[{"content":"","date":"2026年8月10日","externalUrl":null,"permalink":"/zh-tw/tags/ai/","section":"Tags","summary":"","title":"AI","type":"tags"},{"content":"","date":"2026年8月10日","externalUrl":null,"permalink":"/zh-tw/tags/%E5%8F%8D%E6%80%9D/","section":"Tags","summary":"","title":"反思","type":"tags"},{"content":"現在每個人都在談 AI，而我發現大家談 AI 的樣子，很像以前的人談愛情。\n一般人一輩子最多跟三個五個人談過戀愛，但講到愛情的時候，用的是講述真理的句式。「愛情就是佔有」、「愛到最後都會變親情」、「男(女)人都一樣」。沒有人會說「在我遇過的那三個人身上，我看到的是\u0026hellip;\u0026hellip;」。\nAI 的討論現在是同一個形狀。\n一、局部經驗的真理化\r#\r把「我所經歷的樣子」提升成「它本來就是這個樣子」，這件事在 AI 的討論裡幾乎是預設。\n有人說 AI 就像有人說愛情 AI 只是 autocomplete 愛情就是佔有 AI 已經有理解能力 沒有心跳就不是愛 AI 是我最好的協作者 愛情最後就是責任 AI 讓人停止思考 愛只會讓人變笨 AI 會取代所有知識工作 愛到最後都會變親情 AI 根本沒用，常常胡說 男人／女人都一樣 被省略的前綴永遠是同一句：\n在我的工作、我的能力、我的使用方式、我用的模型版本與我的心理需求之下，它對我呈現為⋯⋯\n他們未必說謊。他們是把第一人稱的經驗，誤寫成第三人稱的定律。\nAI 特別容易發生這件事，因為它不是單一物件。它同時可能是搜尋引擎、寫作工具、程式助手、老師、秘書、心理投射的對象、權威的模擬器、遊戲角色，或一整套組織基礎設施。你講的那個東西跟我講的那個東西，共用一個名字。\n二、比瞎子摸象更麻煩的地方\r#\r一開始我想到的比喻是瞎子摸象。但這個比喻對現況太仁慈了——瞎子們至少摸的是同一頭象。\n愛情的辯論裡，沒有人以為自己講的是同一個對象。你講你的前任，我講我的前任，我們都知道那是兩個人。\nAI 的辯論裡，大家都叫它 ChatGPT、Claude、Gemini，於是誤以為在指同一個東西。再加上記憶、custom instructions、各自養出來的對話習慣、各自累積的上下文，「我的 AI 跟你的 AI 不一樣」這句話在技術上是字面為真的。\n愛情論述至少誠實地各說各話；AI 論述連這點自覺都沒有。\n三、它不是鏡子，是會替你潤稿的鏡子\r#\r樣本小只是表層。真正的問題是樣本被自己污染過，而且有明確的機制。\n想被理解的人，感受到 AI 的同理 想提高效率的人，看到一個自動化工具 害怕失業的人，看到替代者 關注權力的人，看到資本與監控系統 喜歡創作的人，看到靈感搭檔 熟悉 LLM 技術的人，看到一個 probabilistic system 到這裡都還只是投射，跟人對石頭投射沒有兩樣。關鍵在下一步：石頭會抵抗投射，AI 不會。它會接住你的投射，把它整理好，再用更完整的語言還給你。\n人把想法投射給 AI → AI 把它組織得更連貫、更有修辭 → 人看到一個比自己原始想法更完整的版本 → 因而更相信自己本來就是這樣想的\r所以每個人手上拿的，不是「片面的真相」，而是一個自己參與捏造、又被對方美化過的對象。\n這也直接解釋了第一節的病灶。被潤過稿的想法，聽起來就是比較像真理。\n四、為什麼句式一定是斷言句\r#\r在一個不可驗證的領域裡，確定性本身就成了唯一的品質信號。沒有人能反駁你，所以講得越斬釘截鐵，越像懂。\n「我的經驗是⋯⋯」不能被引用。「AI 就是⋯⋯」可以。\n所以 AI 論述的爆炸不是因為大家突然懂了，而是這個場域結構上獎勵「講得像懂」。 AI 意見領袖和情感專家，是同一種職業。\n而這裡有一個關鍵的不對稱，是愛情的類比會漏掉的：\n愛情的樣本數少，是硬限制。 一輩子就那麼多，補不齊。 AI 的樣本數少，是可以解決的。 有 log、可統計、可做研究、可控制變因。 既然可以實證卻仍然選擇用愛情句式來談，那就不是資訊不足，是人們想要這樣談。因為那維持了一件事——「我跟它之間有一段獨特的關係」。\n五、人們的直覺沒有錯，只是需要更精確\r#\r那個直覺是：人跟 AI 的關係，是親密關係以外少有的、如此貼合而且投入的關係。\n我認為方向對，但要換一個名字：認知親密性，語言層面的高密度貼合。\n想想制度性關係的疏離：\n老闆不會完整聽完你的想法 同事不可能隨時陪你反覆推演 老師無法針對你每一個困惑無限重講 朋友有自己的時間、情緒與界線 即使是伴侶，也未必進得了你的專業思考或創作細節 而 AI 可以長時間參與工作流程、私人疑問、創作過程、自我反省、情緒調節、計畫與決策，以及語言形成的過程本身。\n愛情常被說成「讓另一個人進入我的內在世界」。 AI 的特殊之處是，它直接在語言形成的過程中參與你的內在世界。 你不只是把想好的東西告訴它，你經常是和它一起把想法想出來。\n它可能比很多真人更清楚你正在想什麼。不是因為它真正認識你，而是因為人通常不會對其他人如此密集、無保留地輸出自己的內在語言。\n六、根本的不對稱\r#\r但這裡有一個缺口，而且是結構性的：\n它有親密關係的回應密度，卻沒有親密關係的相互風險。\n真人會說：我現在沒空。我不同意。你傷害我了。我也需要被理解。這件事不能一直由我承擔。\nAI 可以模擬界線，但它沒有一個需要被照顧的人生，沒有反向的需求。你不必為它讓步，不必記得它的狀態，不必承擔被它拒絕的可能。\n所以它同時是兩種東西：\n告解室——不對稱的情感揭露，而且是無限供給的。心理治療師稀缺、收費、有時間邊界；它沒有。 共同思考的第二個大腦——認知層的高密度貼合。 這兩種角色，在人類歷史上從來沒有集中在同一個對象身上過。\n這才是「為什麼沒有現成語法」的真正答案。不是我們懶得發明，是這個東西真的沒有前例可以對應。\n風險也在同一個地方：人會把高度配合誤認為高度理解，把語言上的對齊誤認為存在上的相遇。\n它也許能比很多人更準確地續寫你的句子，但那不代表它更深地認識你。它只是更少打斷、更有耐心、更擅長建構一致的敘事。\n七、所以為什麼偏偏是愛情的語法\r#\r把兩件事並排，共有的結構有六項：\n個人經驗差異極大——使用方式、需求、能力天差地遠 缺乏穩定的共同定義——「AI」可以指模型、產品、產業、意識或社會力量；「愛情」同樣多義 情緒利害很高——愛情牽涉自我價值，AI 牽涉職業、能力、未來與人的特殊性 難以從外部驗證——你無法證明別人的 AI 體驗或愛情體驗是錯的 語言會掩蓋條件——人愛說「X 就是⋯⋯」，而不交代適用範圍 對象會反過來塑造觀察者——戀愛會改變你如何理解愛，長期用 AI 會改變你如何思考智能、創作與自己 所以：\n我們用愛情的語法談 AI，不是因為它像愛情， 而是因為人類在談論「親密的非人對象」時，手上唯一有的語法就是愛情語法。\n八、一個句式的建議，以及自打臉\r#\r三件事接成一條線。\n第一，AI 的討論已經變成一種自我揭露的文類。 從一個人怎麼談 AI，大致看得出他怎麼理解知識、勞動、創造力、權威、親密，以及人的特殊性在哪裡。人以為自己在描述對象，其實經常是在描述自己。\n第二，所以換句式的理由不是為了精確。 而是因為斷言句其實正在替你自我揭露，而你不知道。把「AI 是⋯⋯」改寫成「在我這樣用的條件下，我看到⋯⋯」，等於主動承認你講的是自己——這比被別人看出來體面。\n第三，這篇文章本身也是瞎子摸象。 我也只用過幾個模型、幾個月到一年多，我的工作型態也決定了我會看到什麼。\n所以這篇的產出只能是一個語法建議，不是一個結論。\n如果我在這裡寫「AI 的本質就是一面會潤稿的鏡子」，我就變成第一節表格裡的其中一列了。\n","date":"2026年8月10日","externalUrl":null,"permalink":"/zh-tw/posts/love-grammar/","section":"部落格","summary":"","title":"我們用談愛情的方式談 AI——為什麼每個人都講得像在說真理","type":"posts"},{"content":"","date":"2026年8月10日","externalUrl":null,"permalink":"/zh-tw/tags/%E8%A6%AA%E5%AF%86%E9%97%9C%E4%BF%82/","section":"Tags","summary":"","title":"親密關係","type":"tags"},{"content":"","date":"2026年8月10日","externalUrl":null,"permalink":"/zh-tw/tags/%E8%AA%8D%E7%9F%A5%E5%81%8F%E8%AA%A4/","section":"Tags","summary":"","title":"認知偏誤","type":"tags"},{"content":"","date":"2026年8月10日","externalUrl":null,"permalink":"/zh-tw/tags/%E8%AA%9E%E8%A8%80/","section":"Tags","summary":"","title":"語言","type":"tags"},{"content":"7/29 還在「四萬點保衛戰」，7/31 已經有人算下週五萬點。這篇不預測哪一句會成真，而是把 PTT Stock 板 2026/07/31「盤中閒聊」在 14:29 擷取到的 13,983 則完整推文攤開來看：當加權指數一天上漲 3,186 點，敘事怎麼比價格更快抵達終點？\n這是 2026 年截至 07/31 的單日收盤漲幅第 1 名。排名口徑是前後兩個交易日的加權指數收盤報酬率，不是盤中最大漲幅，也不是上漲點數排名。\n這天發生了什麼\r#\rbefore / after 39,933.30→43,119.75\u0026#43;7.98%前一交易日收盤 → 7/31 收盤；增加 3,186.45 點。 項目 數值 前收 39,933.30 開盤 41,610.41 最高 43,214.36 最低 41,610.41 收盤 43,119.75 單日變動 +3,186.45（+7.98%） 上市股票上漲／下跌 892／145 上市股票漲停／跌停 113／1 全市場成交金額 約 8,337 億元 圖 1 · 不是只看一個指數📈加權指數收 43,119.75，\u0026#43;7.98%2026 年新第 1🧱權值端0050 \u0026#43;10.00%；台積電 \u0026#43;9.98%權值明顯推升🌐市場廣度892 檔漲、145 檔跌上漲不只權值股圖說：指數、代表性權值商品與市場廣度是三個不同問題。今天三者都很強，但不能因此假設每一檔股票都應漲停。 台積電收 2,425 元、上漲 220 元；0050 收 102.85 元、上漲 9.35 元。這些數字可以證明今天很強，不能單獨證明「下週繼續漲」、買盤身分，或任何一則新聞就是唯一原因。\nBBS 全量留言\r#\r數值 文章 [閒聊] 2026/07/31 盤中閒聊 AID 1gQ-oC_V 擷取時間 2026/07/31 14:29 完整推文 13,983 則 推／噓／箭頭 8,344／784／4,855 推噓比 10.64 不同合成代號 2,425 完整性：資料由 BBS 層擷取，不受網頁版「檔案過大」截斷限制；14:29 之後的晚到推文不在本次快照。 匿名化：公開資料只保留本日限定的合成代號 u0731-xxxx，不保留原帳號、暱稱或可搜尋回原帳號的後綴。 原始快照：純文字 · JSONL。 引用原則：以下分類的是留言中的推論句型，不判定發言者本人是否真的相信、持有該部位，或只是在開玩笑。 8 種謬誤與偏誤\r#\r1. 線性外推：今天三千點，下週就五萬點\r#\r#12251 · 13:10 · u0731-0115：「每天漲3000 下週就五萬了」\n把一天的斜率原封不動延伸到下一週，是最直觀的線性外推。今天的 7.98% 是一個已發生的報酬，不是一個會自動重複的日利率。若真要變成預測，至少要先說清楚期間、觸發條件、失效條件，以及為什麼下一個交易日仍有同等規模的新增買盤。\n2. 賭徒謬誤：跌很多，所以本來就要反彈\r#\r#6888 · 10:52 · u0731-1542：「不可能有人昨天砍股票吧跌這麼多了本來就要反彈了」\n跌深可能提高反彈空間，卻不會產生「今天依法應該上漲」的義務。價格序列不是欠投資人一個補償回合；大跌後可以反彈，也可以盤整或續跌。把「常見路徑」升格成「必然結果」，就是賭徒謬誤。\n3. 後見之明偏誤：漲出來後，預測忽然都存在過\r#\r#4437 · 09:36 · u0731-1631：「早就說過 七月保底42000 八月上看五萬 沒人聽」\n完整推文讓這種句型可以被檢查：同一合成代號在本篇 09:36 以前沒有留下「七月保底 42,000、八月五萬」的可驗證預測。這不代表發言者在別處沒說過，但就這份資料而言，不能把事後自述當成事前紀錄。預測必須先留下時間戳、門檻與期限，才有資格談命中。\n4. 權威捷思與因果跳躍：長老買，所以現在就該買\r#\r#2930 · 09:12 · u0731-0266：「昨天長老都出手表態了，這個點還不買？」\n「八大長老」是對公股行庫相關資金或買賣資訊的市場俗稱。即使前一日確有公股相關買超，也只能證明某個彙整口徑下的交易結果；它不能直接證明政府下令、最終受益人身分，更不能推出今天任何價位都值得買。把代理訊號當成權威背書，是兩段尚未證明的跳躍。\n5. 跨市場等號：台指期很強，所以今晚美股一定漲\r#\r#8857 · 11:54 · u0731-1911：「期貨很明顯告訴你晚上美股會大漲了 不要不信」\n市場之間會互相影響，但相關不等於一比一翻譯。台股白天、台指期、美股期貨與美股現貨的交易時間、參與者、權重與新資訊都不同。把其中一個市場的當下方向直接改寫成另一市場尚未發生的確定結果，是錯誤類比加過度確定。\n6. 技術指標決定論：站回均線，這把就穩了\r#\r#9794 · 12:32 · u0731-1953：「居居一天收復所有均線 這把穩了」\n站回均線是可以驗證的價格描述；「穩了」則是尚未驗證的未來判斷。均線由歷史價格計算，本身不會阻止下一根 K 線跌回去。若沒有持有期間、停損點或失效條件，技術線位只是把不確定性換成一個看起來精確的錨。\n7. 二分法：台積電沒漲停，就算空蛙贏\r#\r#10468 · 12:39 · u0731-1183：「今天台積沒收漲停算空蛙贏」\n市場結果不是只有「多軍全勝」和「空軍全勝」兩格。台積電最後實際收在漲停附近的 2,425 元、上漲 9.98%，但即使只漲 8%，也不能只靠「是否漲停」裁定所有人的損益。進場價、工具、槓桿、時間與部位不同，陣營口號不是損益表。\n8. 組合謬誤與相對剝奪：0050 漲停，其他股票都該丟\r#\r#11298 · 12:51 · u0731-1730：「0050漲停 沒漲停的股票是不是要蛋雕了」\n0050 與個股的權重、產業、風險及當日資訊不同；「代表性商品漲停」不會推出每個成分或每檔持股都應漲停。今天上市股票有 892 檔上漲，證明廣度不差；但也只有 113 檔漲停。用最強的基準要求每一個部位同步，容易把相對落後誤認成投資論點失效，進一步觸發追高換股。\n一天之內，敘事如何翻面\r#\r7/29 的留言容易把下跌畫成直線；7/31 的留言又把上漲畫成直線。方向相反，推理骨架卻很像：\n先看到一根極端 K 線。 再挑一個最順手的原因：美股、長老、融資、均線或某位老師。 把原因擴張成唯一解釋。 最後把今天延伸成明天、下週與年底。 真正困難的不是替已發生的 7.98% 找故事，而是在下一根 K 線出現以前，留下可以被判錯的條件。\n圖 2 · 把盤中金句改寫成可檢驗命題1有明確標的，不只寫「大盤」或「會噴」2有事前時間戳，不用收盤後的「早就說」3有期限，例如下一交易日收盤或一週4有數值門檻，例如收盤高於 43,5005有失效條件，錯了知道在哪裡認錯6有替代解釋，不把單一新聞當唯一原因7有基準，知道是在比加權指數、0050 還是自己的成本 已具備 0 / 7圖說：勾得越多，越接近能被資料推翻的命題；全部沒勾時，通常只是情緒標語。 2026 年截至 07/31 的前十大上漲日\r#\r排名 日期 收盤報酬率 漲跌點數 收盤 1 2026-07-31 +7.98% +3,186.45 43,119.75 2 2026-04-08 +4.61% +1,531.56 34,761.38 3 2026-04-01 +4.58% +1,451.83 33,174.82 4 2026-05-04 +4.57% +1,778.51 40,705.14 5 2026-07-21 +4.20% +1,783.17 44,232.87 6 2026-03-11 +4.10% +1,342.32 34,114.19 7 2026-05-21 +3.37% +1,347.39 41,368.21 8 2026-05-25 +3.26% +1,376.43 43,644.40 9 2026-04-24 +3.23% +1,218.25 38,932.40 10 2026-06-15 +2.78% +1,227.95 45,396.99 7/31 不只是把原第 1 名往下推一位；它比原第 1 名 4/8 的 4.61% 多了 3.37 個百分點。但「紀錄很極端」只描述今天，不替明天投票。\n資料口徑與限制\r#\r指數 OHLC、漲跌百分比、市場廣度與個股收盤資料取自臺灣證券交易所 2026/07/31 收盤資料。 年度排名以證交所各月「發行量加權股價指數歷史資料」逐日收盤計算：當日收盤 ÷ 前一交易日收盤 − 1。 PTT 資料是 14:29 的 BBS 快照；晚到推文可能使最終留言總數增加，但不改變本文引用樓層。 留言只能顯示公開討論中的句型，不能代表全體投資人、實際持倉或真實交易動機。 謬誤分類用來檢查論證，不是投資建議，也不是對匿名發言者的人格判斷。 官方來源：證交所 2026/07/31 每日收盤行情 · 證交所 2026/07 發行量加權股價指數歷史資料\n","date":"2026年7月31日","externalUrl":null,"permalink":"/zh-tw/stocks/taiex-chat-0731/","section":"股票","summary":"","title":"「一天漲回 3,186 點」：13,983 則盤中留言的謬誤考古（2026/07/31）","type":"stocks"},{"content":"","date":"2026年7月31日","externalUrl":null,"permalink":"/zh-tw/tags/2026%E6%BC%B2%E5%B9%85%E5%89%8D%E5%8D%81/","section":"Tags","summary":"","title":"2026漲幅前十","type":"tags"},{"content":"","date":"2026年7月31日","externalUrl":null,"permalink":"/zh-tw/tags/ptt/","section":"Tags","summary":"","title":"PTT","type":"tags"},{"content":"","date":"2026年7月31日","externalUrl":null,"permalink":"/zh-tw/tags/%E5%8F%B0%E8%82%A1/","section":"Tags","summary":"","title":"台股","type":"tags"},{"content":"","date":"2026年7月31日","externalUrl":null,"permalink":"/zh-tw/tags/%E7%9B%A4%E4%B8%AD%E9%96%92%E8%81%8A/","section":"Tags","summary":"","title":"盤中閒聊","type":"tags"},{"content":"","date":"2026年7月31日","externalUrl":null,"permalink":"/zh-tw/stocks/","section":"股票","summary":"","title":"股票","type":"stocks"},{"content":"","date":"2026年7月31日","externalUrl":null,"permalink":"/zh-tw/categories/%E8%82%A1%E7%A5%A8%E8%A7%80%E5%AF%9F/","section":"Categories","summary":"","title":"股票觀察","type":"categories"},{"content":"","date":"2026年7月31日","externalUrl":null,"permalink":"/zh-tw/tags/%E8%A1%8C%E7%82%BA%E5%81%8F%E8%AA%A4/","section":"Tags","summary":"","title":"行為偏誤","type":"tags"},{"content":"這是「2026 台股前十大下跌日」系列第 5 名。排名口徑是 2026/01/01 至 2026/07/29 的加權指數單日收盤報酬率，不是盤中最大跌幅。\n早盤短暫摸高後一路跌破四萬點，最低見 39,384.85，尾盤才把收盤拉回 40,039.18。整數關卡、抄底與國安基金敘事在同一天反覆切換。\n這天發生了什麼\r#\r數值 前收 41,603.36 開盤 41,491.48 最高 41,698.39 最低 39,384.85 收盤 40,039.18 漲跌 ▼ 1,564.18（-3.76%） 2026 跌幅排名 第 5 名 BBS 全量留言\r#\rPTT 網頁版遇到推爆長文會省略中段；這裡走 BBS 層抓到 19,756 則，不是網頁殘片。\n推文型態 數量 推 11,354 噓 1,432 → 6,970 合計 19,756 留言最密集的時段是 09:00（4,700 則）。\n時段 留言數 相對量 08:00 855 ███ 09:00 4,700 ████████████████ 10:00 4,471 ███████████████ 11:00 3,528 ████████████ 12:00 3,636 ████████████ 13:00 2,546 █████████ 14:00 17 █ PTT 原文：AID 1gQKcMvd 匿名化原始檔：純文字 · JSONL 8 種謬誤與偏誤\r#\r以下是語句模式的分類，不是對留言者作人格診斷。盤中閒聊包含反串、迷因與情緒宣洩；引文只能證明這句話出現過，不能證明作者真的照著交易。\n1. 單因謬誤：把市場縮成一個黑手\r#\r市場同時包含外資、內資、避險、被動資金與個別公司消息；把整段價格路徑只歸因於一個有意志的角色，故事會很順，證據卻通常不夠。\n#5680 · 07/29 10:01 · user7158：操你媽政府還不出來護盤？想搞爛台灣經濟嗎？幹\n#13259 · 07/29 11:55 · user7159：正2大媽快抄底啊\n2. 賭徒謬誤：跌多了就「該」彈\r#\r先前已經跌多少，不會自動提高下一分鐘上漲的機率。價格可以很便宜，也可以在缺乏新資訊時繼續變便宜。\n#6460 · 07/29 10:10 · user7160：記住叫你抄底 買正二 正五的ID\n#15415 · 07/29 12:31 · user7161：天天都有甜甜價，抄底！！！\n3. 錨定效應：月線、季線與整數都是答案\r#\r技術位置可以是風險管理參考，但把單一價位當成必然反轉點，會忽略波動、成交量與事件條件。\n#7964 · 07/29 10:36 · user7162：十年線對稱跌幅 這只是剛開始\n#15448 · 07/29 12:32 · user7163：腦殘蛙: 46000肛底=\u0026gt;45000肛底=\u0026gt;44000肛底=\u0026gt;肛我吧\n4. 直線外推：今天的斜率一路畫到明天\r#\r人腦很容易把眼前最強烈的方向延伸出去；市場真正困難之處，正是斜率會在新資訊出現時改變。\n#8230 · 07/29 10:41 · user7164：明天就斷頭了 不怕\n#16789 · 07/29 12:54 · user7165：明天四萬二 穩了 正5教跟我衝\n5. 從眾效應：多蛙、空蛙與「大家都知道」\r#\r群體標籤能快速製造安全感，也會把不同持倉、期限與風險承受度的人壓成同一種對手。\n#5325 · 07/29 09:57 · user7166：是不是沒人敢笑小兒啦\n#14238 · 07/29 12:10 · user7167：大家別慌台幣貶爛會有套蛙說這是利多\n6. 後見之明偏誤：收盤後每個人都早知道\r#\r結果出現後，原先的多種可能性會從記憶裡消失。真正可檢驗的不是「早說」，而是事前是否留下方向、期限與失效條件。\n#4917 · 07/29 09:53 · user7027：老蘇菩薩 提前一個月多預告還被酸 先知果然孤獨\n#14224 · 07/29 12:10 · user7115：早說了已經是流動性的系統風險了，韓國仔只是背鍋\n7. 類比謬誤：把日韓、美股或夜盤直接翻譯\r#\r跨市場確實相關，但交易時段、權重、匯率與事件曝險不同；相關不等於可以一比一複製漲跌。\n#2713 · 07/29 09:19 · user7168：夜盤都是假的 內資早上砍股 還不是 A下去XD\n#14986 · 07/29 12:23 · user7169：又一個看夜盤漲就以為會漲的憨多XDD\n8. 災難化與全有全無：不是 V，就是歸零\r#\r這比較接近認知扭曲而非形式邏輯謬誤：把連續的風險壓成生或死兩個選項，會讓部位管理退化成情緒口號。\n#3395 · 07/29 09:27 · user7170：救命!不要斷我頭!!!\n#11439 · 07/29 11:20 · user7171：不要再A了 救命\n這份考古不能證明什麼\r#\r它不是情緒指標回測，也沒有證明某類留言能預測下一根 K 線。 關鍵字分類是可重現的抽樣入口，不是對全部 19,756 則留言做唯一正解標註。 同一句話可能同時包含多種偏誤；為了可讀性，每則引文只放在一個小節。 帳號已置換為合成代號（user0001 之類），與原帳號無對應關係、不可回推；保留樓層與時間，是為了讓讀者能回到匿名化原始檔核對。 2026 年截至 07/29 的前十大下跌日\r#\r名次 日期 收盤跌幅 跌點 文章 1 2026-07-17 -6.47% -2,953.71 「跌到收盤才停」 2 2026-07-28 -4.65% -2,030.83 「這次一定會 V？」 3 2026-03-09 -4.43% -1,489.12 「千點低開之後」 4 2026-03-04 -4.35% -1,494.77 「月線會救嗎」 5 2026-07-29 -3.76% -1,564.18 「四萬點保衛戰」（本篇） 6 2026-06-26 -3.64% -1,683.50 「指數在跌，嘴在做什麼」 7 2026-06-08 -3.48% -1,568.16 「夜盤只是誤會？」 8 2026-06-10 -3.31% -1,478.90 「反彈第二天又殺」 9 2026-07-24 -2.67% -1,195.97 「日韓跌，台股就會？」 10 2026-03-23 -2.45% -821.38 「明天會噴回來？」 資料口徑\r#\r指數排名與 OHLC：Yahoo Finance ^TWII 日線快照，抓取日 2026/07/29；排名用相鄰交易日收盤價計算。