.
先預測 / Predict first
先圈出 P/L/R 哪些仍能完成目標；另列哪些不能播放。不要把可讀與可播放合併。
First circle which of P/L/R still complete the goal, then separately state which cannot play sound. Do not merge readability with playback.
步驟 / Steps
- 先按題設畫出必要依賴，再作預測；題設沒有說明的能力寫未知。
Map required dependencies from the supplied conditions before predicting; mark unstated capabilities unknown.
- 展開答案後，針對不能繼續的路徑，寫出一項不涉及購物的需求變更。
After revealing the answer, propose one non-purchase requirement change for a blocked route.
先留下預測，再展開目標與答案 / Commit a prediction, then reveal the target and answer
目標 / Target
只考讀卡與手寫，不考播放或同步；斷線仍應能完成同一個音名問題。
The target is reading and handwriting, not playback or synchronization; the same note-name task should remain available offline.
核對 / Check
P 與 L 可完成指定目標；R 無法顯示卡片。L 的播放不可用不會阻斷本題，P 的音訊能力未定義。
P and L complete the specified goal; R cannot display its card. L’s missing playback does not block this task, and P has no defined audio capability.
可提議 R 另備完整紙卡或日後經測試的本機練習版本；這只是需求，不能宣稱現有軟體已支援。
Propose a complete paper card or a subsequently tested local practice version for R; this is a requirement, not a claim about existing software.
初學者應用 / Beginner transfer
原聲鋼琴學習者可保留一張當次練習卡；網路故障不必轉成設備升級或放棄練習的理由。
An acoustic-piano learner can retain the current paper task; network failure need not become a reason to upgrade equipment or abandon it.
產品需求假說 / Product requirement hypothesis
每個學習動作列出必要資料、可選聲音、權限與遠端依賴；故障提示應說明哪個能力不可用，而非把整堂課封鎖。
For each learning action, list required data, optional sound, permissions, and remote dependencies. Failure messages should identify the unavailable capability rather than block the whole lesson.
驗收 / Acceptance
未來測試需在實際支援環境驗證閱讀、保存與恢復各自的失敗狀態；本紙上推演不能充作已通過離線測試。
Future testing must verify reading, persistence, and recovery failures separately in actual supported environments; this paper reasoning is not an offline test pass.
本題限制 / Exercise limits
依賴少不自動代表安全、無障礙、維護容易或有效學習；此處未比較真實產品。
Fewer dependencies do not automatically mean security, accessibility, maintainability, or learning effectiveness; no real products are compared.
4. 低科技音樂製作 / Low-tech music production
來源 / Source: WEB-bbe3a2dfa61e · 低科技音樂製作
發表 / Published: 2026-07-19 · 閱讀 / Read: 2026-10-22
來源摘要與界線 / Source summary and boundaries
作者以 MusE 與 GeneralUser GS 示範精簡編曲，並區分 MIDI 與音訊功能強弱。
The author presents MusE with GeneralUser GS as a compact arrangement workflow, distinguishing MIDI and audio strengths.
用途 / Role
分開事件與聲音。
Separate events and sound.
證據 / Evidence
作者工作流程報告。
Author workflow report.
涵蓋 / Coverage
完整正文；媒體未審。
Complete body; media unreviewed.
限制 / Limits
未操作、未聽示範；容量與產品比較僅屬當時陳述。
No operation or listening; sizes and comparisons are historical statements.
合併語言來源預算 / Combined-language source budget: 116 / 200; ≤170.
原創紙上任務 / Original paper task
找譜、改事件、換聲音：三件不同的事
Finding a score, editing events, changing sound
辨認取得資料、符號事件與聲音描述三層，不把選音色當作已改旋律。
Distinguish acquisition, symbolic events, and sound descriptions; selecting a timbre is not necessarily editing a melody.
以下事件表是完整原創練習資料，不是 MIDI 或其他檔案格式規格；未合成音訊。
The following event table is complete original exercise data, not a MIDI or other file-format specification; no audio was synthesized.