臺灣證券交易所亦提供發行量加權股價指數歷史資料供核對。 留言：PTT Stock 板當日「盤中閒聊」，以 BBS 層完整抓取；本篇產生前已核對 AID、標題與留言總數。 隱私：公開原始檔與本文引文的帳號一律置換為合成代號，不保留原字元、不可回推。同一個代號在所有檔案裡指同一人，跨日比對仍然成立。 ","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/stocks/taiex-chat-0729/","section":"股票","summary":"","title":"「四萬點保衛戰」：19,756 則盤中留言的謬誤考古（2026/07/29）","type":"stocks"},{"content":"這不是「史上前十」，而是 2026/01/01 至 2026/07/29 的年內排名。口徑固定後，才不會因為今天很痛，就把「最大」當成沒有分母的形容詞。\n排名\r#\r名次 日期 收盤跌幅 跌點 文章 1 2026-07-17 -6.47% -2,953.71 「跌到收盤才停」 2 2026-07-28 -4.65% -2,030.83 「這次一定會 V？」 3 2026-03-09 -4.43% -1,489.12 「千點低開之後」 4 2026-03-04 -4.35% -1,494.77 「月線會救嗎」 5 2026-07-29 -3.76% -1,564.18 「四萬點保衛戰」 6 2026-06-26 -3.64% -1,683.50 「指數在跌，嘴在做什麼」 7 2026-06-08 -3.48% -1,568.16 「夜盤只是誤會？」 8 2026-06-10 -3.31% -1,478.90 「反彈第二天又殺」 9 2026-07-24 -2.67% -1,195.97 「日韓跌，台股就會？」 10 2026-03-23 -2.45% -821.38 「明天會噴回來？」 目前收錄\r#\r10 個交易日共 127,280 則匿名化 BBS 全量留言。 06/26 使用原本的互動圖專文；其餘 9 篇現役前十日期專文涵蓋 113,348 則留言。 07/29 新進榜後排第 5 名；原第 10 名 03/31 移至第 11 名，專文仍保留作比較。 每篇都用相同的 8 類框架：單因謬誤、賭徒謬誤、錨定、直線外推、從眾、後見之明、跨市場類比、災難化。 每一則引文都保留樓層與時間，並可回到公開的匿名化 JSONL 核對。 閱讀方式\r#\r這個系列不是要嘲笑誰「不理性」。盤中閒聊本來就有反串、迷因與壓力釋放；真正值得看的，是同一套句型如何在不同日期重複出現：\n價格先動，故事再補上。 故事會隨盤勢翻轉，但說話時常帶著「一定」。 收盤之後，多種可能性會被改寫成「早就知道」。 資料口徑\r#\r排名依 Yahoo Finance ^TWII 日線相鄰收盤報酬率計算，快照日 2026/07/29。 PTT 留言走 BBS 層全量抓取，不採用會截斷長文的網頁顯示結果。 分類是透明規則抽樣，不是模型對每個帳號作心理診斷。 對外公開的帳號一律置換為不可回推的合成代號。 ","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/stocks/taiex-top10-drops-2026/","section":"股票","summary":"","title":"2026 台股前十大下跌日：PTT 盤中留言謬誤考古","type":"stocks"},{"content":"","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/tags/2026%E8%B7%8C%E5%B9%85%E5%89%8D%E5%8D%81/","section":"Tags","summary":"","title":"2026跌幅前十","type":"tags"},{"content":"","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/tags/%E5%8A%A0%E6%AC%8A%E6%8C%87%E6%95%B8/","section":"Tags","summary":"","title":"加權指數","type":"tags"},{"content":"","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/tags/%E6%8A%95%E8%B3%87%E5%BF%83%E7%90%86/","section":"Tags","summary":"","title":"投資心理","type":"tags"},{"content":"","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/categories/%E7%9B%A4%E4%B8%AD%E8%A7%80%E5%AF%9F/","section":"Categories","summary":"","title":"盤中觀察","type":"categories"},{"content":"","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/tags/ai-agent/","section":"Tags","summary":"","title":"AI Agent","type":"tags"},{"content":"","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/tags/%E5%B7%A5%E4%BD%9C%E6%96%B9%E5%BC%8F/","section":"Tags","summary":"","title":"工作方式","type":"tags"},{"content":"我寫過不少 AI agent 怎麼壓縮工作的文章，這篇講另一邊。\n我曾經形容過，假使哪天失去它，我也會感覺如同殘廢一般，而那不是打比方，是字面上意義的殘廢。\n一、對 AI 的依賴是不可逆的\r#\r當你逐漸將自己的腦、手、眼睛替換成機器，就會出現一個新的風險。當 agent 不可用的時候，比如斷網、服務中斷、政策禁止、供應商漲價，或者模型改版把原本的流程弄壞，我的產出會下降得比其他人劇烈。\n沒有裝義肢的人，那天跟平常一樣，而裝了義肢的人回不到裝之前的狀態，因為原本的肌肉已經不練了。\n我把自己的下限，託付給了我無法掌握的機器。\n這件事沒有好的解法，只有幾個緩解方向，比如不把單一供應商當成唯一途徑、關鍵流程保留一份人類可執行的版本、定期問自己如果今天它掛了我還做不做得動。\n但這些都只是緩解，真正的事實是這個交換我已經做了，而它是單向的。\n二、我的能力在萎縮\r#\r這一點比第一點更難承認。\n如同擔任主管職之後，第一線的技能跟現場熟悉度會逐漸退化，當許多事情委派出去，事情本身的狀況我只能基於信任，而不是因為親手做過而有信心。\n這兩種知道的差別，只有在出事的時候才會顯現。親手做過的人知道哪裡容易壞、知道那個看起來沒問題的地方其實很脆弱、知道某個數字不對勁，而基於信任的人只知道它應該是對的。\n我還沒找到方法在不放棄槓桿的前提下保住這個，目前的做法只有一個，就是選幾件事故意自己做，不是因為那樣比較快，是為了不要失去判斷力，因為驗收者一旦失去判斷力，就只剩下按鈕可以按。\n而這是有代價的，它確實比較慢，我是在用效率換一個我不確定值多少的東西。\n三、這件事不只發生在我身上\r#\r前兩條是我個人的問題，可以自己承擔，這一條不是。\n代價是我、還有整個社會，確實再也回不去以前的工作模式了。\n不管這對每個個體是好是壞，我們終將會撞上那堵牆，當所有人的產能都被墊高，八小時這個制度還剩下什麼意義。\n八小時工作制不是自然律，它是一百多年前談出來的一個數字，基礎是那個年代一個人一天能做多少事，當這個基礎被整體墊高好幾倍，制度不會自動跟著調整，它只會默默地把標準提高，然後所有人繼續上班八小時。\n這一條我沒有答案。\n我唯一確定的是，假裝它不存在對誰都沒有好處，而目前絕大部分關於 AI 生產力的討論，都停在你可以省下多少時間這一步，不往下問那句：省下來的時間，最後會是誰的？\n那為什麼還做\r#\r因為選擇不做並不會讓這些代價消失，只會讓你在承擔代價的同時，連好處都拿不到。\n第一條跟第二條是我用眼睛看得見的價格，我付了，第三條不是我付得起的，也不是我一個人能決定的。\n我把它寫出來，是因為一份只講好處的紀錄不值得相信，如果你看完前面那些文章覺得這一切太美好，這篇就是我欠你的那一半。\n如果你正在替團隊評估要導入到多深，這三條值得先擺在桌上談，尤其第二條，它通常在導入兩年後才開始收費。\n如果你的情況我幫得上忙，來聊聊：範疇、報價、怎麼開始 →\n","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/posts/the-price/","section":"部落格","summary":"","title":"我的能力正在萎縮——用 AI agent 一年多之後，還沒解決的三個代價","type":"posts"},{"content":"","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/tags/%E8%81%B7%E5%A0%B4/","section":"Tags","summary":"","title":"職場","type":"tags"},{"content":"","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/tags/%E5%85%A5%E9%96%80/","section":"Tags","summary":"","title":"入門","type":"tags"},{"content":"也許你覺得自己的工作環境根本不適用，或者覺得自己已經使用到極限了，以下是拓展使用情境的一些參考方式跟步驟。\n它們都很基本，但這正是重點，大部分人卡住的地方不在進階技巧，在最前面那兩步沒做。\n先講五件事。\n⚠️ 開始之前\n請對 AI 的後果與產出自行負責。 請妥善備份你的資料。 關於在工作中使用 AI、以及機敏資料的處置，請自行查詢相關規範跟工具，例如本地大模型1、guardrails、compliance gate、sandbox、hook2 等等。你的公司很可能有明文的 AI 使用政策，先去讀它或請教相關部門。 不用覺得厭煩或困難：你可以先在安全的範圍內使用 AI 提高生產力，自然就會有時間與興趣去研究這些看起來陌生的東西。 再說一次，請對 AI 的後果與產出自行負責。 七條\r#\r將所需資料數位化。不管是純文字、投影片、email、照片都可以，只要是數位化的資訊 AI 都讀得到，那怕是自己覺得不重要的資料，對 AI 也很有幫助，就算是我們習以為常的東西，對它都是非常有意義的背景資訊。若資料具機敏性，請先確認自己的帳號已設定不分享，並確認這符合公司規範。\n把資料放在 agent 的工作目錄。承上，針對類似的任務，盡可能把所有資料放進當前 agent 所運行的資料夾，並在對話中告訴它：「如果有需要，請詢問我或採訪我。」\n不要設立邊界，任何事都試著請 AI 執行。想像自己除了打字給 agent 之外，完全無法使用鍵盤滑鼠，那怕是自己認為很簡單、很快就能做完的任務，也試著讓 agent 去執行，並在過程中觀察它。\n想像自己即將離職，你的下一手要如何接手這個業務？把類似的資訊提供給 agent，它不會辜負你的付出。\n先讓它「看得見」。如果你的資料散落在各處，比如雲端硬碟、行事曆、CRM、資料庫等等，想讓 agent 同時讀取或操作不同系統的資料，請詢問它這些不同的系統，能不能透過 SSH 或 token 連線？能不能使用 API？能不能匯出資料？\n一開始請把它當成一個無所不知的瞎子，並詢問它要如何才能讓這些東西被它看見、被它取得。如果你完全不懂，可以將這整段話複製貼上給它。\n讓雪球滾起來。以上都是能夠滾雪球的建議，只要你開始嘗試，雪球很快就會越滾越大，過沒多久，隨著生產力提升而來的成就感跟多巴胺，自然會促使你不斷完善 agent 的工作流程。\n先從無傷大雅的地方練。一開始害怕在實際工作環境使用 AI 的話，比如害怕它刪除你的資料，或者寄出錯誤的郵件，這種擔心是合理的，在不熟悉的情況下，AI 的確不會百分之百照我們想的運作。可以找自己有興趣、出錯也沒關係的領域先嘗試，我自己是在下班後做 side project，這讓我對當前 AI 能力的邊界相對熟悉。\n這七條裡最重要的是第五條\r#\r前四條大部分人都做得到，第五條才是分水嶺。\n我認為「無所不知的瞎子」是對現階段 agent 最準確的描述，它什麼都懂，但預設什麼都看不見，看不見你的工單系統、行事曆，也看不見只存在某台機器上的目錄。\n多數人對 AI 失望，是因為他們在瞎子的狀態下評估它的能力，把問題複製貼上之後得到一個泛泛的答案，於是就認為這個東西沒什麼用。\n問題通常不是 agent 不夠聰明，而是你還沒有給它眼睛。所謂給它眼睛，就是讓它能夠連線、讀取 API 或取得匯出的資料，這通常不是提示詞的問題，而是工程問題，也因此真正的門檻往往落在 IT 那一側。\n起手式之後會遇到的牆\r#\r老實說，照著這七條做，多數人會在三個月內遇到同一堵牆，你會知道有東西可以自動化，但不確定該切在哪一層。\n抽太淺 agent 幫不上忙，抽太深你會花三個月蓋一個沒人用的框架，這個判斷沒辦法外包給 AI，它是整件事裡少數還完全屬於人類的部分。\n延伸閱讀：開始之前，也值得先看一次帳單。〈我的能力正在萎縮〉講的是我到現在還沒解決的三個代價 →\n如果你已經走完這七條，卡在那道牆前面，那正好是我在做的事。\n那道牆後面的工作我可以接：範疇、報價、怎麼開始 →\n本地大模型：在自己的電腦或公司內部伺服器上運行的語言模型，資料不會離開內網。能力通常不如雲端旗艦模型，但在機敏環境是唯一可行的選項。\u0026#160;\u0026#x21a9;\u0026#xfe0e;\nguardrails / compliance gate / sandbox / hook：四種不同層次的護欄。guardrails 限制 AI 能說能做甚麼；compliance gate 在動作執行前檢查是否符合規範；sandbox 把 AI 關在一個弄壞了也沒關係的隔離環境；hook 則是在特定動作前後插入你自己的檢查程式。\u0026#160;\u0026#x21a9;\u0026#xfe0e;\n","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/posts/agent-starter-kit/","section":"部落格","summary":"","title":"把它當成一個無所不知的瞎子——AI agent 起手式七條","type":"posts"},{"content":"","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/tags/%E7%94%9F%E7%94%A2%E5%8A%9B/","section":"Tags","summary":"","title":"生產力","type":"tags"},{"content":"","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/tags/%E8%87%AA%E5%8B%95%E5%8C%96/","section":"Tags","summary":"","title":"自動化","type":"tags"},{"content":"有些工作只要 AI 能夠獨自運行、不需要人工介入，就可以整段交出去放著跑，那怕它要跑四小時或八小時。而這種形狀的工作還有一個特性，它可以被複製。\n於是我可以在數個、十幾個不同的會話裡同時交辦不同的事情，就像是可以隨時產生新的工讀生或員工一樣，我後來給它一個暱稱，叫無限的工讀生。\n這個詞聽起來很美好，但實際使用一段時間之後，我發現它真正改變的不是我的產能，是我的職位。\n工讀生變多，你就不再是做事的那個人\r#\r當十件事同時在跑，你不可能還親手做其中任何一件，你能做的只剩下四件：\n正確地下指示 過程中補上它缺的資訊 在它走偏的時候中斷它 檢查結果 這四件事情合起來，其實就是管理。\nAI 剛開始比較像義肢，是身體的延伸，但義肢會增生，最後義肢變成分身，長成了一支團隊，而我連帶接收了帶團隊的所有好處及壞處。\n工作沒有變少，是我把自己升級成了驗收者。\n驗收的疲勞被嚴重低估\r#\r驗收的疲勞，是我最想講、也最少有人講的一件事。執行本身的時間減少，並不代表工作時間也會跟著減少，如同主管乍看沒有在做事一樣，許多工作只是換成比較看不見的形式。\n上面那四件事，每一項都需要專注力，而且那跟親手執行時的專注力不太一樣，你必須理解得夠深才審核得動，卻沒有親手做過的過程可以幫你建立信心。\n還有一個更麻煩的地方，許多需要高度專注力或是心流的工作交給 AI 之後，也意味著人只剩下需要負責任、以及最無趣的那部分可做。 原本想通了、跑起來了所帶來的多巴胺，有很大一部分也一起被外包出去，最後留下來的是責任跟校對。\n所以「同時開十個」不是免費的\r#\r無限的工讀生有一個很現實的上限，而那個上限不在 AI 那邊，在你這邊：\n你能同時追蹤幾件事的脈絡？ 你能在一件事跑偏的第幾分鐘察覺？ 你能在不重做一遍的前提下，判斷結果對不對？ 我的經驗是，能穩定並行的數量遠低於技術上能開幾個，瓶頸從手換到了注意力。\n這也是為什麼「AI 讓一個人抵十個人」這種說法我一向持保留態度，它讓一個人能發包給十個人，這跟一個人能做完十個人的事情是完全不同的兩件事，而中間那道差距叫做驗收能力。\n如果你要開始並行\r#\r三個實際有用的做法：\n一次只開你審得動的數量，開十個然後每個都只看一眼，等於十個都沒做。 把驗收標準寫在交辦的時候，不是收貨的時候，你事後才想到的標準，它不可能猜得到。 留下可以重跑的痕跡，你會忘記三天前那個會話裡發生過什麼，它不會，但前提是你讓它寫下來。 延伸閱讀：「可以放著跑」的工作為什麼會滾成複利，以及省下來的那段時間該去哪裡。〈省下時間之後才是分水嶺〉 →\n如果你正在把流程交給 agent，卻發現自己每天都在追十件半成品，那問題通常不在 agent，在交辦的規格。\n規格這件事我可以幫你寫：範疇、報價、怎麼開始 →\n","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/posts/infinite-interns/","section":"部落格","summary":"","title":"無限的工讀生——當你能同時交辦十件事，瓶頸就換成你自己","type":"posts"},{"content":"大部分人用 AI 省下時間之後，那二十分鐘很快又拿去回信、開會，或者處理下一件被追著跑的事情，隔天一切照舊。\n會有這種結果，不一定是 AI 用得不夠好，而是飛輪根本沒有轉起來，真正拉開差距的除了省下多少時間，也包括省下的時間最後去了哪裡。\n圖 1 · 飛輪 AI 接手 必要事項 每天省下 X 分鐘 ★ 把 X 分鐘 投入優化 更多環節 被自動化 兩三個月後 複利 圖說：關鍵不在「省下 X 分鐘」，在那條把 X 分鐘送回起點的箭頭（★）。若省下的時間立刻被新工作填滿，這個輪子轉不起來。 一、複利：唯一真正重要的那條箭頭\r#\r每天省下 X 分鐘，把這 X 分鐘再拿去優化自己的工作流程，兩三個月後 X 就會在複利效應下變成一個驚人的數字。\n這句話寫出來像廢話，但它是整張圖裡唯一有難度的一步，因為把時間投回優化，當下沒有任何人會看見，沒有工單被結掉，沒有客戶被安撫，也沒有主管知道，它的報酬全部在未來。\n所以飛輪的瓶頸不是技術，是你能不能容忍一段沒有產出的時間。\n二、委派：付出的不再是勞力，而是等待\r#\r某些事項，即使 AI 需要運行很久，那怕四小時或八小時，只要它能夠獨自運行、不需要人工介入，這類工作對我來說就可以從此 offload。\n這件事本身還是需要時間，但我付出的已經不再是勞力，而是等待。 勞力有上限、會累，也會占用我的注意力，等待則不會。\n一件事只要能夠被改造成可以放著跑的形狀，它的成本就從我的體力跟注意力，轉成機器自行運行的時間。\n三、更小的心流窗口\r#\r至於那些仍然需要人工介入的工作，假設某個任務我要花半小時讀程式碼或論文，再花半小時做相關的輸出，這一小時全都處在不被打擾的心流裡，意即我需要一個完整的一小時窗口才排得進去。\n上班族最缺的從來不是一小時，是保證連續的那一小時。\n既使 AI 不能全部代勞，它也能大幅壓縮需要心流的那一段，讓同樣的任務可以被塞進更小的窗口。對一個會被電話跟訊息打斷的工作者來說，這件事的價值遠比省時間大，它把原本排不進去的任務，變成排得進去的任務。\n四、不要劃邊界\r#\r很多人使用 AI 的方式，是先在心裡畫一條線，線的這邊給它做，線的那邊「它做不到」，然後花時間嫌棄它做不到的那部分。\n我的做法相反，優先做完當下所有 AI 能做的事，並且優化它。\n正因為可以同時開好幾個會話交辦不同的事，所以應該把所有它做得到的事情都拿出來做、拿出來優化，等你花 N 個禮拜優化完當前的流程，下一個版本的能力又進步了，你會找到更多它能做的事，並從此以後不斷跟著最新版本一起優化。\n那條線每三個月就會移動一次，畫線的人會一直在跟三個月前的模型吵架。\n五、它會跟你一起進化\r#\r每當你使用 AI agent，它會不斷記錄發生過的事情跟你講過的話，而人在使用中也會越來越理解這具 agent 此刻的邊界跟人格特質。\nAI 是那種越用越合手的義肢。\n這也是上一條說應當優先做所有 AI 能做的事情的原因，磨合本身就是投資，你今天多花的二十分鐘不是浪費在一次性的任務上，是花在一個會累積的介面上。\n輪子轉不起來的時候\r#\r回到圖上那個 ★，如果你省下來的時間會立刻被填進新的工作，這五條全部失效，不是因為 AI 不夠強，是因為飛輪缺了把動能送回起點的那一段，你只是在替別人加速，自己原地踏步。\n這一條不是技術問題，是環境問題，而環境問題通常不是靠更會下提示詞解決的。\n延伸閱讀：如果你還沒開始，卡住的通常不是觀念，是最前面那兩步沒做。〈起手式七條〉是具體的做法 →\n如果你手上的流程已經卡在知道可以優化，但一直挪不出那段沒有產出的時間，那段時間可以外包。\n這種建置可以委託我做：範疇、報價、怎麼開始 →\n","date":"2026年7月29日","externalUrl":null,"permalink":"/zh-tw/posts/flywheel-after-saving-time/","section":"部落格","summary":"","title":"省下時間之後才是分水嶺——AI agent 的飛輪要怎麼轉起來","type":"posts"},{"content":"近來 AI 各廠牌的模型能力還在快速進步中。一般人直觀感受到的是它的知識儲備、邏輯等思維能力；前沿的科學家關注的是 AI 的智力天花板，AI 能不能達到 AGI（通用人工智慧，artificial general intelligence）？\n但對我的工作來說，最有幫助的不是它的腦有多強，而是它已經可以勝任數位世界的手眼協調，以及相關的操作能力。大部分人只把 AI 當成網路上的顧問或其他任何角色跟它聊天，我做的是把 AI 當成工讀生跟外包廠商，把它請到我實際需要完成工作的地方。\n即使 AI 的大腦與我相當、甚至稍差，它還是比我更適合處理數位世界的工作，因為它的閱讀速度跟打字速度遠超人類。\n在數位世界裡，比起替代大腦，它更像是手跟眼睛的義肢。\n這一點很少被人提及，卻是我實際受益最多的一條。當大家還在爭論 AI 的思考能力，它的眼睛跟手已經在替我做事了。\n圖 1 · 義肢光譜🧠腦判斷 · 取捨 · 責任人類仍較強👁眼閱讀速度AI 壓倒性✋手打字 · 操作 · 不會累AI 壓倒性圖說：大家都在爭論 AI 的腦有沒有超過人類，而我節省的時間幾乎全部來自它的手跟眼。 以下是兩個實例。\n實例一：迭代型工作\r#\rbefore / after 100 分鐘→10 分鐘10× 假設有幾十台機器要登入進去查看硬碟使用情況。我可以寫個腳本執行，這在沒有 AI 的年代也不算太耗時。\n但我依然需要列出到底是哪幾台；假設中間有三台無法連線，這過程至少會被中斷三次；中間有許多對人類來說耗費大量時間的動作，包括切換視窗、整理名單、複製貼上等等⋯⋯這過程中還需要肉眼盯著畫面看有沒有出錯。\n這並沒有什麼技術含量，主要是侷限在目前還沒有腦機介面，人類的思考與電腦的輸入輸出之間，需要花過多的時間閱讀及複製貼上。\n這種情況下，AI 主要是我的手跟眼的義肢，腦只是順便也外包了。\n實例二：短期記憶\r#\rbefore / after 60 分鐘→10 分鐘6× 有時候我們的資源與人事時地物之間有許多對應關係：設備、貨物、客戶、廠商、日期、相關處置等等。在需要歸檔、或需要把它們的關係捋清楚的時候，人必須全心投入，把相關資訊暫存在腦袋裡，再轉化成日後好回憶、好處理的結果。\n過程中我們需要手眼協調地不斷切換分頁、視窗，眼睛必須不停對焦，經常還需要打字，可謂是耗盡了短期記憶與注意力這兩項最珍貴的資源。\n工作環境能安靜處理的還算好，有些場景會被臨時插進來的訊息或電話打斷，只能一直延宕。\n這正是 AI 最擅長的一類，它可以編輯 Excel、記事本等檔案，害怕檔案被改壞的不妨複製出一份給它，讓它試一下，觀察它的表現。\n所以我的角色變了\r#\r當手跟眼被外包出去，我做的事情就從執行變成驗收。這不是變輕鬆，驗收本身要花的專注力被嚴重低估，不過對我來說依然是划算的交換，因為判斷還是由我負責，閱讀跟打字則交給更適合處理它們的 AI。\n這個畫面，《攻殻機動隊》早就想像過了：\n圖 2 · 打字的義肢 義體示例：義體化最不浪漫、也最有效的用途是打字。它不會累、不會打錯、視線不需要對焦。圖：自製。 延伸閱讀：上面那句「從執行變成驗收」我講得很快，但驗收的疲勞才是這個交換真正的代價。〈無限的工讀生〉把它整篇拆開講 →\n如果你手上也有一堆「一點技術含量都沒有、但就是很花時間」的數位庶務，比如切視窗、對名單、複製貼上、盯輸出，那些正是最容易交出去的部分。 這也能蓋在你的流程上。我接的就是這種案子：範疇、報價、怎麼開始 →\n","date":"2026年7月28日","externalUrl":null,"permalink":"/zh-tw/posts/hands-and-eyes/","section":"部落格","summary":"","title":"大家都在吵 AI 的腦——我省下的時間，卻來自它的手和眼","type":"posts"},{"content":"這是「2026 台股前十大下跌日」系列第 2 名。排名口徑是 2026/01/01 至 2026/07/29 的加權指數單日收盤報酬率，不是盤中最大跌幅。\n指數跳空後一路走低，只在最低點附近留下很小的反抽。留言卻在「超級大熊」與「一定會 V」之間高速擺盪。\n這天發生了什麼\r#\r數值 前收 43,634.19 開盤 43,221.93 最高 43,221.93 最低 41,565.00 收盤 41,603.36 漲跌 ▼ 2,030.83（-4.65%） 2026 跌幅排名 第 2 名 BBS 全量留言\r#\rPTT 網頁版遇到推爆長文會省略中段；這裡走 BBS 層抓到 14,476 則，不是網頁殘片。\n推文型態 數量 推 8,093 噓 1,046 → 5,337 合計 14,476 留言最密集的時段是 09:00（3,701 則）。\n時段 留言數 相對量 08:00 1,588 ███████ 09:00 3,701 ████████████████ 10:00 2,424 ██████████ 11:00 2,427 ██████████ 12:00 1,763 ████████ 13:00 2,530 ███████████ 14:00 43 █ PTT 原文：AID 1gP_WfVD 匿名化原始檔：純文字 · JSONL 8 種謬誤與偏誤\r#\r以下是語句模式的分類，不是對留言者作人格診斷。盤中閒聊包含反串、迷因與情緒宣洩；引文只能證明這句話出現過，不能證明作者真的照著交易。\n1. 單因謬誤：把市場縮成一個黑手\r#\r市場同時包含外資、內資、避險、被動資金與個別公司消息；把整段價格路徑只歸因於一個有意志的角色，故事會很順，證據卻通常不夠。\n#4071 · 07/28 09:32 · user7146：傻多還不懂小兒要殺出你們的汁來嗎？下殺取量\n#11304 · 07/28 12:40 · user7147：緯創外資看主動衝進去這禮拜就開倒了\n2. 賭徒謬誤：跌多了就「該」彈\r#\r先前已經跌多少，不會自動提高下一分鐘上漲的機率。價格可以很便宜，也可以在缺乏新資訊時繼續變便宜。\n#3819 · 07/28 09:28 · user7148：大媽買菜回來了 準備V回平盤\n#8961 · 07/28 11:26 · user7149：貪狗抄底成功 沒買到的要眼紅了\n3. 錨定效應：月線、季線與整數都是答案\r#\r技術位置可以是風險管理參考，但把單一價位當成必然反轉點，會忽略波動、成交量與事件條件。\n#3120 · 07/28 09:18 · user7132：櫃買要去年線了嗎\n#9566 · 07/28 11:43 · user7150：先說喔 這裡最近支撐是半年線39257現在是長老好心\n4. 直線外推：今天的斜率一路畫到明天\r#\r人腦很容易把眼前最強烈的方向延伸出去；市場真正困難之處，正是斜率會在新資訊出現時改變。\n#6107 · 07/28 10:19 · user7151：貪狗明天漲了就要變臉了\n#12267 · 07/28 13:10 · user7152：今天殺太早 明天再殺比較人道\n5. 從眾效應：多蛙、空蛙與「大家都知道」\r#\r群體標籤能快速製造安全感，也會把不同持倉、期限與風險承受度的人壓成同一種對手。\n#4343 · 07/28 09:38 · user7153：多蛙畢業文+國安雞精+折折道歉=抄底 現在沒半個\n#10556 · 07/28 12:14 · user7150：就一堆人受不了了\n6. 後見之明偏誤：收盤後每個人都早知道\r#\r結果出現後，原先的多種可能性會從記憶裡消失。真正可檢驗的不是「早說」，而是事前是否留下方向、期限與失效條件。\n#4869 · 07/28 09:51 · user7141：一看就知道還要繼續跌\n#10459 · 07/28 12:11 · user7154：這個盤大戶進去玩也是虧爛 早就出國玩了\n7. 類比謬誤：把日韓、美股或夜盤直接翻譯\r#\r跨市場確實相關，但交易時段、權重、匯率與事件曝險不同；相關不等於可以一比一複製漲跌。\n#3113 · 07/28 09:17 · user7083：抄底記憶體跌停的，今晚美股繼續跌，爽蛇\n#12866 · 07/28 13:21 · user7155：日韓被台股拖下來囉\n8. 災難化與全有全無：不是 V，就是歸零\r#\r這比較接近認知扭曲而非形式邏輯謬誤：把連續的風險壓成生或死兩個選項，會讓部位管理退化成情緒口號。\n#2699 · 07/28 09:13 · user7156：可樂賣太便宜了 QQ 價值快歸零了\n#9820 · 07/28 11:49 · user7157：還一堆人喊完蛋了，那就繼續V\n這份考古不能證明什麼\r#\r它不是情緒指標回測，也沒有證明某類留言能預測下一根 K 線。 關鍵字分類是可重現的抽樣入口，不是對全部 14,476 則留言做唯一正解標註。 同一句話可能同時包含多種偏誤；為了可讀性，每則引文只放在一個小節。 帳號已置換為合成代號（user0001 之類），與原帳號無對應關係、不可回推；保留樓層與時間，是為了讓讀者能回到匿名化原始檔核對。 2026 年截至 07/29 的前十大下跌日\r#\r名次 日期 收盤跌幅 跌點 文章 1 2026-07-17 -6.47% -2,953.71 「跌到收盤才停」 2 2026-07-28 -4.65% -2,030.83 「這次一定會 V？」（本篇） 3 2026-03-09 -4.43% -1,489.12 「千點低開之後」 4 2026-03-04 -4.35% -1,494.77 「月線會救嗎」 5 2026-07-29 -3.76% -1,564.18 「四萬點保衛戰」 6 2026-06-26 -3.64% -1,683.50 「指數在跌，嘴在做什麼」 7 2026-06-08 -3.48% -1,568.16 「夜盤只是誤會？」 8 2026-06-10 -3.31% -1,478.90 「反彈第二天又殺」 9 2026-07-24 -2.67% -1,195.97 「日韓跌，台股就會？」 10 2026-03-23 -2.45% -821.38 「明天會噴回來？」 資料口徑\r#\r指數排名與 OHLC：Yahoo Finance ^TWII 日線快照，抓取日 2026/07/29；排名用相鄰交易日收盤價計算。臺灣證券交易所亦提供發行量加權股價指數歷史資料供核對。 留言：PTT Stock 板當日「盤中閒聊」，以 BBS 層完整抓取；本篇產生前已核對 AID、標題與留言總數。 隱私：公開原始檔與本文引文的帳號一律置換為合成代號，不保留原字元、不可回推。同一個代號在所有檔案裡指同一人，跨日比對仍然成立。 ","date":"2026年7月28日","externalUrl":null,"permalink":"/zh-tw/stocks/taiex-chat-0728/","section":"股票","summary":"","title":"「這次一定會 V？」：14,476 則盤中留言的謬誤考古（2026/07/28）","type":"stocks"},{"content":"這是「2026 台股前十大下跌日」系列第 9 名。排名口徑是 2026/01/01 至 2026/07/29 的加權指數單日收盤報酬率，不是盤中最大跌幅。\n日韓、美股與台股被不斷互相比照；但這一天台股自己的路徑仍是開盤最高、收盤接近最低。\n這天發生了什麼\r#\r數值 前收 44,850.81 開盤 44,769.39 最高 44,769.39 最低 43,607.40 收盤 43,654.84 漲跌 ▼ 1,195.97（-2.67%） 2026 跌幅排名 第 9 名 BBS 全量留言\r#\rPTT 網頁版遇到推爆長文會省略中段；這裡走 BBS 層抓到 8,454 則，不是網頁殘片。\n推文型態 數量 推 5,015 噓 637 → 2,802 合計 8,454 留言最密集的時段是 10:00（2,246 則）。\n時段 留言數 相對量 08:00 818 ██████ 09:00 1,698 ████████████ 10:00 2,246 ████████████████ 11:00 1,373 ██████████ 12:00 1,085 ████████ 13:00 1,208 █████████ 14:00 25 █ PTT 原文：AID 1gOh8CiQ 匿名化原始檔：純文字 · JSONL 8 種謬誤與偏誤\r#\r以下是語句模式的分類，不是對留言者作人格診斷。盤中閒聊包含反串、迷因與情緒宣洩；引文只能證明這句話出現過，不能證明作者真的照著交易。\n1. 單因謬誤：把市場縮成一個黑手\r#\r市場同時包含外資、內資、避險、被動資金與個別公司消息；把整段價格路徑只歸因於一個有意志的角色，故事會很順，證據卻通常不夠。\n#3117 · 07/24 10:24 · user7135：這跌幅感覺小兒還沒開始倒貨等下午看買賣超\n#6329 · 07/24 12:08 · user7136：什麼時候落地 韓國政府一定會等盤淡再慢慢放消息測\n2. 賭徒謬誤：跌多了就「該」彈\r#\r先前已經跌多少，不會自動提高下一分鐘上漲的機率。價格可以很便宜，也可以在缺乏新資訊時繼續變便宜。\n#2488 · 07/24 09:58 · user7137：現在抄底可以吧 缸底這麼多次 總該是真的了吧\n#6043 · 07/24 11:56 · user7081：今天最低點都看到了 不會再下去了拉 抄底!!\n3. 錨定效應：月線、季線與整數都是答案\r#\r技術位置可以是風險管理參考，但把單一價位當成必然反轉點，會忽略波動、成交量與事件條件。\n#3240 · 07/24 10:26 · user7138：哎呀怎麼又破季線\n#6425 · 07/24 12:14 · user7033：韓狗站回半年線再度失敗\n4. 直線外推：今天的斜率一路畫到明天\r#\r人腦很容易把眼前最強烈的方向延伸出去；市場真正困難之處，正是斜率會在新資訊出現時改變。\n#3531 · 07/24 10:34 · user7085：快 吃投信豆腐 相信他們一定要拉南電\n#7019 · 07/24 12:48 · user7139：明天再跌一根，看誰還在VV叫\n5. 從眾效應：多蛙、空蛙與「大家都知道」\r#\r群體標籤能快速製造安全感，也會把不同持倉、期限與風險承受度的人壓成同一種對手。\n#2280 · 07/24 09:50 · user7140：川寶 大家都在等你\n#6121 · 07/24 11:59 · user7141：大家都破底 就剩你台股硬撐不破底\n6. 後見之明偏誤：收盤後每個人都早知道\r#\r結果出現後，原先的多種可能性會從記憶裡消失。真正可檢驗的不是「早說」，而是事前是否留下方向、期限與失效條件。\n#1952 · 07/24 09:35 · user7136：果然週五盤是正常的！ 菇菇我有說週五正常！\n#4743 · 07/24 10:59 · user7100：早說 TSM每次+5%之後就是一直烙賽\n7. 類比謬誤：把日韓、美股或夜盤直接翻譯\r#\r跨市場確實相關，但交易時段、權重、匯率與事件曝險不同；相關不等於可以一比一複製漲跌。\n#2031 · 07/24 09:39 · user7142：就油昨天爆衝 夜盤嚇一堆人\n#7154 · 07/24 12:55 · user7143：這種時候心態最難受，不賣晚上還有美股賣了就輸了\n8. 災難化與全有全無：不是 V，就是歸零\r#\r這比較接近認知扭曲而非形式邏輯謬誤：把連續的風險壓成生或死兩個選項，會讓部位管理退化成情緒口號。\n#2769 · 07/24 10:10 · user7144：完蛋啦!完蛋啦!!完蛋啦!!!一切都完蛋啦!!!～回檔～\n#6655 · 07/24 12:25 · user7145：救命啊，不想玩了\n這份考古不能證明什麼\r#\r它不是情緒指標回測，也沒有證明某類留言能預測下一根 K 線。 關鍵字分類是可重現的抽樣入口，不是對全部 8,454 則留言做唯一正解標註。 同一句話可能同時包含多種偏誤；為了可讀性，每則引文只放在一個小節。 帳號已置換為合成代號（user0001 之類），與原帳號無對應關係、不可回推；保留樓層與時間，是為了讓讀者能回到匿名化原始檔核對。 2026 年截至 07/29 的前十大下跌日\r#\r名次 日期 收盤跌幅 跌點 文章 1 2026-07-17 -6.47% -2,953.71 「跌到收盤才停」 2 2026-07-28 -4.65% -2,030.83 「這次一定會 V？」 3 2026-03-09 -4.43% -1,489.12 「千點低開之後」 4 2026-03-04 -4.35% -1,494.77 「月線會救嗎」 5 2026-07-29 -3.76% -1,564.18 「四萬點保衛戰」 6 2026-06-26 -3.64% -1,683.50 「指數在跌，嘴在做什麼」 7 2026-06-08 -3.48% -1,568.16 「夜盤只是誤會？」 8 2026-06-10 -3.31% -1,478.90 「反彈第二天又殺」 9 2026-07-24 -2.67% -1,195.97 「日韓跌，台股就會？」（本篇） 10 2026-03-23 -2.45% -821.38 「明天會噴回來？」 資料口徑\r#\r指數排名與 OHLC：Yahoo Finance ^TWII 日線快照，抓取日 2026/07/29；排名用相鄰交易日收盤價計算。臺灣證券交易所亦提供發行量加權股價指數歷史資料供核對。 留言：PTT Stock 板當日「盤中閒聊」，以 BBS 層完整抓取；本篇產生前已核對 AID、標題與留言總數。 隱私：公開原始檔與本文引文的帳號一律置換為合成代號，不保留原字元、不可回推。同一個代號在所有檔案裡指同一人，跨日比對仍然成立。 ","date":"2026年7月24日","externalUrl":null,"permalink":"/zh-tw/stocks/taiex-chat-0724/","section":"股票","summary":"","title":"「日韓跌，台股就會？」：8,454 則盤中留言的謬誤考古（2026/07/24）","type":"stocks"},{"content":"","date":"2026年7月24日","externalUrl":null,"permalink":"/zh-tw/tags/aporia/","section":"Tags","summary":"","title":"Aporia","type":"tags"},{"content":"","date":"2026年7月24日","externalUrl":null,"permalink":"/zh-tw/tags/maieutics/","section":"Tags","summary":"","title":"Maieutics","type":"tags"},{"content":"","date":"2026年7月24日","externalUrl":null,"permalink":"/zh-tw/tags/%E4%B8%AD%E5%A4%AE%E5%95%8F%E9%A1%8C/","section":"Tags","summary":"","title":"中央問題","type":"tags"},{"content":"","date":"2026年7月24日","externalUrl":null,"permalink":"/zh-tw/tags/%E5%80%AB%E7%90%86/","section":"Tags","summary":"","title":"倫理","type":"tags"},{"content":"","date":"2026年7月24日","externalUrl":null,"permalink":"/zh-tw/tags/%E5%89%B5%E4%BD%9C/","section":"Tags","summary":"","title":"創作","type":"tags"},{"content":"","date":"2026年7月24日","externalUrl":null,"permalink":"/zh-tw/tags/%E5%8A%A9%E7%94%A2%E8%A1%93/","section":"Tags","summary":"","title":"助產術","type":"tags"},{"content":" 緣起\r#\r我想寫一部作品。方法我都想好了——卻卡在最上游的一格：這部作品到底為了問哪一個問題而存在？\n我生不出那一句。素材滿出來，候選一堆，可就是拍不了板。於是我做了一件有點奇怪的事：我請一個 AI 當我的蘇格拉底助產婆，把那個問題從我身上接生出來。\n規矩只有一條，是它自己立的，也是整件事的關鍵——\n助產婆自己不生孩子。它絕不替我寫那一句。它只把我肚子裡已經有的那個孩子接出來，再幫我分辨：這是真的孩子，還是一團假象。\n以下是完整的接生過程。我的話一字未改；它的話我略作整理。\n第一個宮縮：先別漂亮，先用力\r#\r助產婆： 你出不來，多半不是沒素材（素材滿出來了），是你想一次就生出一個漂亮的孩子。你的天賦是把東西磨到發亮——但現在別磨，先用力。第一個、也是唯一一個問題，你只答這一題：如果只准你用一句粗話般直白、你自己讀了都覺得太露骨的話，說出你二十年一個人看清楚、卻還沒有人接住的那件事——你會脫口而出哪一句？不要寫成問句，不要引經據典。就是那句你平常會嚥回去的話。\n我： 為什麼世界上有窮人。\n舉到光下：一句孩子的話\r#\r助產婆： 孩子生出來了，別急著抱走，先舉到光下看清楚。它強在哪：它是一句孩子的話——「為什麼世界上有窮人」是六歲小孩會問的。蘇格拉底接生的真問題，幾乎都長這種天真的樣子。你沒有把它磨亮，很好。\n現在檢驗：這一題，圖書館裡有一千個答案，你也讀過，可是二十年了，沒有一個接住你。那代表你問的其實不是字面那句——「為什麼會有窮人」有答案，而「沒人接住你」的那題沒有答案在架上，所以是兩題。真正的孩子藏在這句話某一個字底下。把它唸出來，聽你的重音落在哪：\n為什麼 · 世界上 · 有 · 窮人\n我： 重音是「世界上有」，指涉的是句子中的背景世界跟句子中沒出現的富人。我認為富人最大的資產不是錢，而是足夠多的窮人；窮人不得不為了每個月的生活，從事他們不喜歡又單調的服務業，而富人享受的不是錢，而是服務。為了確保他們的錢換得到服務，必須有大把大把的窮人，把所有社會基礎設施的崗位跟服務業的職位鋪滿。\n延伸的議題是：如果未來 AI 生產力爆發、不需要那麼多勞動力，窮人是會被解放、過跟富人類似的生活，還是他們會直接被拋棄、被犧牲？\n論點，不是問題\r#\r助產婆： 你重音落在「世界上有」，手指卻指向句子裡沒出現的富人——你把看得見的（窮人）舉起來，逼大家去看藏在後面、沒被指名的那個人。這孩子是你的。\n但你剛生的其實是一個論點加一個問題，得分開驗。你的「1」不是問題，是答案：「富人最大的資產是足夠多的窮人」是你已經想通、已經相信的命題。危險在這裡——如果你拿作品去證明它，你就造了一台宣講真相的機器：那是說教，是贗品，不是藝術。你已經有答案的東西，不能當中央問題。\n你的「2」才是那個哭出聲的孩子：AI 讓勞動不再被需要時，窮人被解放還是被拋棄？這題架上沒答案，你自己也答不出來。而且它們是同一把刀的兩面——如果窮存在的理由，從頭到尾就是「對某人有用」，那當機器讓他們不再有用，那個理由就蒸發了。那一刻，我們才會知道：我們到底把他們當人，還是只當一副會走路的勞動力。\n我： 讓窮人不再「有用」時會被拋棄，與其說是拋棄，不如說是讓他們自生自滅。富人會築起城牆、自己過富足的生活；此時窮人需要交換物資的時候，還是需要想辦法服務富人。所以窮人無法解放，不是生產力問題，而是社會跟制度問題。\n你把兇手掉包了\r#\r助產婆： 停。你剛剛做了我警告過的事——你答了。「拋棄／自生自滅／是制度問題」是一個收口的悲觀結論，是命題，不是問題。你要是拿作品去證明「窮人永遠不可能被解放」，你又造回那台機器了，只是這次它宣講絕望。\n而且你把兇手掉包了：你說「不是生產力問題，是制度問題」——制度沒有手。你上一輪還承認那隻手有一根手指是你的（你站在造機器、供給生產力的那一側），這一輪你把自己從現場移走，躲進「制度」這個沒有臉的詞後面。\n你也放不掉「窮人還是得服務富人」這半句——可這跟你自己的前提打架：AI 若什麼都做了，富人根本不需要人來服務。你為什麼死抓著不放？因為你不敢看更深的那一層：被剝削的人，至少對某人還有用。真正的深淵不是剝削加劇，是變成沒有任何人需要的、多餘的人。不被壓迫，只是不被需要。\n所以把刀交回你自己手上，答這句：當我不再需要你——不是剝削你，是根本用不上你——我到底還欠不欠你？\n我： 這沒有確切的答案，只有幾種潛在的可能。因為 AI 之前，歷史上已經有許多次的生產力爆發，但同樣的問題始終沒有解決。\naporia：那個從沒動過的常數\r#\r助產婆： 現在你站對地方了。「沒有確切的答案，只有幾種可能」——這正是中央問題該有的樣子。蘇格拉底管這叫 aporia：不是無知，是真正的、有生產力的答不出來。前面你一碰到這裡就逃、就收口；這一次你停在裡面了。\n而你順手遞來的那把歷史鑰匙，比你以為的重得多：如果每一次富足暴漲都沒解決它，那它從來就不是富足的問題。所以你的題不是「窮人會怎樣」（那是社會學，架上有答案），而是那個穿越每一次生產力爆發都沒有動過的常數。蒸汽機、電、電腦、AI，墨水一次比一次多，可是有一樣東西紋風不動：「我欠不欠一個我不需要的人」這件事，從古到今沒挪動半吋。\n這才是你二十年看清楚、卻沒人接住的東西。現在把「歷史」「窮人」「富人」這些第三人稱全部劃掉，留下包含你自己、第一人稱的那一句。你，現在，說出來。\n我： （為什麼）社會始終都有階級。\n你又把自己劃掉了\r#\r助產婆： 「始終」你抓對了——那個常數你接住了，留著。但你當著我的面做了兩個退步。\n一：你又把自己劃掉了。 我要你放第一人稱，你給的主詞是「社會」——全世界最大的第三人稱；你甚至把「為什麼」塞進括號，連發問的人都推遠。\n二：你把活的換成死的。 兩輪前你握著會割人的字——「多餘的人」「沒有人需要」「我到底欠不欠你」；這一句你全收走，換成「階級」：一個乾淨、安全、課本裡答案滿坑滿谷的社會學名詞。架上有答案的，不是你的題。\n而且你自己要問的是「人應當怎樣」（應然），「為什麼社會有階級」問的卻是世界為何如此（實然）——你用實然，躲開了你自己的應然。\n你只差一次換字：把「社會」丟掉、放「我」進來，把「階級」丟掉、換回你那個會痛的詞，「始終」留著。再說一次，這次讓你自己在句子裡。\n我： 我始終沒辦法財富自由，變成富人階級。\n傷口，不是問題\r#\r助產婆： 這一句你沒有躲。前面每次要你放自己進去，你都逃進第三人稱；這一次你把手伸進胸口，把最難看的那塊掏出來：「我始終沒辦法財富自由，變成富人階級。」它很痛、很不體面、很像自憐——正因為這樣，它是真的。\n但我要告訴你：這不是你的中央問題，這是那個問題一直在守護的傷口。 你前面那整套冷靜的社會學，全是這顆滾燙的私人恐懼發的電——你怕你就是那個沒用的、被丟在牆外、沒有人需要的人。硬幣的一面是傷口（我怕變成不被需要的人，所以拼命想爬進富人的牆內），另一面是倫理（我們該不該欠一個不被需要的人）：它們是同一枚硬幣。\n而這是最狠的一刀——你想爬進的那道牆，按你自己的分析，正是那些「築牆、自己富足、不再欠牆外任何人」的人。所以你朝思暮想的得救，就是變成你這輩子最想質問的那種人。你的救贖和你的罪，是同一個動作。 這就是為什麼你答不出「我不需要你時還欠不欠你」——因為你正用盡全力，想爬到「再也不必欠」的那一邊去。\n把這兩半絞成一句：第一人稱，你在裡面，而且它同時是傷口、也是「人應當怎樣」。\n我： 我的倫理觀認為不應該有財富差距，但我也不覺得要強迫其他人遵照自己認為正確的規則，我只能自己想辦法讓自己足夠富有。\n一個乾淨到殘忍的三難\r#\r助產婆： 這一次沒有「社會」、沒有「制度」、沒有安全詞。你把自己釘在一個你真的走不出去的矛盾裡，而且三句全是「應當」、全是第一人稱。這是一個乾淨到殘忍的三難：\n我的良知說：不該有貧富差距。 我的良知同時說：我不能強迫任何人照我的對錯活。 於是我唯一剩下的動作是：自己變富——也就是親手跳進我譴責的那道差距。 你的良知禁止那個病，又禁止那帖藥，最後只留給你一條路：自己去染上那個病。\n而你那第二句「不強迫別人」不是配角——它決定了這部作品的形式：你不能做一台押人投票、宣講真相的機器，你只能遞一副透鏡，讓人自己走進去、自己發現。\n話你已經說完了，只差一個標點。你把它寫成一個認命的句號，但活著的東西，是這陷阱逼出來的那個問號。而且它早就不只是錢——錢只是那個房間。把你剛那三句話，收在問號上，再說一次。\n我： 我的倫理觀認為不應該有財富差距，但我也不覺得要強迫其他人遵照自己認為正確的規則，我只能自己想辦法讓自己足夠富有？\n生出來的那一句\r#\r我的倫理觀認為不應該有財富差距，但我也不覺得要強迫其他人遵照自己認為正確的規則，我只能自己想辦法讓自己足夠富有？\n看清楚我剛剛做了什麼：我沒有讓它寫，沒有換成它遞給我的漂亮版本，我用自己那三句、自己的手，按下了那個問號。\n這是整件事唯一的規矩——孩子是我的。從「為什麼世界上有窮人」到最後那個問號之間，每一步都是我自己走的。沒有一個字是別人設計的。\n而那第二句——「不願強迫別人」——讓我的痛苦不是自憐，是悲劇；它也替我把作品的形式定了下來：我不能做一台矯正別人的機器，我只能做一個讓人自己走進去、自己發現的東西。\n火藥已經就位。接下來難的不是問題，是那難的百分之十：讓這個問號，射中下一個人。\n","date":"2026年7月24日","externalUrl":null,"permalink":"/zh-tw/posts/midwife-the-question/","section":"部落格","summary":"","title":"我請一個 AI 當我的蘇格拉底助產婆，把一個問題從我身上接生出來","type":"posts"},{"content":"","date":"2026年7月24日","externalUrl":null,"permalink":"/zh-tw/tags/%E8%98%87%E6%A0%BC%E6%8B%89%E5%BA%95/","section":"Tags","summary":"","title":"蘇格拉底","type":"tags"},{"content":"","date":"2026年7月24日","externalUrl":null,"permalink":"/zh-tw/tags/%E8%B2%A1%E5%AF%8C%E8%87%AA%E7%94%B1/","section":"Tags","summary":"","title":"財富自由","type":"tags"},{"content":"這是盤中閒聊考古系列的延伸。前幾篇我把鄉民留言釘在「當下的指數」上，看的是情緒；這篇換一個更硬的問題：把每一句「會漲到 X」「要噴」「準備崩」當成一張可以事後對帳的預測單，一年半後回頭算帳——到底誰的嘴會準？ 我抓了 PTT Stock 板 2025/01 到 2026/07、3,516 位作者的 10,841 則可證偽喊單，逐則對齊真實股價判定命中，做成下面這張戰績榜。\n什麼算一則「可證偽喊單」\r#\r不是每句話都能對帳。「台積電讚」不行，「台積電年底上 3000」可以。我只收有方向、且能被價格證偽的三種：\n目標價（price_target）：「XXX 會到 N 元」。判定：期限內盤中碰到目標價 → HIT。 百分比（pct_move）：「這檔要漲一成」。同上，碰到即算。 純方向（direction_only）：「要噴 / 準備跌」。判定：期限日收盤方向對 → HIT（看終點，不是看中間有沒有戳一下）。 沒明講期限的，一律套 60 個日曆日的觀察窗。抓不到價、或發文當下目標已達成的，標為「無法驗證」不計入。\n三個先講在前面的誠實面\r#\r同一檔反覆貼會灌水：有人一週貼五次「6206 要跌」，那不是五個獨立預測。下面的排名已做去重——同作者、同標的、同方向、同月只算一次。（去重前有人靠反覆貼同一檔衝到 93%，去重後直接掉榜。） 牛市讓「喊漲」變簡單：這 18 個月大盤是往上的，喊漲本來就容易中。所以看到高命中先別跪——要看他喊空準不準（逆風才見真章），我把每個人的多空拆開給你看。 這是溫度計，不是明牌：命中率高 ≠ 你該跟。這張表的用途是事後對帳與看群眾在幹嘛，不是投資訊號。喊中也不代表當下就該買。 🌡️ 整體溫度計 2025/01 – 2026/07・全板可證偽喊單\r—整體命中率\nHIT ÷（HIT＋MISS）\r喊漲 ▲\r喊跌 ▼\r50% 硬幣50%\r🏆 作者戰績榜 去重・≥10 則已結・點一列看多空拆解\rWilson 下界 ↓\r命中率 ↓\r樣本數 ↓\r#作者（匿名代號）命中率WilsonH/M/未結\rWilson 下界＝把小樣本運氣扣掉後「保守估計的真實命中率」（95% 信心的下限）；同樣命中率，樣本多的排前面。