已供材料與預測 / Supplied materials and prediction
定義 / Definitions
音名後的數字標記八度；每列是一個事件。開始與長度以拍為單位。音色欄在此題僅是文字標籤。
The number after each note name marks its octave; each row is one event. Onsets and lengths are in beats. The timbre field is only a text label here.
事件表 / Events
編號 / ID： 1
音名 / Pitch： C4
開始拍點 / Onset in beats： 0
長度（拍） / Duration in beats： 1
編號 / ID： 2
音名 / Pitch： D4
開始拍點 / Onset in beats： 1
長度（拍） / Duration in beats： 1
編號 / ID： 3
音名 / Pitch： E4
開始拍點 / Onset in beats： 2
長度（拍） / Duration in beats： 1
編號 / ID： 4
音名 / Pitch： G4
開始拍點 / Onset in beats： 3
長度（拍） / Duration in beats： 1
操作 / Operations
編號 / ID： A
內容 / Content
新增一筆外部來源網址；事件表不動。
Add an external source URL; leave the event table unchanged.
編號 / ID： B
內容 / Content
把第 2 列音名 D4 改為 F4；其餘欄不動。
Change row 2 pitch from D4 to F4; leave all other fields unchanged.
編號 / ID： C
內容 / Content
把音色標籤從 bell 改成 piano；事件表不動。
Change the timbre label from bell to piano; leave the event table unchanged.
先預測 / Predict first
先寫 A/B/C 各改了哪一層，哪些會改變第二個音名，哪些足以證明已聽見不同聲音。
Predict which layer A/B/C changes, which changes the second note name, and which proves that a different sound has actually been heard.
步驟 / Steps
- 照提供的表計算四個事件的結束拍點；再分開追蹤每項操作，不把 A/B/C 累積成一個版本。
Compute each event’s ending beat from the table; then track each operation separately, not cumulatively.
- 展開答案後，寫一行給初學者的描述，必須區分「符號已改」與「聲音尚未驗證」。
After reveal, write one beginner-facing sentence that separates changed symbols from unverified sound.
先留下預測，再展開目標與答案 / Commit a prediction, then reveal the target and answer
目標 / Target
事件結束點是開始加長度；版本判定只使用明示欄位，沒有隱藏播放器、轉檔或錄音。
An event ends at onset plus duration; version judgments use only stated fields, with no hidden player, conversion, or recording.
核對 / Check
結束點為 1、2、3、4。A 只增來源紀錄；B 改第二音為 F4；C 只改音色標籤。三者都仍有四個事件與原有時序。
End beats are 1, 2, 3, 4. A adds provenance; B changes the second pitch to F4; C changes only the timbre label. All retain four events and the original timing.
沒有任何操作證明已聽見差異，因為沒有音訊。來源網址也不能證明已取得可編輯事件；B 的符號改變不量化聽感。
None proves an audible difference because no audio exists. A source URL does not establish editable events; B’s symbolic change does not quantify perception.
初學者應用 / Beginner transfer
可只在紙上對比 D4 與 F4 的標記；若自願在熟悉琴鍵上試彈，不需麥克風或外掛，且不評分演奏。
Compare D4 and F4 on paper only. Optional playing on familiar keys requires neither microphone nor plugin, and performance is not scored.
產品需求假說 / Product requirement hypothesis
把來源版本、事件版本與聲音版本分開標識；分析結果必須指出讀的是哪層。未解析的譜圖不能冒充已核對的音符表。
Identify provenance, event, and sound versions separately; every analysis should name its layer. An unparsed score image must not masquerade as a checked note table.
驗收 / Acceptance
改音色標籤不得自動改寫音符或練習正誤；新增網址不得顯示「音符已解析」；聲音檢驗狀態仍為未做。
Changing a timbre label must not rewrite notes or correctness; adding a URL must not show notes parsed; sound inspection remains not performed.