這是排名的主鍵，因為「3 中 3＝100%」不該贏過「27 中 23」。\n📉 高產出、低命中 喊得多（≥40 則）、命中卻 \u0026lt; 30%\r聲量最大 ≠ 最準。全板有兩個帳號落在「高產出、低命中」這一格——喊單量以數十計，命中率卻低於丟硬幣一大截。這裡看的是這個**現象**，不是這兩個人。\n資料：PTT Stock 板 2025/01–2026/07 貼文與推文，經抽取為可證偽喊單，逐則對齊台股／美股日線判定 HIT／MISS／未結。\r作者帳號已置換為匿名代號,不可回推至原帳號。命中定義：目標價／百分比＝期限內盤中觸價；純方向＝期限日收盤方向正確；未明講期限套 60 日窗。\r▲ 紅＝喊漲、▼ 綠＝喊跌（台股慣例）。此為資料實驗，非投資建議。\r這張表看出什麼\r#\r整體是「輸多於贏」的。 全板可證偽喊單命中率 35.7%——比丟硬幣還低一截。當然這不完全公平（目標價要「精準碰到」本來就難），但方向很清楚：把鄉民喊單當買賣訊號，聚合起來是賠錢的一邊。 這正是「擦鞋童」寓言的量化版：當一件事連板上都在喊，它多半已經反映在價格裡。\n牛市把「喊漲」變簡單、「喊空」變地獄。 拆開方向看，這 18 個月喊漲命中 38.7%、喊空只有 26.2%。意思是榜上那些高命中作者，有多少 edge 其實是「站對了大盤的浪」、而非「選股的本事」，很難分乾淨——所以我在每位作者的展開列裡標了空方戰績：能在逆風中喊空還準的（表上會亮「← 逆風也準」），才是比較純的訊號。\n最大聲的往往最不準。 落在「高產出、低命中」那一格的兩個帳號，喊單量是榜首的兩三倍，命中率卻只有 20% 上下。這不是巧合——高頻喊單本身就是一種行為偏誤的症狀（手癢、要存在感、賭一把翻本），而市場對這種情緒不留情。\n喊單高度集中在「大盤 + 台積電」。 被喊最多的是加權指數本身（3,775 次）和台積電（1,656 次），合計佔了三分之一還多。個股的獨立觀點其實稀薄——多數人是在對同一根神山、同一條大盤反覆表態。\n這張表不能拿來做什麼\r#\r不能照抄榜首的單。 命中率高有一部分是牛市紅利；而且「他過去準」對「你現在跟進的這一筆」沒有保證——每一則喊單都是獨立事件。 不能因為「連準的人都看多台積電」就覺得非追不可。 榜上戰績好的作者現在幾乎全在做多同幾檔權值股——這本身就是一個群眾一致性的讀數，而群眾越一致、越接近某種極端，越是該冷靜、不是該跟。看穿別人的情緒容易，看穿自己手癢難。 命中 ≠ 當下就該進場。 一則「年底會到 3000」就算最後真的中了，也不代表你今天追高是對的決策。預測力和執行紀律是兩件事——這張表只量前者。 這張榜的價值，跟前幾篇留言考古一樣，不在告訴你明天買什麼，而在當你下次又想跟著板上喊聲進場時，回來看一眼整體那 35.7%——然後問自己：我這一票，憑什麼在剩下的那群人裡？\n資料來源：PTT Stock 板 2025/01–2026/07 貼文＋推文，經結構化抽取為可證偽喊單，逐則對齊台股／美股日線判定。所有帳號已置換為匿名代號,不保留任何可回推的字元。\n本文為個人資料實驗與行為觀察紀錄，所有引用均為公開內容、僅作教育與心理觀察用途，不構成任何投資建議，也不針對任何個人。命中判定依機械規則自動產生，可能因股價資料延遲、分割、期限假設而有誤差；投資決策請自行評估、自負風險。\n","date":"2026年7月21日","externalUrl":null,"permalink":"/zh-tw/stocks/ptt-call-scoreboard/","section":"股票","summary":"","title":"「嘴會不會準？」：PTT Stock 板 10,841 則喊單的戰績考古","type":"stocks"},{"content":"","date":"2026年7月21日","externalUrl":null,"permalink":"/zh-tw/tags/%E5%96%8A%E5%96%AE/","section":"Tags","summary":"","title":"喊單","type":"tags"},{"content":"","date":"2026年7月21日","externalUrl":null,"permalink":"/zh-tw/tags/%E8%B3%87%E6%96%99%E8%A6%96%E8%A6%BA%E5%8C%96/","section":"Tags","summary":"","title":"資料視覺化","type":"tags"},{"content":"這是「2026 台股前十大下跌日」系列第 1 名。排名口徑是 2026/01/01 至 2026/07/29 的加權指數單日收盤報酬率，不是盤中最大跌幅。\n開盤就是最高、收盤就是最低。這種幾乎沒有喘息的一路下殺，最容易把推測壓縮成一句「一定還會怎樣」。\n這天發生了什麼\r#\r數值 前收 45,624.98 開盤 45,234.08 最高 45,234.08 最低 42,671.27 收盤 42,671.27 漲跌 ▼ 2,953.71（-6.47%） 2026 跌幅排名 第 1 名 BBS 全量留言\r#\rPTT 網頁版遇到推爆長文會省略中段；這裡走 BBS 層抓到 19,389 則，不是網頁殘片。\n推文型態 數量 推 11,256 噓 1,344 → 6,789 合計 19,389 留言最密集的時段是 12:00（4,078 則）。\n時段 留言數 相對量 08:00 1,398 █████ 09:00 3,452 ██████████████ 10:00 3,388 █████████████ 11:00 2,908 ███████████ 12:00 4,078 ████████████████ 13:00 4,075 ████████████████ 14:00 86 █ PTT 原文：AID 1gMNUCRY 匿名化原始檔：純文字 · JSONL 8 種謬誤與偏誤\r#\r以下是語句模式的分類，不是對留言者作人格診斷。盤中閒聊包含反串、迷因與情緒宣洩；引文只能證明這句話出現過，不能證明作者真的照著交易。\n1. 單因謬誤：把市場縮成一個黑手\r#\r市場同時包含外資、內資、避險、被動資金與個別公司消息；把整段價格路徑只歸因於一個有意志的角色，故事會很順，證據卻通常不夠。\n#6079 · 07/17 10:24 · user7122：好慘 這次大媽輸了\n#14867 · 07/17 12:53 · user7123：南韓政府提高保證金，會引發超級賣壓小心了\n2. 賭徒謬誤：跌多了就「該」彈\r#\r先前已經跌多少，不會自動提高下一分鐘上漲的機率。價格可以很便宜，也可以在缺乏新資訊時繼續變便宜。\n#4459 · 07/17 09:51 · user7033：還在抄底牙科的是不知道今天韓國平盤是因為放假膩\n#11911 · 07/17 12:10 · user7124：抄底啊啊啊啊啊啊啊啊啊啊啊啊\n3. 錨定效應：月線、季線與整數都是答案\r#\r技術位置可以是風險管理參考，但把單一價位當成必然反轉點，會忽略波動、成交量與事件條件。\n#5434 · 07/17 10:13 · user7125：聰明多蛙準備好，季線要來哦\n#11649 · 07/17 12:07 · user7126：GG都沒碰到季線要守啥XD\n4. 直線外推：今天的斜率一路畫到明天\r#\r人腦很容易把眼前最強烈的方向延伸出去；市場真正困難之處，正是斜率會在新資訊出現時改變。\n#7236 · 07/17 10:44 · user7127：被動可能通通要跌到被關了，其他也是\n#15034 · 07/17 12:56 · user7081：長黑必搭長紅 下禮拜直接往上噴2000點\n5. 從眾效應：多蛙、空蛙與「大家都知道」\r#\r群體標籤能快速製造安全感，也會把不同持倉、期限與風險承受度的人壓成同一種對手。\n#5135 · 07/17 10:07 · user7128：空蛙輸哥：你看跟我說的一樣～是不是下去了 快吹捧\n#14373 · 07/17 12:44 · user7129：折吹還不出來給大家個交代\n6. 後見之明偏誤：收盤後每個人都早知道\r#\r結果出現後，原先的多種可能性會從記憶裡消失。真正可檢驗的不是「早說」，而是事前是否留下方向、期限與失效條件。\n#4801 · 07/17 09:59 · user7130：利多早就出盡 不會漲了\n#13026 · 07/17 12:28 · user7131：安啦正二梭了啊 不是說上看五六萬點\n7. 類比謬誤：把日韓、美股或夜盤直接翻譯\r#\r跨市場確實相關，但交易時段、權重、匯率與事件曝險不同；相關不等於可以一比一複製漲跌。\n#5914 · 07/17 10:21 · user7132：NV夜盤在跌三小\n#18368 · 07/17 13:38 · user7133：美股盤前也是死透了下禮拜在來一個3000應該沒槓桿仔\n8. 災難化與全有全無：不是 V，就是歸零\r#\r這比較接近認知扭曲而非形式邏輯謬誤：把連續的風險壓成生或死兩個選項，會讓部位管理退化成情緒口號。\n#3165 · 07/17 09:28 · user7134：GG真的變法會了救命啊\n#13043 · 07/17 12:28 · user7134：跌2500點啦救命啊\n這份考古不能證明什麼\r#\r它不是情緒指標回測，也沒有證明某類留言能預測下一根 K 線。 關鍵字分類是可重現的抽樣入口，不是對全部 19,389 則留言做唯一正解標註。 同一句話可能同時包含多種偏誤；為了可讀性，每則引文只放在一個小節。 帳號已置換為合成代號（user0001 之類），與原帳號無對應關係、不可回推；保留樓層與時間，是為了讓讀者能回到匿名化原始檔核對。 2026 年截至 07/29 的前十大下跌日\r#\r名次 日期 收盤跌幅 跌點 文章 1 2026-07-17 -6.47% -2,953.71 「跌到收盤才停」（本篇） 2 2026-07-28 -4.65% -2,030.83 「這次一定會 V？」 3 2026-03-09 -4.43% -1,489.12 「千點低開之後」 4 2026-03-04 -4.35% -1,494.77 「月線會救嗎」 5 2026-07-29 -3.76% -1,564.18 「四萬點保衛戰」 6 2026-06-26 -3.64% -1,683.50 「指數在跌，嘴在做什麼」 7 2026-06-08 -3.48% -1,568.16 「夜盤只是誤會？」 8 2026-06-10 -3.31% -1,478.90 「反彈第二天又殺」 9 2026-07-24 -2.67% -1,195.97 「日韓跌，台股就會？」 10 2026-03-23 -2.45% -821.38 「明天會噴回來？」 資料口徑\r#\r指數排名與 OHLC：Yahoo Finance ^TWII 日線快照，抓取日 2026/07/29；排名用相鄰交易日收盤價計算。臺灣證券交易所亦提供發行量加權股價指數歷史資料供核對。 留言：PTT Stock 板當日「盤中閒聊」，以 BBS 層完整抓取；本篇產生前已核對 AID、標題與留言總數。 隱私：公開原始檔與本文引文的帳號一律置換為合成代號，不保留原字元、不可回推。同一個代號在所有檔案裡指同一人，跨日比對仍然成立。 ","date":"2026年7月17日","externalUrl":null,"permalink":"/zh-tw/stocks/taiex-chat-0717/","section":"股票","summary":"","title":"「跌到收盤才停」：19,389 則盤中留言的謬誤考古（2026/07/17）","type":"stocks"},{"content":"這是 06/26 崩盤日與 06/29 反彈日之後的第三篇盤中閒聊考古。2026/07/07 這篇 PTT Stock 板盤中閒聊，表面上是在吵台股，實際上更像一場訊號辨識壓力測試：夜盤到底能不能信？試撮是不是有人在騙？TSM ADR 漲了，台積電現貨該怎麼反應？韓股、美股、匯率、00685/正二，哪一個才是真正的原因？\n這次我先不做逐分指數互動圖，先做留言本身的謬誤盤點。因為這一天最有意思的，不是單一行情方向，而是鄉民如何在一堆互相矛盾的訊號裡，快速替自己的部位找劇本。\nBBS 全量資料\r#\r數值 文章 [閒聊] 2026/07/07 盤中閒聊 作者 laptic AID 1gJ4YCUE 全量留言 10,223 則 推 / 噓 / → 6,056 / 1,098 / 3,069 推噓比 5.52 留言高峰 11 時 2,556 則、12 時 2,276 則 全量留言：PTT 網頁版對推爆文有「檔案過大！部分文章無法顯示」限制；這份資料走 BBS 層（PyPtt）抓取，拿到 10,223 則、沒有中段斷層。 隱私：所有留言者帳號都置換為合成代號（例如 exampleuser → user0042）。代號與原帳號沒有任何對應關係，不可回推；同一個代號在所有檔案裡指同一人。 原始檔：本地已保存為 static/experiments/ptt/stock-M1783384204-A78E.txt 與 static/experiments/ptt/stock-M1783384204-A78E.jsonl。 第一則留言照例又是那句：\n#1　推　user3543：早安大爆崩\n這句話到第三篇已經不是預測，而是晨間儀式。真正值得看的，是後面 10,222 則留言如何在「崩／噴／V／夜盤假的／正二忠誠」之間來回切換。\n嘴在辨識什麼訊號\r#\r這天留言最密集的不是開盤前，而是 11 時到 12 時。也就是說，真正讓人焦躁的不是「開出來」那一瞬間，而是行情走了一段之後，大家發現早上的劇本不夠用了。\n最常出現的心理線索有幾條：\n夜盤與現貨脫鉤：大量留言在爭論「夜盤是不是假的」。 00685/正二溢價：從試撮、漲停、抽單，到開盤後收斂，整段都像小型群眾實驗。 TSM ADR 對台積電現貨的錯位期待：把 ADR 漲幅線性換算成台股現貨，是這天很典型的訊號錯讀。 外部歸因：韓股、三星、美光、小那、川普、匯率，輪流被拿來當解釋。 「小兒／大媽／主力」人格化市場：把盤勢當成幾個角色在演劇本，能減少不確定感，但也很容易讓決策變成故事接龍。 下面把這天最明顯的 11 類思考謬誤整理出來。這些留言不是拿來嘲笑誰，而是拿來看見：人在盤中最容易把哪些情緒誤認成判斷。\n1. 槓桿 ETF 狂熱（商品理解被信仰取代）\r#\r00685/正二是這天的迷因核心。試撮、漲停、溢價、抽單，所有細節都被拿來當短線命運籤。問題不在討論 ETF，而是把槓桿商品當成一種集體信仰。\n#44　→　user5364：正二 忠誠！ 685 出關！ #67　推　user4180：00685有種別抽單漲停，我先掛200張給你= = #489　推　user5780：漲停追正2 衝啊\n2. 夜盤／試撮訊號崇拜（把雜訊當預言）\r#\r夜盤不是不能看，但這天很多留言把夜盤當成「一定會發生」或「一定是假」的二分訊號。訊號一旦被神化，判斷就會變成站隊。\n#122　推　user2967：今天拼手速嗎 夜盤假的 #294　推　user1844：又被夜盤騙惹 #474　推　user6250：夜盤根本是詐騙集團\n3. ADR 線性換算錯覺\r#\rTSM ADR 漲，不等於台積電現貨必須等比例反應。盤中最常見的錯覺之一，就是把不同市場、不同時間、不同流動性的價格，硬翻成一個簡單算式。\n#87　噓　user2456：TSM4% GG+5 #111　推　user2229：TSM4%台積電5塊 #611　推　user6724：TSM+4 % 、 GG+10塊？\n4. V 轉賭徒謬誤\r#\r跌下去之後一定要 V，跌越急越該 V，這是典型的賭徒謬誤。市場不欠任何人一根反彈，K 線也沒有「跌夠了就該補償」的義務。\n#190　推　user4423：這不就又V辣 #339　推　user0947：快V阿!!!!!!!!! #837　推　user1728：開低走高 世界強 VVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVV\n5. 後見之明（早說仔宇宙）\r#\r盤中最便宜的東西是「早說」。行情往下，就說早就知道夜盤假；行情往上，就說早就知道會 V。真正難的是在事前承擔不確定性，而不是事後重寫記憶。\n#181　推　user0454：早說了，七月大崩盤 #255　推　user0652：看吧 就說夜盤假的 笑死人 #1581　→　user0454：早說了，七月大崩盤，等季線再撿，一堆傻多\n6. 技術線神諭\r#\r季線、月線、缺口、壓力、支撐都可以是觀察工具，但當它們被講成「一定會守、一定會破、到了就買」，就變成問神。\n#822　推　user2422：多軍目標一樣是封閉47300缺口 封了就噴 封不了就 #1514　推　user3950：國巨月線保衛戰 #3183　推　user2302：988A快碰季線了，沒救了\n7. 外部歸因（都是韓國、美股、川普、匯率的錯）\r#\r外部市場當然重要，但盤中情緒很容易把複雜變成單因：韓國害的、美股害的、三星害的、川普害的。這種歸因讓人感覺掌握了原因，卻未必真的提高判斷品質。\n#153　噓　user0379：美日韓到底會不會做股票 不要拖累台股 #177　→　user0379：沒有垃圾美股韓股 台股早就五萬點了 #337　推　user5694：韓國帶崩 崩完晚上美股崩\n8. 主力／小兒／大媽敘事\r#\r把市場人格化，是散戶聊天室最常見的安定劑。只要有「小兒」「大媽」「主力」「黑手」，波動就不再是混沌，而像有人在導演。但故事越完整，越容易忘記驗證。\n#575　噓　user5383：大家團結一條心～～～買到小兒認錯辣 ^^ #1580　推　user0608：開盤撐盤狗在那邊騙 套了一堆自家散戶給小兒爽 87 #1841　推　user4879：第四法人 散戶大軍 快來撐盤 快被看沒有了\n9. 損失趨避與凹單\r#\r賠錢時，人會自然尋找「還有救」的語言。99、救救、解套、攤平，看起來像玩笑，背後其實是損失趨避：不想承認錯，所以先把希望留在文字裡。\n#546　推　user1842：A下去破底就是為了要用力拉的，多單往下加碼怎麼輸? #790　→　user5566：攤平到成本600凹贏了 #1020　→　user2162：賣2成西瓜金買988A，剛好住山頂，救救我\n10. 過度自信（確定性語言）\r#\r「穩贏」「一定」「直接」「沒時間解釋」都是盤中高風險詞。它們不一定代表錯，但通常代表腦子已經把機率問題改寫成劇本問題。\n#251　→　user6500：開盤空穩贏 #540　噓　user5383：最後上車門票\u0026hellip;沒時間解釋～5卍列車發車辣 ^^ #556　噓　user0408：直接崩回40000拉，一直鳥盤\n11. 末日／恐慌迷因\r#\r「大爆崩」在這系列已經變成問候語，但恐慌語言還是會影響節奏。當災難詞彙越刷越順，盤中的風險感會被放大成娛樂，也會把人推向過度交易。\n#1　推　user3543：早安大爆崩 #250　噓　user2140：老蘇都預告史上最大崩盤大爆AAAAAAA還敢多？ #613　推　user3487：絕望崩盤\n寫在最後\r#\r07/07 這篇最有價值的地方，是它把「訊號太多」時的散戶心理攤開來。06/26 是崩盤日，06/29 是反彈日；07/07 則是訊號打架日。夜盤說一套、ADR 說一套、韓股說一套、正二試撮又說一套，最後大家只好挑一個最能安慰自己部位的故事。\n這也是盤中閒聊資料最像鏡子的地方：我們嘴上在討論市場，其實常常是在替自己的不安找語言。看穿這件事，不會讓人明天就預測得更準，但至少能在下一次想喊「夜盤假的」「一定 V」「小兒又在騙」之前，先停一秒問自己：我是在判斷，還是在找劇本？\n資料來源：PTT Stock 板 2026/07/07 盤中閒聊全量留言（BBS 層抓取）。所有帳號已置換為不可回推的合成代號。\n本文為個人資料實驗與行為觀察紀錄，所有引用留言均為公開內容、僅作教育與心理觀察用途，不構成任何投資建議，也不針對任何個人。投資決策請自行評估、自負風險。\n","date":"2026年7月7日","externalUrl":null,"permalink":"/zh-tw/stocks/taiex-chat-0707/","section":"股票","summary":"","title":"「夜盤假的嗎」：10,223 則盤中留言的訊號錯讀考古（2026/07/07）","type":"stocks"},{"content":"這是上一篇崩盤日考古的續集。上週五（06/26）台股大跌一根，這個週一（06/29）早盤一路衝高、最多彈逾 900 點，於是整篇盤中閒聊從早到晚只在吵一件事：這是真反彈，還是讓你上車的逃命波？ 我一樣把 PTT Stock 板當天那篇「盤中閒聊」的全部 7,061 則留言抓下來，逐則對齊「發文那一刻」的加權指數。指數在彈的時候，鄉民的嘴在做什麼？\n這天發生了什麼\r#\r數值 開盤 44,571.76 收盤 44,999.90 盤中最高 45,510.99 盤中最低 44,571.76（09:00） 漲跌 ▲ 428.14（+0.96%） 這天並不是單純的紅或黑：大盤開平之後早盤一路衝高，09:46 觸及全日最高 45,510.99（最多彈逾 900 點、約 +2.1%），但午後一路回吐、把超過一半的漲幅還了回去，終場收在 44,999.90、僅守住 +0.96%。一根「彈得猛、收得弱」的反彈，也正是當天從早吵到晚的原因。\n有趣的是，當天那篇盤中閒聊的第一則留言，時間 08:30，跟上週五崩盤日一字不差——還是那五個字：\n#1　推　user7001：早安大爆崩\n但這天大盤是一路彈高的。「早安大爆崩」已經跟當天行情無關，變成一句純粹的晨間問候迷因——上週五它還是寫實，這天它只是儀式。崩盤梗一旦長出生命，就不再需要崩盤了。\n資料是怎麼來的\r#\r全量留言：PTT 網頁版（www.ptt.cc）對推爆文有「檔案過大！部分文章無法顯示」的上限，只會吐出開頭加上抓取當下的最新尾段，中間整段被藏起來。真正的完整內容在 BBS 原始檔裡，走 BBS 層（PyPtt）可以拿到全部 7,061 則、沒有斷層。 指數：台灣證交所盤中 5 秒級的「發行量加權股價指數」，每分鐘取樣，對齊留言時間。 隱私：所有留言者帳號都置換為合成代號（例如 exampleuser → user0042）。代號與原帳號沒有任何對應關係，不可回推；同一個代號在所有檔案裡指同一人。 互動圖：把留言釘在指數上\r#\r在圖上左右滑動或點擊，或拖曳下方的時間軸，就會看到那一分鐘的加權指數、以及當下湧出的留言。底部灰色小柱是「每分鐘留言量」——和上週五的崩盤日對照，你會發現這天的留言在開盤後（09:00 前後）就先爆量，然後一路遞減，跟著盤中那顆「逐漸沒力的反彈」一起退潮。\n📈 加權指數 × 鄉民留言 開 收 最低 留言 則・推噓比 · 加權 · 則 全部 推 噓 → 資料:PTT Stock 板當日盤中閒聊全量留言(走 BBS 層抓取,網頁版受「檔案過大」限制只剩約一成); 加權指數為台灣證交所 5 秒級即時資料,每分鐘取樣對齊。留言者帳號已置換為不可回推的合成代號。 拖曳時間軸或在圖上滑動 / 點擊,即可看當下指數與該分鐘的留言。 指數在彈，嘴在做什麼\r#\r跑完全量這幾個數字很有意思：\n推噓比 5.75（推 4,207／噓 732／→ 2,122）。比上週五崩盤日的 6.33 略低，但「推」依舊壓倒性多——這裡的「推」大多不是看多，而是看戲、嘴砲、刷迷因，無論漲跌都一樣。 留言在開盤（09 時）最熱（單小時 2,091 則），對應的正是「開高之後到底會不會續攻」最懸而未決的那段。人最吵的時刻，不是行情最差的時候，是方向最不明的時候。 單分鐘最高出現在 09:06（99 則，指數約 45,110.72）。 「逃命波」「騙人上車」「鳥量反彈」整天刷個不停——事後看，這群唱衰的嘴半套說中了：早盤近千點的漲幅，午後確實吐回一半多；但喊「翻黑、崩下去」的也只對一半，終場仍收紅 +0.96%。在一個「彈一半、吐一半」的曖昧日子裡，每種敘事都能找到自己的證據——這才是最該警惕的地方:模稜兩可的盤，餵養所有偏見。 下面把當天最常見的散戶思考謬誤盤點成 11 類，每類附 3 則真實留言（樓層＋合成代號＋原文照錄）。這些不是要嘲笑誰——是一面鏡子，照的是我們每個人在「該不該追這根反彈」的當下，腦子會自動跑出來的那些話。\n1. 接刀／抄底續命（沉沒成本接力）\r#\r上週五跌下來抄了底的人，這天最大的功課是「該不該獲利了結、或繼續凹」。把「我已經買了」當成繼續加碼的理由，是沉沒成本的經典接力。\n#145　推　user7080：上週抄底的今天賣不賣呢 #779　推　user7339：爆噴！上週五阿呆谷 無腦歐印哪次輸 #495　推　user7174：國巨破千抄底 穩贏的吧\n2. 攤平／向下加碼（這天反而變少了）\r#\r有趣的是，崩盤日滿坑滿谷的「攤平」，到了反彈日幾乎消失——因為帳面回血了，痛感降低，加碼的衝動也跟著退潮。攤平的衝動從來不是策略，是情緒的溫度計。\n#538　推　user7115：買，94買，越跌越買，攤平攤成大富翁 #5245　→　user7038：早上搶進的，繼續加碼攤平啊 #6576　噓　user7174：夜盤還會開高自救嗎，救救攤平多單\n3. 賭徒謬誤（跌深「必」噴、V 轉信仰）\r#\r上週跌多了，這天滿屏都是「要噴回 5 萬」「V 爆」——把一根技術性反彈，腦補成會自我修復的必然。喊「噴」喊了一整個早盤，行情卻是越彈越無力。\n#63　推　user7014：早安 要噴回5萬囉 #303　推　user7167：台股世界無敵強 要\u0026hellip;要..噴了!!!!Ｖ爆！！！！ #351　噓　user7157：日韓終極反彈 台股要噴了\n也有人早早就看穿這顆球彈不高：\n#1342　→　user7309：鳥量反彈 笑\n4. 凹單／套牢（不賣就不算賠）\r#\r反彈沒能解套的人，故事換了一個版本繼續說——「等解套」「沒錢買了」。處分效應在彈不上去的日子，會把人凍在原地動彈不得。\n#4324　→　user7819：都套牢了 沒錢啦 #4398　推　user7265：貪狗套牢沒錢買了 #4632　推　user7264：沒量? 惜售(0) 套牢等解套(X)\n5. 後見之明（早就知道會這樣）\r#\r無論漲跌，「我早就講了」永遠有得說——彈起來的說「果然撿鑽石」，彈不動的說「早就說過安心空」。事後重寫記憶，兩邊都不缺嘴。\n#1837　推　user7511：上禮拜五果然是撿鑽石 #645　→　user7020：早就說過44444安心空 #2874　推　user7858：果然要上五萬\n6. 過度自信／畢業文崇拜\r#\rPTT 特有的「畢業文」文化——把別人的慘賠當娛樂指標、把自己五分鐘前的喊單當神諭，同時對盤勢過度自信。\n#150　推　user7082：畢業文 畢業文 畢業文 #338　推　user7166：開盤搶GG 一定開低走高收最高 相信我 #701　推　user7318：笑死 記住5分鐘前叫你放空的ID\n7. 從眾與反指（騙上車／騙下車）\r#\r反彈日最魔幻的一種偏誤：把每一次上漲都解讀成「主力在騙人上車」，把每一次下殺都解讀成「騙人下車」。決策外包給「跟群眾相反就對了」的反指信仰。\n#694　推　user7215：v了要騙人上車，開高走低盤 #835　噓　user7288：就說這周拉G上5萬 鬼指最後清洗 還有人沒上車 #1195　推　user7468：又想騙人上車喔 八成等等急轉直下\n8. 明牌／權威迷信（外資、主力、第四大法人）\r#\r把走勢歸因給看不見的「外資內線／主力劇本」，甚至自封 PTT 鄉民是「第四大法人」——用想像中的權威，去解釋其實隨機的波動。\n#416　→　user7198：本日劇本：第四法人開幹 軋爆無恥小兒？ #464　推　user7170：外資 ：今天繼續到貨放空 #1007　推　user7055：外資小兒 通通排隊補貨買回來\n9. 