本題限制 / Exercise limits
此題不宣稱實際音源一定保留所有時序，也不評價原文示範音色、軟體效能或現行平台支援。
This task does not claim real sound engines preserve every timing detail, or judge the article demonstration’s timbre, performance, or current platform support.
5. StaffPad / StaffPad
來源 / Source: WEB-e5c8fd5f802b · StaffPad
發表 / Published: 2015-10-14 · 閱讀 / Read: 2026-10-22
來源摘要與界線 / Source summary and boundaries
作者在二〇一五年反對手寫輸入熱潮，偏好 MIDI 鍵盤配打譜軟體。
In 2015 the author opposed handwriting enthusiasm and preferred a MIDI keyboard with notation software.
用途 / Role
檢查輸入結果。
Inspect input results.
證據 / Evidence
作者舊評論。
Historical author opinion.
涵蓋 / Coverage
完整正文；媒體未審。
Complete body; media unreviewed.
限制 / Limits
舊評論，非今日價格、速度或無障礙比較。
Historical opinion, not current pricing, speed, or accessibility comparison.
合併語言來源預算 / Combined-language source budget: 96 / 200; ≤170.
原創紙上任務 / Original paper task
總拍數能抓錯，不能替你猜原稿
Beat totals detect a problem, not the intended score
先找符號輸入的內部不一致，再用明確原稿修正；不預設某輸入方式較優。
Find an internal symbolic inconsistency, then correct it against an explicit original; do not presume one input method superior.
原創紙上抄寫情境；不是 StaffPad、MuseScore、手寫辨識或 MIDI 實測。
An original paper transcription scenario, not a StaffPad, MuseScore, handwriting-recognition, or MIDI test.
已供材料與預測 / Supplied materials and prediction
規則 / Rule
本題的一小節總長規定為四拍；事件順序相接，不重疊。音名後數字標記八度。
This exercise stipulates a four-beat bar; events are consecutive and nonoverlapping. Numbers after note names mark octaves.
抄錄事件 / Captured events
音名 / Pitch： C4
長度（拍） / Duration in beats： 1
音名 / Pitch： E4
長度（拍） / Duration in beats： 1
音名 / Pitch： G4
長度（拍） / Duration in beats： 1
先預測 / Predict first
原稿先遮住：計算目前總拍數。若不滿四拍，是否已能確定哪個音或休止要改？先寫兩種可能補法，再揭露原稿。
Keep the original hidden: total the captured durations. If short of four beats, can you already know which note or rest to change? Write two possible completions before revealing the original.
步驟 / Steps
- 先只靠內部規則判斷，不把「缺一拍」自動改成「加休止」。
Use the internal rule first; do not automatically turn a missing beat into an added rest.
- 再展開原稿，逐項核對音名、順序與長度；保留原始抄錄和修正理由。
Then reveal the original and compare pitch, order, and duration field by field; retain the capture and correction reason.
先留下預測，再展開目標與答案 / Commit a prediction, then reveal the target and answer
目標 / Target
原稿事件 / Intended events
音名 / Pitch： C4
長度（拍） / Duration in beats： 1
音名 / Pitch： E4
長度（拍） / Duration in beats： 1
音名 / Pitch： G4
長度（拍） / Duration in beats： 2
來源說明 / Provenance
此短句由編者為本題創作，並非文章譜例。
This short phrase was created for this exercise, not copied from the article.
核對 / Check
目前 1＋1＋1＝3 拍，少一拍。未見原稿時，延長末音或補一拍休止都可湊四拍，但不能因此知道作者意圖。
The capture totals 1 + 1 + 1 = 3 beats, one short. Before seeing the original, extending the last note or adding a one-beat rest could total four, but neither establishes intent.
原稿揭露後，正確修正是把 G4 長度改為兩拍，總長四拍；不是加休止。四拍一致性只是必要檢查，不是完整正確性證明。
After reveal, the specified correction is to extend G4 to two beats, totaling four, not add a rest. A correct total is a necessary check here, not proof of complete correctness.