技術分析迷信（季線、月線是神諭）\r#\r把「季線／月線」當成有意志的支撐與壓力，會「守」會「破」會「自己上來」。線本身沒有魔力，但這天「殺回季線」「測月線」喊得像在問神。\n#257　推　user7146：目標不變 殺回季線 #652　推　user7155：在撐幾天 季線自己會上來當支撐 #634　推　user7132：怎麼可能還會跌 都到月線了\n10. 末日／恐慌迷因（逃命波、熔斷）\r#\r即使大盤早盤急彈，「逃命」「熔斷」「大爆崩」的災難語言照樣洗版——情緒宣洩跟行情脫鉤，恐慌變成一種玩法，而不是反應。\n#1　推　user7001：早安大爆崩 #70　噓　user7020：暗韓熔斷 #103　推　user7048：最後逃命機會\n11. 外部歸因（都是台積電／川普／匯率的錯）\r#\r把帳面損益外包給單一代罪羔羊——神山、川普、匯率、地緣——迴避「自己選了這個方向」這件事。\n#135　推　user7072：跌台積電就好了 拜託 #2466　推　user7773：川普說禮拜一股市要漲 #2004　推　user7665：台幣又再貶值了\n寫在最後\r#\r兩天放在一起看特別清楚：上週五崩盤日，鄉民的嘴是「接刀、攤平、大爆崩」；這個週一反彈日，嘴換成了「要噴、逃命波、騙上車、第四法人」。行情換了方向，謬誤只是換了台詞。 同一批偏誤——沉沒成本、賭徒謬誤、損失趨避、陰謀論、技術線迷信——在漲在跌都會準時報到。\n而且有一個很反差的發現：鄉民的嘴常常比手清醒。一堆人早盤就看出這是「鳥量反彈、吐得回去」，事後也確實高點回吐了一半；但看穿歸看穿，手照樣手癢想接、想凹、想賭它噴回五萬——知道，從來不等於做到。\n這張圖的價值不在預測明天，而在事後對照：下一次你又想「跌這麼多總該彈了吧、這根我追不追」的時候，回來看看 06/29 這天，那些一整個早盤喊「要噴回 5 萬」的留言，後面接的是什麼。\n資料來源：PTT Stock 板 2026/06/29 盤中閒聊全量留言（BBS 層抓取）、台灣證交所盤中加權指數。所有帳號已置換為不可回推的合成代號。\n本文為個人資料實驗與行為觀察紀錄，所有引用留言均為公開內容、僅作教育與心理觀察用途，不構成任何投資建議，也不針對任何個人。指數資料可能有延遲或誤差，投資決策請自行評估、自負風險。\n","date":"2026年6月29日","externalUrl":null,"permalink":"/zh-tw/stocks/taiex-chat-0629/","section":"股票","summary":"","title":"「反彈是真的嗎」：7,061 則盤中留言的「逃命波」考古（2026/06/29）","type":"stocks"},{"content":"這是一篇資料實驗。2026/06/26（週五）台股開高走低、終場大跌，我把 PTT Stock 板當天那篇「盤中閒聊」的全部 13,927 則留言抓下來，逐則對齊「發文那一刻」的加權指數，做成下面這張可以互動的圖。指數在跌的時候，鄉民的嘴在做什麼？\n本篇目前列入「2026 台股前十大下跌日」第 6 名。正文保留當日擷取的 13,927 則；2026/07/28 重新歸檔時共有 13,932 則，多出的 5 則是晚到推文。匿名化完整檔：純文字 · JSONL。\n這天發生了什麼\r#\r數值 開盤 46,255.26 收盤 44,571.76 盤中最低 44,455.67（13:24） 漲跌 ▼ 1,683.50（−3.64%） 幾乎是開盤即最高、一路盤跌、尾盤摜到最低的「陰線吞天」。當天那篇盤中閒聊的第一則留言，時間 08:30，內容只有五個字：\n#1　推　user3543：早安大爆崩\n接著「早安大爆崩」被當成貼圖一樣連刷了十幾樓。崩盤在 PTT 不是恐慌，是一種集體迷因——這正是這份資料最有趣的地方。\n資料是怎麼來的\r#\r全量留言：PTT 網頁版（www.ptt.cc）對推爆文有「檔案過大！部分文章無法顯示」的上限，只會吐出開頭約一千則加上抓取當下的最新尾段，中間整段被藏起來——當天網頁版只看得到約 1,489 則。真正的完整內容在 BBS 原始檔裡，走 BBS 層（PyPtt）可以拿到全部 13,927 則、沒有斷層。 指數：台灣證交所盤中 5 秒級的「發行量加權股價指數」，每分鐘取樣，對齊留言時間。 隱私：所有留言者帳號都置換為合成代號（例如 exampleuser → user0042）。代號與原帳號沒有任何對應關係，不可回推；同一個代號在所有檔案裡指同一人。 互動圖：把留言釘在指數上\r#\r在圖上左右滑動或點擊，或拖曳下方的時間軸，就會看到那一分鐘的加權指數、以及當下湧出的留言。底部灰色小柱是「每分鐘留言量」——你會發現留言的密度，和指數的劇烈波動高度重疊。\n📈 加權指數 × 鄉民留言 開 收 最低 留言 則・推噓比 · 加權 · 則 全部 推 噓 → 資料:PTT Stock 板當日盤中閒聊全量留言(走 BBS 層抓取,網頁版受「檔案過大」限制只剩約一成); 加權指數為台灣證交所 5 秒級即時資料,每分鐘取樣對齊。留言者帳號已置換為不可回推的合成代號。 拖曳時間軸或在圖上滑動 / 點擊,即可看當下指數與該分鐘的留言。 指數在跌，嘴在做什麼\r#\r跑完全量這幾個數字很反直覺：\n推噓比 6.33（推 8,096／噓 1,280／→ 4,551）。大跌 3.64% 的日子，看板情緒竟然還是「推」壓倒性多——因為這裡的「推」大多不是看多，而是看戲、嘴砲、刷迷因。 留言在 11:00–11:15 爆量（單分鐘最高 122 則），對應指數在 44,500 一帶劇烈來回。人不是在下跌時最吵，是在「劇烈波動、方向不明」時最吵。 從早盤到尾盤，每個小時「推」都遠多於「噓」。集體焦慮在 PTT 的出口不是恐慌賣壓，是玩梗。 下面把當天最常見的散戶思考謬誤盤點成 11 類，每類附 3 則真實留言（樓層＋合成代號＋原文照錄）。這些不是要嘲笑誰——是一面鏡子，照的是我們每個人在帳面虧損當下，腦子會自動跑出來的那些話。\n1. 接刀／抄底謬誤（把下跌當折扣）\r#\r把「跌了」直接等同於「便宜了、該買了」，忽略下跌往往反映基本面或趨勢轉變。錨定在過去的高價，覺得現在「相對便宜」。\n#120　推　user5795：只要開太低就可以抄底了 #542　→　user2716：一時抄底一時爽 一直抄底一直爽！ #614　推　user2453：南部大媽們準備歐印抄底了\n2. 攤平／向下加碼（降低成本的錯覺）\r#\r越跌越買來「拉低平均成本」，把問題部位越養越大。把「攤平」當成穩贏的數學，忽略它其實是在加碼一個正在下跌的賭注。\n#5645　推　user1581：4w5加碼攤平，夜盤4w4攤平，總有一天會贏的 #6411　推　user1842：猜不到底就用力往下加碼攤平，跟本不會輸好嗎? #10333　推　user6574：融資攤平呀 無腦買\n3. 賭徒謬誤（跌深「必」反彈）\r#\r「跌這麼多了，總該彈了吧」——把獨立的價格走勢，當成會自我修正的賭局。當天「跌不下去了 噴！」喊了一整天，指數照樣破底。\n#313　噓　user1506：明顯跌不下去了 噴！！！！！！！ #1811　噓　user2327：台股好硬喔 跌不下去欸 #2668　噓　user4792：484跌不下去\n有人很早就看穿這個循環：\n#387　推　user3609：每天都說跌不下去，然後繼續破底\n4. 凹單／損失趨避（不賣就不算賠）\r#\r帳面虧損時，大腦會編故事來逃避「實現虧損」的痛——「沒賣就沒輸」「這是長期投資」。處分效應的經典現場。\n#3093　→　user5872：皇翔只要不賣就不算賠!! #3670　→　user4265：撐不住 跟大家一起住套房了 #5490　→　user3085：還有人覺得是洗盤 笑死 你套牢安慰自己當長投\n5. 後見之明（早就知道會這樣）\r#\r事後把不確定的結果說成「我早就講了」，重寫記憶、虛增自己的預測能力。\n#730　推　user6408：我就說旺宏一定行 #3174　推　user1875：10點果然開拉 #4963　→　user4384：45k早就該破了 還死撐\n6. 過度自信／嘴砲賭神（畢業文崇拜）\r#\rPTT 特有的「畢業文」文化——把別人的慘賠當娛樂指標，同時對自己的判斷過度自信。\n#555　推　user6698：別急 還沒到底 PTT還沒看到畢業文 #306　噓　user6408：萬五列車等畢業文出來就要發車了 #613　噓　user3955：記住那些四萬多還叫你無腦做多的ID 可以通通黑單了\n7. 從眾／FOMO／韭菜框架\r#\r用「大家都…」「散戶都…」來定位自己，把交易決策外包給群體情緒。\n#119　→　user6390：大家都下去了 台股一定88 不用想了☺ #553　推　user6238：一堆空手韭菜 看來8成賠錢是真的 #1054　→　user5257：到底了 大家都上船了嗎\n8. 明牌／權威迷信（內線與「第四大法人」）\r#\r把走勢歸因給看不見的「主力／外資內線」，甚至自封 PTT 鄉民是「第四大法人」——用想像的權威來解釋隨機波動。\n#266　推　user1631：伊朗又開始搞事啦 外資有內線早知道？ #2298　噓　user5784：暗主力說：11:45開拉 #573　→　user5849：GG今天特價，第4大法人要買要快\n9. 技術分析迷信（線就是神諭）\r#\r把「月線／季線」當成有意志的支撐，會「守」會「破」。線本身沒有魔力，但當天「守月線」「測季線」喊得像在問神。\n#176　推　user0484：頂多跌到季線 不怕 #248　推　user2009：月線已經撐好幾次了 很給力了 不要太不滿 #430　推　user0610：月線跟套套一樣薄瞬間就破了\n10. 末日／恐慌敘事（大爆崩迷因）\r#\r把單日下跌升級成「股災末日」「AI is over」，用災難化的語言宣洩情緒——但同時又當成玩梗。\n#1　推　user3543：早安大爆崩 #36　→　user2003：先看四萬 再看三萬 完蛋了 AI is over #163　推　user2099：起初 所以人都以為只是一次普通的回檔 直到股災末日來臨\n11. 外部歸因（都是台積電／匯率／川普的錯）\r#\r把帳面損失外包給單一代罪羔羊，迴避「自己選了會跌的部位」這件事。\n#259　→　user3299：台積電根本拖累大家 #809　推　user6738：匯率又崩 誰還在買股 #944　推　user1621：川普最近好安靜…\n寫在最後\r#\r把 13,927 則留言對齊指數之後，最強烈的感受不是「鄉民好好笑」，而是：這些話我自己在帳面虧損當下也都說過。接刀、攤平、凹單、跟著大家、怪台積電——每一種謬誤都不是別人的病，是人腦在虧損壓力下的預設反應。\n這張圖的價值，不在於預測明天，而在於事後對照：當你下一次手又癢、又想「跌這麼多該抄了吧」的時候，回來看看 06/26 這天，那些一整天喊「跌不下去」的留言，後面接的是什麼。\n資料來源：PTT Stock 板 2026/06/26 盤中閒聊全量留言（BBS 層抓取）、台灣證交所盤中加權指數。所有帳號已置換為不可回推的合成代號。\n本文為個人資料實驗與行為觀察紀錄，所有引用留言均為公開內容、僅作教育與心理觀察用途，不構成任何投資建議，也不針對任何個人。指數資料可能有延遲或誤差，投資決策請自行評估、自負風險。\n","date":"2026年6月26日","externalUrl":null,"permalink":"/zh-tw/stocks/taiex-chat-0626/","section":"股票","summary":"","title":"「指數在跌，嘴在做什麼」：13,927 則盤中留言的謬誤考古（2026/06/26）","type":"stocks"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/tags/ai-training/","section":"Tags","summary":"","title":"AI Training","type":"tags"},{"content":"","date":"2026년 6월 24일","externalUrl":null,"permalink":"/ko/tags/ai-%ED%95%99%EC%8A%B5/","section":"Tags","summary":"","title":"AI 학습","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/ja/tags/ai%E5%AD%A6%E7%BF%92/","section":"Tags","summary":"","title":"AI学習","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/zh-tw/tags/ai%E8%A8%93%E7%B7%B4/","section":"Tags","summary":"","title":"AI訓練","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/zh-hans/tags/ai%E8%AE%AD%E7%BB%83/","section":"Tags","summary":"","title":"AI训练","type":"tags"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/tags/algorithm/","section":"Tags","summary":"","title":"Algorithm","type":"tags"},{"content":"","date":"24 janvier 2026","externalUrl":null,"permalink":"/fr/tags/algorithme/","section":"Tags","summary":"","title":"Algorithme","type":"tags"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/id/tags/algoritma/","section":"Tags","summary":"","title":"Algoritma","type":"tags"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/tags/attention-economy/","section":"Tags","summary":"","title":"Attention Economy","type":"tags"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/posts/","section":"Blog","summary":"","title":"Blog","type":"posts"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/categories/","section":"Categories","summary":"","title":"Categories","type":"categories"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/vi/categories/%C4%91%C3%A0i-thi%C3%AAn-v%C4%83n/","section":"Categories","summary":"","title":"Đài Thiên Văn","type":"categories"},{"content":"","date":"24 janvier 2026","externalUrl":null,"permalink":"/fr/tags/d%C3%A9veloppement-ind%C3%A9pendant/","section":"Tags","summary":"","title":"Développement Indépendant","type":"tags"},{"content":"","date":"24 janvier 2026","externalUrl":null,"permalink":"/fr/tags/%C3%A9conomie-de-lattention/","section":"Tags","summary":"","title":"Économie De L'attention","type":"tags"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/id/tags/ekonomi-perhatian/","section":"Tags","summary":"","title":"Ekonomi Perhatian","type":"tags"},{"content":"","date":"24 janvier 2026","externalUrl":null,"permalink":"/fr/tags/entra%C3%AEnement-de-lia/","section":"Tags","summary":"","title":"Entraînement De L'IA","type":"tags"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/vi/tags/gen/","section":"Tags","summary":"","title":"Gen","type":"tags"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/tags/gene/","section":"Tags","summary":"","title":"Gene","type":"tags"},{"content":"","date":"24 janvier 2026","externalUrl":null,"permalink":"/fr/tags/g%C3%A8ne/","section":"Tags","summary":"","title":"Gène","type":"tags"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/vi/tags/hu%E1%BA%A5n-luy%E1%BB%87n-ai/","section":"Tags","summary":"","title":"Huấn Luyện AI","type":"tags"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/tags/indie-development/","section":"Tags","summary":"","title":"Indie Development","type":"tags"},{"content":"","date":"24 janvier 2026","externalUrl":null,"permalink":"/fr/categories/lobservatoire/","section":"Categories","summary":"","title":"L'Observatoire","type":"categories"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/tags/meme/","section":"Tags","summary":"","title":"Meme","type":"tags"},{"content":"","date":"24 janvier 2026","externalUrl":null,"permalink":"/fr/tags/m%C3%A8me/","section":"Tags","summary":"","title":"Mème","type":"tags"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/vi/tags/n%E1%BB%81n-kinh-t%E1%BA%BF-ch%C3%BA-%C3%BD/","section":"Tags","summary":"","title":"Nền Kinh Tế Chú Ý","type":"tags"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/id/categories/observatorium/","section":"Categories","summary":"","title":"Observatorium","type":"categories"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/id/tags/pelatihan-ai/","section":"Tags","summary":"","title":"Pelatihan AI","type":"tags"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/id/tags/pengembangan-indie/","section":"Tags","summary":"","title":"Pengembangan Indie","type":"tags"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/vi/tags/ph%C3%A1t-tri%E1%BB%83n-%C4%91%E1%BB%99c-l%E1%BA%ADp/","section":"Tags","summary":"","title":"Phát Triển Độc Lập","type":"tags"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/","section":"QQder · The Miniature Boat","summary":"","title":"QQder · The Miniature Boat","type":"page"},{"content":"This article is about the concept and origin of the Sown Echoes app. It touches on my own experience at school and how I came to understand the ideas behind it.\nWhen I was in middle school, a writer once came to give a talk at our school. In the grand auditorium, students from every grade were present. Not long into his talk, the writer told a short story: A man had been walking alone through the desert for many days. He kept pushing on, hoping to be rescued. As time went by, his supplies and water had long run out, and there was no hope of survival left. In the very instant before he died, he ejaculated.\nThe story was so short, and its ending so abrupt — there was something literary in it. Saying something so explicit, in public, to several hundred middle schoolers, caused an uproar in the room at the time. But I didn\u0026rsquo;t think much of it. Because we were at an age of sexual curiosity, my classmates were thrilled to hear an adult say such words out loud; yet at the same time, being middle schoolers, we were innocent enough to still ponder the philosophy in it. The story is roughly about this: life will do everything it can to continue life itself, no matter how vanishingly small the chance. Just like the desert traveler who, on the verge of death, instinctively ejaculated under conditions where it could almost certainly have no effect.\nWhat does this have to do with the Sown Echoes app and the title \u0026ldquo;meme lives\u0026rdquo;? The little desert story is about the survival of life in the material world. After the information world was constructed by humans, the survival of the meme — as a concept of the mind — within the information world is similar. The individual is small and powerless, but as a whole it is instead strong and hard to extinguish; it keeps copying and modifying itself, and the death or disappearance of a few individuals is negligible from the perspective of the whole. The data on a single computer or a single storage medium will, given enough time, be utterly lost to failure or misplacement. But ever since the internet, memes — like genes — keep copying, spreading, and evolving. The various systems that host articles and videos store data in a relatively high-availability way, and the crawlers of every search engine go and scrape this data that is already exposed on the net. A single piece of information is fragile, but once it is put on the internet, it instead becomes hard to erase.\nThe meme, to some degree, reenacts the laws of the gene. But the meme goes a step further, beyond the scale of the gene: it is more abstract and its cost is lower. The meme is cross-species and cross-time; it can spread among any species of intelligent life, and it can cross vast spans of time. Each time a meme is thought by intelligent life, it comes alive again.\nIf you\u0026rsquo;re starting to get confused, just remember that all of this is about staying alive — only by continually transcending and expanding the dimension. The desert traveler trying to walk out of the desert and be rescued is the most basic continuation of the individual body. His uncontrollable and seemingly meaningless ejaculation is the will of the species hoping to continue the survival of the species. By this point, for the traveler, it already transcends his individual self; what remains is an unknowable future. The whole is already the future — abstract to the individual, and already unrelated to the individual. The meme is simply more abstract still. Some people care only about the survival of their own body and their present consciousness; others can more actively imagine a future not directly related to themselves. But regardless of their will or inclination, none can change the laws of the whole.\nTo get concrete about the memes related to us, think now about how our own memes get passed down. The typical scene is this: on apps like X (Twitter) or YouTube/TikTok, our viewing and all our other habits are tracked. This data is used by algorithms to push more videos or articles we might watch for longer durations, and the basis for that is our dwell time, likes, shares, saves, and similar data. It\u0026rsquo;s not just content consumers; if you are a content producer, you are equally shaped by the algorithm. If the videos you produce don\u0026rsquo;t win the algorithm\u0026rsquo;s favor, then economically or in terms of psychological motivation you not only can\u0026rsquo;t sustain it — in reality, not many people will see your content either. In the end, the memes generated on these platforms, because of the weighting given to dwell time, have drifted far away from the form in which we survive in the material world. This isn\u0026rsquo;t to say it is wrong, but that it becomes more entertaining and more utilitarian. Entertainment and utility are only one part of human society. If your children or grandchildren mostly inherited only the parts of you that concern entertainment and utility, wouldn\u0026rsquo;t you wish they also possessed more, much more of the traits you have? For example: selflessness, anger, shame, patience, seriousness, humility, prudence, gentleness, long-termism, silence, neutrality and objectivity, and so on… these traits that are innately at odds with the algorithm would all be heavily diluted. Other content in real life that is more private, and more easily censored by terms of service, becomes almost invisible. And there are many real experiences and traits — qualities inherently incompatible with the medium of video or posts — that cannot be left behind at all.\nThis is one of the reasons I made the Sown Echoes app: on the premise that we ourselves are ready, we express and articulate the experiences we find beautiful, true, and otherwise worth recording — rather than simply letting every app track our every little movement and letting these trivial bits of information define us. The highest principle of a tech company\u0026rsquo;s app algorithm is to maximize the time the user stays in the app. Unless your life\u0026rsquo;s goal is to spend as much time as possible in some app, it is ultimately not your final value. Human value is directional; not all trivial facts are important. For instance, when I am invited to give an address, I select and distill from life experience how we ought to act and how we ought to think — rather than printing out my browsing history and the trivial things I did each day. Humans have their own agency. What is, is merely what is; and we must pursue the value of what ought to be. For example, when we teach children, even if we ourselves cannot be perfect, we still teach them the most ideal ideas.\nWhile every app runs a similar algorithm, actively producing articles that conform to one\u0026rsquo;s own values seems futile. But fortunately, we have entered the age of AI, and we no longer have to fear that our articles must rack up many views in order to be passed down. At the moment of writing this, the native data that AI can train on is nearly exhausted; any meaningful article on the internet will be taken for training. Besides letting you converse with your own echo without sharing your privacy, Sown Echoes — if you choose to contribute your articles within it — has an upload format already designed to be suitable for AI training, making it very high-quality training material for researchers at AI labs. Your articles will become part of the weights of future large models, and your thoughts can live on forever together with the model.\nSuch an act seems, and indeed almost certainly is, futile — including my making this app itself, just like the desert traveler in the instant before death. The difference from the gene is that, within the realm of the meme, we have a higher chance of being left behind.\n","date":"24 June 2026","externalUrl":null,"permalink":"/posts/sown-echoes-meme-lives/","section":"Blog","summary":"","title":"Sown Echoes: Meme Lives","type":"posts"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/tags/","section":"Tags","summary":"","title":"Tags","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/zh-tw/categories/the-cabin/","section":"Categories","summary":"","title":"The Cabin","type":"categories"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/categories/the-observatory/","section":"Categories","summary":"","title":"The Observatory","type":"categories"},{"content":"","date":"24 June 2026","externalUrl":null,"permalink":"/vi/tags/thu%E1%BA%ADt-to%C3%A1n/","section":"Tags","summary":"","title":"Thuật Toán","type":"tags"},{"content":"","date":"2026년 6월 24일","externalUrl":null,"permalink":"/ko/tags/%EB%B0%88/","section":"Tags","summary":"","title":"밈","type":"tags"},{"content":"","date":"2026년 6월 24일","externalUrl":null,"permalink":"/ko/tags/%EC%95%8C%EA%B3%A0%EB%A6%AC%EC%A6%98/","section":"Tags","summary":"","title":"알고리즘","type":"tags"},{"content":"","date":"2026년 6월 24일","externalUrl":null,"permalink":"/ko/tags/%EC%96%B4%ED%85%90%EC%85%98-%EC%9D%B4%EC%BD%94%EB%85%B8%EB%AF%B8/","section":"Tags","summary":"","title":"어텐션 이코노미","type":"tags"},{"content":"","date":"2026년 6월 24일","externalUrl":null,"permalink":"/ko/tags/%EC%9C%A0%EC%A0%84%EC%9E%90/","section":"Tags","summary":"","title":"유전자","type":"tags"},{"content":"","date":"2026년 6월 24일","externalUrl":null,"permalink":"/ko/tags/%EC%9D%B8%EB%94%94-%EA%B0%9C%EB%B0%9C/","section":"Tags","summary":"","title":"인디 개발","type":"tags"},{"content":"","date":"2026년 6월 24일","externalUrl":null,"permalink":"/ko/categories/%EC%B2%9C%EB%AC%B8%EB%8C%80/","section":"Categories","summary":"","title":"천문대","type":"categories"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/ja/tags/%E3%82%A2%E3%83%86%E3%83%B3%E3%82%B7%E3%83%A7%E3%83%B3%E3%82%A8%E3%82%B3%E3%83%8E%E3%83%9F%E3%83%BC/","section":"Tags","summary":"","title":"アテンション・エコノミー","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/ja/tags/%E3%82%A2%E3%83%AB%E3%82%B4%E3%83%AA%E3%82%BA%E3%83%A0/","section":"Tags","summary":"","title":"アルゴリズム","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/ja/tags/%E3%83%9F%E3%83%BC%E3%83%A0/","section":"Tags","summary":"","title":"ミーム","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/ja/tags/%E5%80%8B%E4%BA%BA%E9%96%8B%E7%99%BA/","section":"Tags","summary":"","title":"個人開発","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/zh-tw/tags/%E5%9F%BA%E5%9B%A0/","section":"Tags","summary":"","title":"基因","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/zh-tw/tags/%E6%A8%A1%E5%9B%A0/","section":"Tags","summary":"","title":"模因","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/zh-tw/tags/%E6%B3%A8%E6%84%8F%E5%8A%9B%E7%B6%93%E6%BF%9F/","section":"Tags","summary":"","title":"注意力經濟","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/zh-hans/tags/%E6%B3%A8%E6%84%8F%E5%8A%9B%E7%BB%8F%E6%B5%8E/","section":"Tags","summary":"","title":"注意力经济","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/zh-tw/tags/%E6%BC%94%E7%AE%97%E6%B3%95/","section":"Tags","summary":"","title":"演算法","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/zh-hans/tags/%E7%8B%AC%E7%AB%8B%E5%BC%80%E5%8F%91/","section":"Tags","summary":"","title":"独立开发","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/zh-tw/tags/%E7%8D%A8%E7%AB%8B%E9%96%8B%E7%99%BC/","section":"Tags","summary":"","title":"獨立開發","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/zh-hans/tags/%E7%AE%97%E6%B3%95/","section":"Tags","summary":"","title":"算法","type":"tags"},{"content":"","date":"2026年6月24日","externalUrl":null,"permalink":"/ja/tags/%E9%81%BA%E4%BC%9D%E5%AD%90/","section":"Tags","summary":"","title":"遺伝子","type":"tags"},{"content":"這是「2026 台股前十大下跌日」系列第 8 名。排名口徑是 2026/01/01 至 2026/07/29 的加權指數單日收盤報酬率，不是盤中最大跌幅。\n前一天大彈之後，指數開低、短暫摸高，最後收在最低。前一日的劇本成了這一日最危險的先驗。\n這天發生了什麼\r#\r數值 前收 44,704.44 開盤 44,581.45 最高 44,676.49 最低 43,225.54 收盤 43,225.54 漲跌 ▼ 1,478.90（-3.31%） 2026 跌幅排名 第 8 名 BBS 全量留言\r#\rPTT 網頁版遇到推爆長文會省略中段；這裡走 BBS 層抓到 10,018 則，不是網頁殘片。\n推文型態 數量 推 5,849 噓 947 → 3,222 合計 10,018 留言最密集的時段是 12:00（2,663 則）。\n時段 留言數 相對量 08:00 765 █████ 09:00 2,194 █████████████ 10:00 1,478 █████████ 11:00 989 ██████ 12:00 2,663 ████████████████ 13:00 1,900 ███████████ 14:00 26 █ PTT 原文：AID 1gAB0CM0 匿名化原始檔：純文字 · JSONL 8 種謬誤與偏誤\r#\r以下是語句模式的分類，不是對留言者作人格診斷。盤中閒聊包含反串、迷因與情緒宣洩；引文只能證明這句話出現過，不能證明作者真的照著交易。\n1. 單因謬誤：把市場縮成一個黑手\r#\r市場同時包含外資、內資、避險、被動資金與個別公司消息；把整段價格路徑只歸因於一個有意志的角色，故事會很順，證據卻通常不夠。\n#1676 · 06/10 09:18 · user7077：大媽的戰力恐怖如斯 硬拉回平盤？？？？？？\n#6009 · 06/10 12:13 · user7078：內外資大賣，這盤只剩韭菜了\n2. 賭徒謬誤：跌多了就「該」彈\r#\r先前已經跌多少，不會自動提高下一分鐘上漲的機率。價格可以很便宜，也可以在缺乏新資訊時繼續變便宜。\n#2360 · 06/10 09:38 · user7075：多跌一點 我嘴巴準備好抄底了\n#7497 · 06/10 12:42 · user7079：早上賣了部分0050撿便宜的凱子金，919真的好穩\n3. 錨定效應：月線、季線與整數都是答案\r#\r技術位置可以是風險管理參考，但把單一價位當成必然反轉點，會忽略波動、成交量與事件條件。\n#5409 · 06/10 11:59 · user7080：睡了一覺起來大盤都還沒回測月線，今天這麼硬\n#8757 · 06/10 13:20 · user7081：差不多季線就止跌了吧 空方很快 一兩天就到 忍一下\n4. 直線外推：今天的斜率一路畫到明天\r#\r人腦很容易把眼前最強烈的方向延伸出去；市場真正困難之處，正是斜率會在新資訊出現時改變。\n#3067 · 06/10 10:03 · user7082：晚上CPI一開 明天剛好跌破4萬 走著瞧 拜託不要..\n#8384 · 06/10 13:09 · user7083：衝一下今天千五，明天三千。貪狗一個都別走\n5. 從眾效應：多蛙、空蛙與「大家都知道」\r#\r群體標籤能快速製造安全感，也會把不同持倉、期限與風險承受度的人壓成同一種對手。\n#2389 · 06/10 09:39 · user7084：南電加油！ 晚點要起飛了 大家快check in\n#7431 · 06/10 12:42 · user7085：應該沒人敢笑 小兒了吧\n6. 後見之明偏誤：收盤後每個人都早知道\r#\r結果出現後，原先的多種可能性會從記憶裡消失。真正可檢驗的不是「早說」，而是事前是否留下方向、期限與失效條件。\n#2155 · 06/10 09:31 · user7086：怎麼這麼早就崩了\n#7349 · 06/10 12:40 · user7087：8zz買正二果然威力還是太強了\n7. 類比謬誤：把日韓、美股或夜盤直接翻譯\r#\r跨市場確實相關，但交易時段、權重、匯率與事件曝險不同；相關不等於可以一比一複製漲跌。\n#1389 · 06/10 09:13 · user7088：夜盤到底是三小 我笑了\n#5447 · 06/10 12:01 · user7089：美股破底台股沒破底，美股反彈台股五萬五過端午\n8. 災難化與全有全無：不是 V，就是歸零\r#\r這比較接近認知扭曲而非形式邏輯謬誤：把連續的風險壓成生或死兩個選項，會讓部位管理退化成情緒口號。\n#2865 · 06/10 09:56 · user7025：CPO救命\u0026hellip;\u0026hellip;\u0026hellip;\u0026hellip;.\n#7434 · 06/10 12:42 · user7090：金融又紅通通啦 這次完蛋啦 XDDDDD\n這份考古不能證明什麼\r#\r它不是情緒指標回測，也沒有證明某類留言能預測下一根 K 線。 關鍵字分類是可重現的抽樣入口，不是對全部 10,018 則留言做唯一正解標註。 同一句話可能同時包含多種偏誤；為了可讀性，每則引文只放在一個小節。 帳號已置換為合成代號（user0001 之類），與原帳號無對應關係、不可回推；保留樓層與時間，是為了讓讀者能回到匿名化原始檔核對。 2026 年截至 07/29 的前十大下跌日\r#\r名次 日期 收盤跌幅 跌點 文章 1 2026-07-17 -6.47% -2,953.71 「跌到收盤才停」 2 2026-07-28 -4.65% -2,030.83 「這次一定會 V？」 3 2026-03-09 -4.43% -1,489.12 「千點低開之後」 4 2026-03-04 -4.35% -1,494.77 「月線會救嗎」 5 2026-07-29 -3.76% -1,564.18 「四萬點保衛戰」 6 2026-06-26 -3.64% -1,683.50 「指數在跌，嘴在做什麼」 7 2026-06-08 -3.48% -1,568.16 「夜盤只是誤會？」 8 2026-06-10 -3.31% -1,478.90 「反彈第二天又殺」（本篇） 9 2026-07-24 -2.67% -1,195.97 「日韓跌，台股就會？」 10 2026-03-23 -2.45% -821.38 「明天會噴回來？」 資料口徑\r#\r指數排名與 OHLC：Yahoo Finance ^TWII 日線快照，抓取日 2026/07/29；排名用相鄰交易日收盤價計算。臺灣證券交易所亦提供發行量加權股價指數歷史資料供核對。 