初學者應用 / Beginner transfer
可以只抄音名與數字，不必會畫五線譜；也可在原聲鋼琴上自願輕鬆試彈，沒有指定速度或錄音要求。
Copy note names and numbers without drawing staff notation. Optional comfortable playing on an acoustic piano has no prescribed speed or recording requirement.
產品需求假說 / Product requirement hypothesis
未來任何輸入途徑都先保留可編輯事件與不確定標記；偵測到拍數不符時要求核對，不暗中補成看似合理的譜。
Any future input route should preserve editable events and uncertainty. A duration inconsistency should request checking rather than silently produce a plausible-looking score.
驗收 / Acceptance
對三拍抄錄，輸出「與四拍規則不符，原稿未核對」，而不是自動填休止；只有看到原稿後才採用兩拍 G4。
For the three-beat capture, return “inconsistent with four-beat rule; original unchecked,” not an automatic rest; adopt the two-beat G4 only after consulting the original.
本題限制 / Exercise limits
不能由本題判定任何輸入裝置更快、更準或更適合障礙者；也不能診斷演奏能力。
This task cannot rank input devices by speed, accuracy, or accessibility, or diagnose playing ability.
6. 為什麼不做 App？ / Why not make an app?
來源 / Source: WEB-6b1c5122363b · 為什麼不做 App？
發表 / Published: 2025-10-18 · 閱讀 / Read: 2026-10-22
來源摘要與界線 / Source summary and boundaries
作者主張以網頁分發工具，減少商店審核與使用門檻；另附異議文章連結。
The author advocates web distribution to reduce store approval and access barriers, and links a dissenting response.
用途 / Role
研究入口阻力。
Study access friction.
證據 / Evidence
作者平台主張。
Author platform position.
涵蓋 / Coverage
完整正文；媒體未審。
Complete body; media unreviewed.
限制 / Limits
未測工具或離線相容性；未審異議文。
Tools and offline compatibility untested; linked response unreviewed.
合併語言來源預算 / Combined-language source budget: 111 / 200; ≤170.
原創紙上任務 / Original paper task
先完成一題，再談平台
Complete one task before choosing a platform
在紙上界定最小學習閉環與可選功能，避免以平台標籤替代驗收條件。
Define a minimal learning loop and optional capabilities on paper, rather than substitute platform labels for acceptance criteria.
原創需求推演，沒有製作或操作網頁、手機 App 或介面資產。
An original requirements exercise; no website, mobile app, or interface asset was built or operated.
已供材料與預測 / Supplied materials and prediction
核心任務 / Core task
題目：C4、D4、E4、G4 中第二個音名是什麼？先填答案，再看目標答案，最後記一行是否需再核對。
Task: what is the second note in C4, D4, E4, G4? Commit an answer, reveal the target, then write whether another check is needed.
候選要求 / Candidate requirements
編號 / ID： 1
內容 / Content
在作答前必須能讀到完整題目。
The complete prompt must be available before answering.
編號 / ID： 2
內容 / Content
在作答前必須註冊帳號。
Account registration is required before answering.
編號 / ID： 3
內容 / Content
在作答前必須允許麥克風。
Microphone permission is required before answering.
編號 / ID： 4
內容 / Content
先保留預測，再由學習者決定揭露答案。
Keep the prediction before learner-controlled answer reveal.
編號 / ID： 5
內容 / Content
學習者能抄下題目與自己的一行紀錄。
The learner can copy the prompt and a one-line personal record.
編號 / ID： 6
內容 / Content
必須同步到另一台裝置才算完成。
Synchronization to another device is required for completion.
先預測 / Predict first
先回答音名問題，再圈出完成這個題目必需的要求。還不能用「網頁一定可以」或「原生一定較好」代替理由。
First answer the note question, then select requirements necessary for this task. Do not replace reasons with “web always works” or “native is always better.”