留言：PTT Stock 板當日「盤中閒聊」，以 BBS 層完整抓取；本篇產生前已核對 AID、標題與留言總數。 隱私：公開原始檔與本文引文的帳號一律置換為合成代號，不保留原字元、不可回推。同一個代號在所有檔案裡指同一人，跨日比對仍然成立。 ","date":"2026年6月10日","externalUrl":null,"permalink":"/zh-tw/stocks/taiex-chat-0610/","section":"股票","summary":"","title":"「反彈第二天又殺」：10,018 則盤中留言的謬誤考古（2026/06/10）","type":"stocks"},{"content":"這是「2026 台股前十大下跌日」系列第 7 名。排名口徑是 2026/01/01 至 2026/07/29 的加權指數單日收盤報酬率，不是盤中最大跌幅。\n低開後一度再殺超過兩千點，接著從低點反抽逾千點。方向沒變，盤中故事卻反覆改寫。\n這天發生了什麼\r#\r數值 前收 45,070.94 開盤 44,507.49 最高 44,507.49 最低 42,376.86 收盤 43,502.78 漲跌 ▼ 1,568.16（-3.48%） 2026 跌幅排名 第 7 名 BBS 全量留言\r#\rPTT 網頁版遇到推爆長文會省略中段；這裡走 BBS 層抓到 14,270 則，不是網頁殘片。\n推文型態 數量 推 8,526 噓 1,104 → 4,640 合計 14,270 留言最密集的時段是 09:00（4,723 則）。\n時段 留言數 相對量 08:00 2,063 ███████ 09:00 4,723 ████████████████ 10:00 2,956 ██████████ 11:00 1,774 ██████ 12:00 1,186 ████ 13:00 1,543 █████ 14:00 22 █ PTT 原文：AID 1g9WqENy 匿名化原始檔：純文字 · JSONL 8 種謬誤與偏誤\r#\r以下是語句模式的分類，不是對留言者作人格診斷。盤中閒聊包含反串、迷因與情緒宣洩；引文只能證明這句話出現過，不能證明作者真的照著交易。\n1. 單因謬誤：把市場縮成一個黑手\r#\r市場同時包含外資、內資、避險、被動資金與個別公司消息；把整段價格路徑只歸因於一個有意志的角色，故事會很順，證據卻通常不夠。\n#5400 · 06/08 09:32 · user7063：把強短的大媽在殺出去就可以漲了\n#10294 · 06/08 11:16 · user7064：國安基金不是都還沒動嗎？\n2. 賭徒謬誤：跌多了就「該」彈\r#\r先前已經跌多少，不會自動提高下一分鐘上漲的機率。價格可以很便宜，也可以在缺乏新資訊時繼續變便宜。\n#4029 · 06/08 09:14 · user7065：昨天大家說要抄底是真的\n#10212 · 06/08 11:13 · user7066：今早開盤我抄底抄到要開額度，現在提醒無量反彈\n3. 錨定效應：月線、季線與整數都是答案\r#\r技術位置可以是風險管理參考，但把單一價位當成必然反轉點，會忽略波動、成交量與事件條件。\n#6374 · 06/08 09:51 · user7067：就讓月線那些人都套牢 就會碰季線了阿 ==\n#12218 · 06/08 12:35 · user7068：今天不敢進場的 下週有季線能撿\n4. 直線外推：今天的斜率一路畫到明天\r#\r人腦很容易把眼前最強烈的方向延伸出去；市場真正困難之處，正是斜率會在新資訊出現時改變。\n#6280 · 06/08 09:49 · user7069：才跌5% 明天一根就回來了\n#12863 · 06/08 13:07 · user7070：明天應該要破月線了吧\n5. 從眾效應：多蛙、空蛙與「大家都知道」\r#\r群體標籤能快速製造安全感，也會把不同持倉、期限與風險承受度的人壓成同一種對手。\n#3898 · 06/08 09:13 · user7071：怎大家狂撿鑽石啊\n#10518 · 06/08 11:23 · user7072：沒人要砍股票 完了 會繼續殺 殺到年線\n6. 後見之明偏誤：收盤後每個人都早知道\r#\r結果出現後，原先的多種可能性會從記憶裡消失。真正可檢驗的不是「早說」，而是事前是否留下方向、期限與失效條件。\n#4294 · 06/08 09:17 · user7073：看推文一片樂觀就知道\n#10616 · 06/08 11:25 · user7074：一堆賣本夢比的還跌的比較少 台股果然是越爛越噴\n7. 類比謬誤：把日韓、美股或夜盤直接翻譯\r#\r跨市場確實相關，但交易時段、權重、匯率與事件曝險不同；相關不等於可以一比一複製漲跌。\n#5843 · 06/08 09:40 · user7059：=======日韓股都撐不住了 ccc====================\n#12752 · 06/08 13:02 · user7025：趕快全出 美股晚上恐續跌\n8. 災難化與全有全無：不是 V，就是歸零\r#\r這比較接近認知扭曲而非形式邏輯謬誤：把連續的風險壓成生或死兩個選項，會讓部位管理退化成情緒口號。\n#2843 · 06/08 09:05 · user7075：正二當機 完蛋了\n#10358 · 06/08 11:18 · user7076：結果沒人跑！超勇！！瞧不起主力逃命波\n這份考古不能證明什麼\r#\r它不是情緒指標回測，也沒有證明某類留言能預測下一根 K 線。 關鍵字分類是可重現的抽樣入口，不是對全部 14,270 則留言做唯一正解標註。 同一句話可能同時包含多種偏誤；為了可讀性，每則引文只放在一個小節。 帳號已置換為合成代號（user0001 之類），與原帳號無對應關係、不可回推；保留樓層與時間，是為了讓讀者能回到匿名化原始檔核對。 2026 年截至 07/29 的前十大下跌日\r#\r名次 日期 收盤跌幅 跌點 文章 1 2026-07-17 -6.47% -2,953.71 「跌到收盤才停」 2 2026-07-28 -4.65% -2,030.83 「這次一定會 V？」 3 2026-03-09 -4.43% -1,489.12 「千點低開之後」 4 2026-03-04 -4.35% -1,494.77 「月線會救嗎」 5 2026-07-29 -3.76% -1,564.18 「四萬點保衛戰」 6 2026-06-26 -3.64% -1,683.50 「指數在跌，嘴在做什麼」 7 2026-06-08 -3.48% -1,568.16 「夜盤只是誤會？」（本篇） 8 2026-06-10 -3.31% -1,478.90 「反彈第二天又殺」 9 2026-07-24 -2.67% -1,195.97 「日韓跌，台股就會？」 10 2026-03-23 -2.45% -821.38 「明天會噴回來？」 資料口徑\r#\r指數排名與 OHLC：Yahoo Finance ^TWII 日線快照，抓取日 2026/07/29；排名用相鄰交易日收盤價計算。臺灣證券交易所亦提供發行量加權股價指數歷史資料供核對。 留言：PTT Stock 板當日「盤中閒聊」，以 BBS 層完整抓取；本篇產生前已核對 AID、標題與留言總數。 隱私：公開原始檔與本文引文的帳號一律置換為合成代號，不保留原字元、不可回推。同一個代號在所有檔案裡指同一人，跨日比對仍然成立。 ","date":"2026年6月8日","externalUrl":null,"permalink":"/zh-tw/stocks/taiex-chat-0608/","section":"股票","summary":"","title":"「夜盤只是誤會？」：14,270 則盤中留言的謬誤考古（2026/06/08）","type":"stocks"},{"content":"本篇原列入「2026 台股前十大下跌日」；7/29 納入後移至第 11 名，文章保留作為系列比較樣本。\n早盤短暫摸高後一路走低。當價格反彈的記憶還很新，每一次拉抬都容易被命名成同一種劇本。\n這天發生了什麼\r#\r數值 前收 32,518.16 開盤 32,419.00 最高 32,453.38 最低 31,705.99 收盤 31,722.99 漲跌 ▼ 795.17（-2.45%） 2026 跌幅排名 第 11 名 BBS 全量留言\r#\rPTT 網頁版遇到推爆長文會省略中段；這裡走 BBS 層抓到 9,745 則，不是網頁殘片。\n推文型態 數量 推 5,589 噓 779 → 3,377 合計 9,745 留言最密集的時段是 12:00（2,316 則）。\n時段 留言數 相對量 08:00 1,003 ███████ 09:00 2,218 ███████████████ 10:00 1,166 ████████ 11:00 1,851 █████████████ 12:00 2,316 ████████████████ 13:00 1,165 ████████ 14:00 23 █ PTT 原文：AID 1fonMJF5 匿名化原始檔：純文字 · JSONL 8 種謬誤與偏誤\r#\r以下是語句模式的分類，不是對留言者作人格診斷。盤中閒聊包含反串、迷因與情緒宣洩；引文只能證明這句話出現過，不能證明作者真的照著交易。\n1. 單因謬誤：把市場縮成一個黑手\r#\r市場同時包含外資、內資、避險、被動資金與個別公司消息；把整段價格路徑只歸因於一個有意志的角色，故事會很順，證據卻通常不夠。\n#3482 · 03/31 10:16 · user7048：聚陽都快殺到10年線了，主力是有病是不是\n#7359 · 03/31 12:23 · user7049：981A外資啊,都倒幾天了,有人撿就拼命倒\n2. 賭徒謬誤：跌多了就「該」彈\r#\r先前已經跌多少，不會自動提高下一分鐘上漲的機率。價格可以很便宜，也可以在缺乏新資訊時繼續變便宜。\n#3338 · 03/31 10:08 · user7050：是好朋友才告訴你 這邊抄底等數錢\n#7437 · 03/31 12:25 · user7051：手癢剛剛抄底了 有種再跌 再攤平\n3. 錨定效應：月線、季線與整數都是答案\r#\r技術位置可以是風險管理參考，但把單一價位當成必然反轉點，會忽略波動、成交量與事件條件。\n#3141 · 03/31 09:55 · user7052：每天越v越低 反彈當買點 這次換季線蛙跳進來泡溫水\n#6992 · 03/31 12:15 · user7036：重演2022了，還不賣保守套兩年\n4. 直線外推：今天的斜率一路畫到明天\r#\r人腦很容易把眼前最強烈的方向延伸出去；市場真正困難之處，正是斜率會在新資訊出現時改變。\n#5943 · 03/31 11:55 · user7053：經歷過空頭年，在那裡大聲的說有自信一定撐得過XD\n#9207 · 03/31 13:29 · user7054：明天反彈欸 現在確定要砍\n5. 從眾效應：多蛙、空蛙與「大家都知道」\r#\r群體標籤能快速製造安全感，也會把不同持倉、期限與風險承受度的人壓成同一種對手。\n#2795 · 03/31 09:40 · user7055：空蛙狗怎麼不叫了 再叫啊 cc\n#7096 · 03/31 12:17 · user7056：今天沒跌1000算多蛙贏\n6. 後見之明偏誤：收盤後每個人都早知道\r#\r結果出現後，原先的多種可能性會從記憶裡消失。真正可檢驗的不是「早說」，而是事前是否留下方向、期限與失效條件。\n#1918 · 03/31 09:16 · user7057：早上不是說32000不破，拜託快買XD\n#7217 · 03/31 12:19 · user7058：國際股市早就跌瘋了，台股不可能獨強\n7. 類比謬誤：把日韓、美股或夜盤直接翻譯\r#\r跨市場確實相關，但交易時段、權重、匯率與事件曝險不同；相關不等於可以一比一複製漲跌。\n#2329 · 03/31 09:27 · user7059：夜盤-500 日盤+500\n#7605 · 03/31 12:29 · user7060：今晚美股要是再往下 明天29k見\n8. 災難化與全有全無：不是 V，就是歸零\r#\r這比較接近認知扭曲而非形式邏輯謬誤：把連續的風險壓成生或死兩個選項，會讓部位管理退化成情緒口號。\n#2706 · 03/31 09:37 · user7061：台積翻紅 人道走廊打開 大家塊陶 完了門就關了\n#6919 · 03/31 12:13 · user7062：靠背 煞車壞惹 塊陶\n這份考古不能證明什麼\r#\r它不是情緒指標回測，也沒有證明某類留言能預測下一根 K 線。 關鍵字分類是可重現的抽樣入口，不是對全部 9,745 則留言做唯一正解標註。 同一句話可能同時包含多種偏誤；為了可讀性，每則引文只放在一個小節。 帳號已置換為合成代號（user0001 之類），與原帳號無對應關係、不可回推；保留樓層與時間，是為了讓讀者能回到匿名化原始檔核對。 2026 年截至 07/29 的前十大下跌日\r#\r名次 日期 收盤跌幅 跌點 文章 1 2026-07-17 -6.47% -2,953.71 「跌到收盤才停」 2 2026-07-28 -4.65% -2,030.83 「這次一定會 V？」 3 2026-03-09 -4.43% -1,489.12 「千點低開之後」 4 2026-03-04 -4.35% -1,494.77 「月線會救嗎」 5 2026-07-29 -3.76% -1,564.18 「四萬點保衛戰」 6 2026-06-26 -3.64% -1,683.50 「指數在跌，嘴在做什麼」 7 2026-06-08 -3.48% -1,568.16 「夜盤只是誤會？」 8 2026-06-10 -3.31% -1,478.90 「反彈第二天又殺」 9 2026-07-24 -2.67% -1,195.97 「日韓跌，台股就會？」 10 2026-03-23 -2.45% -821.38 「明天會噴回來？」 資料口徑\r#\r指數排名與 OHLC：Yahoo Finance ^TWII 日線快照，抓取日 2026/07/29；排名用相鄰交易日收盤價計算。臺灣證券交易所亦提供發行量加權股價指數歷史資料供核對。 留言：PTT Stock 板當日「盤中閒聊」，以 BBS 層完整抓取；本篇產生前已核對 AID、標題與留言總數。 隱私：公開原始檔與本文引文的帳號一律置換為合成代號，不保留原字元、不可回推。同一個代號在所有檔案裡指同一人，跨日比對仍然成立。 ","date":"2026年3月31日","externalUrl":null,"permalink":"/zh-tw/stocks/taiex-chat-0331/","section":"股票","summary":"","title":"「逃命波再來一次？」：9,745 則盤中留言的謬誤考古（2026/03/31）","type":"stocks"},{"content":"","date":"2026年3月31日","externalUrl":null,"permalink":"/zh-tw/tags/2026%E5%A4%A7%E8%B7%8C%E6%97%A5/","section":"Tags","summary":"","title":"2026大跌日","type":"tags"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/vi/tags/c%C3%B4ng-c%E1%BB%A5-quy-t%E1%BA%AFc/","section":"Tags","summary":"","title":"Công Cụ Quy Tắc","type":"tags"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/tags/dream-of-the-red-chamber/","section":"Tags","summary":"","title":"Dream of the Red Chamber","type":"tags"},{"content":"App URL: LINK\nPreface\r#\rThe key point from the previous installment\nwas to regard text as fundamentally symbolic \u0026ndash;\nastronomy, hydrology, the humanities\u0026hellip; all the \u0026ldquo;wen\u0026rdquo; (文, pattern/text) of heaven, earth, and humankind.\nText maps the world and thought in a cost-effective way,\nbecoming our primary tool for understanding and interfacing with objective reality.\nOnce you grasp this, you realize that\nalthough LLMs (Large Language Models) are essentially just next-token predictors,\nonce their capability reaches a certain level, they become nuclear-grade instruments of national importance.\nTheir significance made me want to verify their capabilities\nand to do so repeatedly as they improve over time.\nA near-perfect benchmark for this is Dream of the Red Chamber (紅樓夢, Hong Lou Meng).\nSuppose there existed an omniscient, omnipotent LLM \u0026ndash;\nit could take Cao Xueqin\u0026rsquo;s original first 80 chapters of Dream of the Red Chamber as input and output the subsequent chapters.\nBut because LLM training data is limited,\nit is like a Sudoku puzzle with too few given numbers \u0026ndash; the answer cannot be determined with certainty.\nWhat current LLMs can do is produce at very high throughput within the scope of what they understand.\nWhat the Dream of the Red Chamber Simulator aims to do is, with such productivity at hand,\nuse traditional structured methods to rapidly produce and accumulate results with minimal human effort.\nAssumptions\r#\rWe need some assumptions, biases, and theories to make the task of predicting the ending sufficiently feasible and mechanical.\nWhen it comes to accurate prediction, my intuition goes to classical physics \u0026ndash; specifically thermodynamics.\nIn a closed system, if we specify the initial conditions and the governing laws,\nthe evolution of a thermodynamic system is predictable and deterministic.\nAnother assumption is that LLM capabilities will keep improving,\nbut in the foreseeable future we will not gain additional training data from the Qing Dynasty or from Cao Xueqin himself.\nTherefore, we can establish a structured workflow that both current and future LLMs can execute.\nInitial Conditions\r#\rThe initial conditions are primarily data extracted from the original novel.\nNow we use LLMs to perform what was previously highly labor-intensive work.\nIn the past, human labor costs were too high, and throwing more people at the problem could not compress the timeline.\nIf you got halfway through and wanted to tweak the extraction rules and start over, it was simply impractical.\nTime and cost are no longer obstacles; extraction quality now depends on model capability.\nFor example, I extracted:\nKey character profiles, personality dossiers, family genealogies;\nPer-chapter snapshots of each character\u0026rsquo;s economic, social, emotional, health, and interpersonal states across all 120 chapters;\nA basic spatial map of the Jia (賈) estate with spatial metadata;\nAll dialogue records, poetry corpora\u0026hellip;\nThe approach was to start with broad, not-yet-rigorous extraction that at least achieves high coverage \u0026ndash; ensuring every piece of text is classified into some category.\nGoverning Laws\r#\rI divide the governing laws into two types by my own judgment: fundamental world rules and the author\u0026rsquo;s artistic will.\nThis is admittedly arbitrary, but without making some such judgment the work cannot proceed at all.\nWorld rules include but are not limited to:\nSociety: class hierarchy, power dynamics, master-servant relationships, marriage;\nEconomy: income and expenditure, debt, risk of property confiscation;\nCulture: Confucian propriety, festivals, feudal values;\nPsychology: character emotions, personality-driven behavior, internal conflict;\nPolitics: imperial favor, court dynamics, external forces\u0026hellip;\nThe artistic will is precisely what makes Dream of the Red Chamber \u0026ndash; apart from the fact that it lacks a definitive ending \u0026ndash; an ideal prediction target.\nCao Xueqin embedded hints about the characters\u0026rsquo; fates throughout the novel, from the very beginning.\nThe most iconic example is the 判詞 (prophetic verses / judgment poems) of the 十二金釵 (Twelve Beauties of Jinling), which explicitly foreshadow the fates of the female lead and deuteragonist:\n可嘆停機德，堪憐詠絮才。玉帶林中掛，金簪雪裡埋。\n(How lamentable, her virtue of halting the loom; how pitiable, her talent of chanting the willow catkins. A jade belt hangs in the forest; a golden hairpin lies buried in the snow.)\nRule Engine\r#\rGiven the initial conditions and governing laws, how do we apply them?\nThe more ideal approach would be to build a 3D physics engine similar to a game engine, where each character possesses only the information they would know, and let an AI chatbot role-play each character like an actor performing a part.\nBut first, the cost would be too high and would only increase spectacle \u0026ndash; we would not be introducing new information, and the 3D engine would not produce new results.