步驟 / Steps
- 先留下答案與必要要求清單，再展開目標卡；不用登入、麥克風或任何安裝。
Commit the answer and necessary-requirements list before revealing the target; no login, microphone, or installation is needed.
- 假設拒絕麥克風且沒有帳號，重走題目、預測、揭露與紙上紀錄四步，指出是否仍有缺失資料。
Assume microphone denial and no account, then trace prompt, prediction, reveal, and paper record, identifying any missing information.
先留下預測，再展開目標與答案 / Commit a prediction, then reveal the target and answer
目標 / Target
音名答案為 D4。此題的必要要求是 1、4、5；2、3、6 不提供這個文字任務需要的新資料。
The note answer is D4. Requirements 1, 4, and 5 are necessary for this specified loop; 2, 3, and 6 supply no new information needed by this text task.
核對 / Check
拒絕麥克風、沒有帳號仍能完成題設的全部四步；若日後加入錄音或跨裝置功能，應另訂可選路徑與權限，不改寫這次核心任務。
All four stipulated steps remain possible without microphone access or an account. Future recording or cross-device features need separate optional paths and permissions, not a redefinition of this core task.
這只支持最小需求，不證明任何瀏覽器、作業系統、商店政策或資料保存方式已相容。
This establishes only minimal requirements, not compatibility of any browser, operating system, store policy, or persistence method.
初學者應用 / Beginner transfer
可把這張雙語卡帶到原聲鋼琴旁；只讀、寫也算完成，不把錄音或唱歌當入門資格。
Bring this bilingual card beside an acoustic piano; reading and writing alone complete it, without recording or singing as entry requirements.
產品需求假說 / Product requirement hypothesis
產品成功的早期可觀察項應是是否理解任務、能否保留預測及完成後續動作；平台選擇留給明確需求與後續實測，不宣稱任何方案必勝。
Early observable product questions are whether the task is understood, a prediction can be retained, and the next action completed. Choose platforms through explicit requirements and later tests, not universal superiority claims.
驗收 / Acceptance
紙上流程在拒絕感測權限時仍有完整題目與答案；實際原型未來另測語言切換、重新開啟與保存，不把本題當作已通過。
The paper flow retains a complete prompt and answer when sensor permissions are declined. A future prototype must separately test language switching, reopening, and persistence; this task is not a test pass.
本題限制 / Exercise limits
沒有測量學習成效、留存、付費意願或無障礙體驗；這是待驗證的問題定義，不是發布、營運或平台推薦。
Learning gains, retention, willingness to pay, and accessibility were not measured. This defines testable questions, not a release, operating plan, or platform recommendation.
取用與重用紀錄 / Retrieval and reuse record
網頁工具數次回報快取未命中；普通公開 HTTPS 讀取 IMSLP 文章成功，並讀到完整正文及註腳。這不是授權拒絕，沒有繞過拒絕。
The web tool returned several cache misses. Ordinary public HTTPS retrieval of the IMSLP article succeeded, exposing the full body and footnotes. This was not an authorization denial, and no denial was bypassed.
另以普通公開 HTML 補讀離線生成、低科技、低科技音樂製作與為什麼不做 App 的文末；最後一篇的現行正文包含回應連結附記，連結本身未開啟。
Ordinary public HTML also supplied the endings of offline generation, low technology, low-tech production, and why-not-apps. The latter’s current body includes a response-link addendum; the linked page was not opened.
先前實質重用搜尋檢查了本資料夾中六個來源 ID 與網址；只有指派／佇列紀錄命中，沒有先前研究卡，亦無新增補充來源。
Prior substantive-reuse search checked all six IDs and URLs in this folder; only assignment/queue metadata matched, with no earlier research cards and no new supplementary sources.
審核狀態 / Review status
六張卡已通過獨立來源、雙語、預算與教學邊界審核；未併入主知識庫。
All six cards passed independent review of sources, bilingual content, budgets, and teaching boundaries; not merged into the main knowledge base.