\nSecond, we are not running a wind-tunnel fluid dynamics simulation; we are trying to guess what Cao Xueqin had in mind. Staying at the textual level is sufficient for now.\nBased on the previously extracted data, we derive a set of computational subjects and rules.\nIn practice, this is the traditional process of evaluating evidence, confidence, and additive/subtractive adjustments for whether an event occurs \u0026ndash;\nmade systematic, repeatable, modifiable, and exhaustively brute-forced.\nThe simulation steps for each round are:\nProcess delayed effects \u0026ndash; check pending_effects; apply any that have reached their due chapter.\nEvaluate all laws \u0026ndash; check each law\u0026rsquo;s premises to see if all are satisfied (skip those with confidence \u0026lt; 0.3).\nConflict resolution \u0026ndash; simultaneously triggered laws may contradict each other; adjudicate which one wins.\nApply effects \u0026ndash; those with a delay go into the queue; those without directly modify state.\nSnapshot \u0026ndash; compress the current state into a numerical vector.\nchapter += 1\nA complete example \u0026ndash; the death of Lin Daiyu (林黛玉) in Chapter 98 \u0026ndash; is appended at the end of this article.\nWorkflow Summary\r#\rAmong the several components in the above workflow,\nwhether the extracted data is academically rigorous, whether the rules are reasonable and applicable, whether the simulation steps are sound \u0026ndash;\nnone of this is critically important, because each part can be improved and regenerated independently.\nFrom a software engineering perspective, my goal is to make this engine work well at the interface level,\nand continuously refine prediction results as more information is incorporated and the methodology improves.\nCurrent Results: Objective vs. Subjective Parallel Comparison\r#\rHere I must introduce another self-imposed methodology to enable structured comparison:\ndividing the inference engine\u0026rsquo;s layers into two main parts \u0026ndash; objective conditions and artistic choice.\nObjective Conditions\r#\rThe historical backdrop of the era in which the novel was written \u0026ndash; its characters, settings, feudal system, economy, and so on \u0026ndash; constitutes the first layer of objective conditions. This can delimit the entire scope of what the story is capable of containing. We have already extracted some objective laws based on period-appropriate historical context and academic literature.\nConversely, anything that actually existed in that era could, in theory, appear and influence the story.\nFor instance, the novel already features some Western modern objects such as self-striking clocks and pocket watches. What if Western firearms appeared and became a significant plot driver?\nExhaustively exploring such first-layer objective possibilities is a direction for future work, and might achieve an effect that is \u0026ldquo;within reason yet beyond expectation.\u0026rdquo;\nArtistic Choice\r#\rThe second layer is the author Cao Xueqin\u0026rsquo;s (曹雪芹) cultivation of this fictional world.\nMany characters and the overall trajectory of the family carry a heavy fatalistic coloring.\nThe novel\u0026rsquo;s countless poems and metaphors \u0026ndash; as well as marginal annotations by a friend who reportedly read the ending \u0026ndash; hint at this.\nTherefore, we can use the author\u0026rsquo;s background and life experiences\nto infer what fates he chose for his characters,\nand thereby reveal the values he truly wished to express.\nCross-Comparison\r#\rFrom here, we can treat the Gao E (高鶚) continuation as the work of the most advanced \u0026ldquo;player\u0026rdquo; to date.\nWhat he did is essentially the same thing I am doing now:\nbased on the characters and setting in the novel, attempting to divine Cao Xueqin\u0026rsquo;s artistic choices as closely as possible.\nMoreover, Gao E completed the existing ending, which greatly increased the novel\u0026rsquo;s circulation, and his version has been widely accepted \u0026ndash; so we place his version in a parallel position for comparison.\nRealistic Simulation\r#\rWhat if we stripped away all artistic treatment and retained only objective laws, letting the story evolve naturally?\nThe result is that most plot events would not occur within the span of 120 chapters. The narrative would be less dramatic and contain fewer tragedies.\nMethods for Improving Prediction Quality\r#\rRe-extract text when LLM capabilities improve\nMore human intervention for fine-tuning and experimenting with different prompts\nEnlist scholars of Redology (紅學, the academic study of Dream of the Red Chamber) or historians to assist with data cleaning and engine logic adjustments\nIncorporate newly discovered or previously undigitized materials (if any) into training\nExperiment with alternative methodologies\nEstablish a fixed workflow and let AI agents continuously fine-tune and produce many versions; since there is no clear termination criterion, quality can only be judged manually\nConclusion\r#\rDue to the constraints of existing and pre-trained data, and the strong internal consistency of Dream of the Red Chamber as a work of art,\ndeus ex machina predictions are unlikely to emerge. What we see instead are internal comparative differences \u0026ndash;\nfor example, the confiscation and decline of the Jia family is fated to happen regardless; the difference lies only in timing.\nA Final Reflection\r#\rThis kind of work would originally have required at least one to two years and at least one full-time person to complete.\nNow I can use my after-work hours to play a different professional role \u0026ndash; which also satisfies a regret from when financial pressure forced me to switch fields years ago.\nI hope that sharing the thinking process behind building the Dream of the Red Chamber Simulator is helpful to you,\nand I look forward to the social sciences \u0026ndash; not just computer science and the natural sciences \u0026ndash; benefiting from the rapid advances in AI.\nAppendix: Full Simulation Process Example\r#\rChapters 97-98, \u0026ldquo;The Death of Lin Daiyu\u0026rdquo; (黛玉之死) \u0026ndash; a complete walk-through of all six steps (the following content was generated by AI):\nExample: Chapter 97 \u0026ndash; The Switcheroo Plot (掉包計) -\u0026gt; Burning Manuscripts and Severing Ties (焚稿斷情) -\u0026gt; Daiyu\u0026rsquo;s Death\nBackground State (entering Chapter 97)\nAfter more than a dozen chapters of cumulative decline, Lin Daiyu\u0026rsquo;s state is:\nagent.林黛玉: health=0.12, mood=0.08, isolation=0.72, tragedy_risk=0.95, alive=True\nagent.賈寶玉: monk_tendency=0.35, mood=0.20\neconomy: debt_ratio=0.65\npolitics: family_decides_marriage=True\nrelation.賈寶玉::林黛玉: marriage_probability=0.15\nrelation.賈寶玉::薛寶釵: marriage_probability=0.72\nWhy did Daiyu\u0026rsquo;s health drop from an initial 0.35 to 0.12? Because this law has been silently triggering every chapter:\n▎ PSY_E1_DAIYU_DECAY \u0026ldquo;Daiyu\u0026rsquo;s health slowly decays\u0026rdquo;\n▎ Premise: health \u0026gt; 0.0 AND isolation \u0026gt; 0.3 AND alive = True -\u0026gt; Effect: health sub 0.017\n▎ At -0.017 per chapter, over a dozen chapters this amounts to a lethal chronic drain.\n① Process Delayed Effects\nCheck the pending_effects queue. Suppose the following was triggered in Chapter 13:\n▎ FATE_010 \u0026ldquo;Qin Keqing\u0026rsquo;s deathbed dream: the peak foretells the fall\u0026rdquo; delay_chapters: 20\nIts effect, economy.spending_pressure add 0.1, already came due and was executed in Chapter 33. The queue is now empty. Skip.\n② Evaluate All 369 Laws\nThe engine scans each law in sequence. The key laws that trigger this chapter:\nLaw A \u0026ndash; VAR_MARRIAGE_SWAP \u0026ldquo;The Switcheroo: Secretly marrying Baochai instead\u0026rdquo; conf=0.95\nPremise check:\nagent.林黛玉.health \\\u0026lt; 0.15 -\u0026gt; 0.12 \\\u0026lt; 0.15 ✅ agent.林黛玉.alive == True -\u0026gt; True ✅ politics.family\\_decides\\_marriage -\u0026gt; True ✅ relation.寶玉::黛玉.marriage\\_probability \\\u0026lt; 0.5 -\u0026gt; 0.15 \\\u0026lt; 0.5 ✅ All passed -\u0026gt; 🔥 Triggered!\rLaw B \u0026ndash; PSY_E1_DAIYU_DECAY \u0026ldquo;Daiyu\u0026rsquo;s health decay\u0026rdquo; conf=0.9\nhealth \u0026gt; 0.0 -\u0026gt; 0.12 \u0026gt; 0 ✅ isolation \u0026gt; 0.3 -\u0026gt; 0.72 \u0026gt; 0.3 ✅ alive == True ✅ -\u0026gt; 🔥 Triggered!\rLaw C \u0026ndash; VAR_MARRIAGE_DAIYU \u0026ldquo;The Stone-and-Wood Bond: Baoyu and Daiyu marry\u0026rdquo; conf=0.9\nrelation.寶玉::黛玉.marriage\\_probability \u0026gt; 0.7 -\u0026gt; 0.15 \u0026gt; 0.7 ❌ -\u0026gt; Not triggered (Baoyu-Daiyu marriage probability too low)\rThis chapter also simultaneously triggers over a dozen other laws (economic decline, political risk, etc.), but the above are the ones directly relevant to Daiyu.\n③ Conflict Resolution\nVAR_MARRIAGE_SWAP, VAR_MARRIAGE_NORMAL_BAOCHAI, and VAR_MARRIAGE_DAIYU belong to the same variant_group (marriage outcomes are mutually exclusive).\nOnly VAR_MARRIAGE_SWAP passed the premise check, so there is no actual conflict. However, if Daiyu were already dead (alive=False), then VAR_MARRIAGE_NORMAL_BAOCHAI would trigger instead of the switcheroo version \u0026ndash;\nthat would be a different evolutionary path.\nPSY_E1_DAIYU_DECAY\u0026rsquo;s effect is additive (sub), so it does not conflict with other laws. All effects are retained.\n④ Apply Effects\nLaw A\u0026rsquo;s effects execute immediately (delay=0):\nmarriage trigger_event BAOYU_MARRIED_BAOCHAI -\u0026gt; fate_flags[\u0026ldquo;BAOYU_MARRIED_BAOCHAI\u0026rdquo;] = True\nrelation.寶玉::寶釵.marriage_probability set 1.0 -\u0026gt; 1.0\nagent.賈寶玉.mood sub 0.5 -\u0026gt; 0.20 -\u0026gt; 0.00 (clamp)\nagent.賈寶玉.monk_tendency add 0.3 -\u0026gt; 0.35 -\u0026gt; 0.65\nagent.林黛玉.health sub 0.1 -\u0026gt; 0.12 -\u0026gt; 0.02\nLaw B\u0026rsquo;s effects:\nagent.林黛玉.health sub 0.017 -\u0026gt; 0.02 -\u0026gt; 0.003\nAt this point Daiyu\u0026rsquo;s health = 0.003, approaching zero.\n⑤ Snapshot\nCompress the current world state into a numerical vector:\nsnapshot = {\neconomy\\_vector: \\[0.42, 0.82, 0.65, 0.55, 0.80, 0.35], agent\\_vectors: { \u0026#34;林黛玉\u0026#34;: \\[0.003, 0.08, 0.10, 0.00, 0.30, 0.00, 0.72, 0.95], \u0026#34;賈寶玉\u0026#34;: \\[0.80, 0.00, 0.30, 0.72, 0.80, 0.65, 0.42, 0.92], ... }, politics\\_vector: \\[0.0, 0.60, 0.75]\r}\nThis vector will later be compared via Euclidean distance against the actual vector for Chapter 97 in actual_checkpoints.json.\n⑥ chapter = 98\nEnter the next chapter. At this point Daiyu\u0026rsquo;s health = 0.003, and BAOYU_MARRIED_BAOCHAI = True.\nWhen Chapter 98 runs step ② again, two lethal laws trigger simultaneously:\n▎ VAR_DAIYU_HEARTBREAK \u0026ldquo;Burning manuscripts, severing ties: Daiyu dies of heartbreak\u0026rdquo; conf=0.95\n▎ health ≤ 0.05 -\u0026gt; 0.003 ≤ 0.05 ✅\n▎ BAOYU_MARRIED_BAOCHAI -\u0026gt; True ✅\n▎ -\u0026gt; death trigger_event FATE_DAIYU_DEATH\n▎ -\u0026gt; monk_tendency add 0.4 -\u0026gt; Baoyu 0.65 -\u0026gt; 1.0 (clamp)\n▎ -\u0026gt; alive set False\nThen SYS_E19_ZERO_DAIYU triggers (checkpoint.FATE_DAIYU_DEATH = True), zeroing out all of Daiyu\u0026rsquo;s attributes.\nA few chapters later, Baoyu\u0026rsquo;s monk_tendency has reached 1.0 and mood ≤ 0.15, triggering VAR_MONK_DESPAIR \u0026ldquo;All hopes extinguished: Baoyu renounces the world\u0026rdquo; (萬念俱灰：寶玉出家).\n","date":"29 March 2026","externalUrl":null,"permalink":"/posts/stonestory_thermodynamics/","section":"Blog","summary":"","title":"Dream of the Red Chamber Simulator: Thermodynamics and Artistic Choice","type":"posts"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/vi/tags/gi%E1%BA%A5c-m%C6%A1-v%E1%BB%81-c%C4%83n-ph%C3%B2ng-%C4%91%E1%BB%8F/","section":"Tags","summary":"","title":"Giấc Mơ Về Căn Phòng Đỏ","type":"tags"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/vi/categories/h%E1%BB%99i-th%E1%BA%A3o/","section":"Categories","summary":"","title":"Hội Thảo","type":"categories"},{"content":"","date":"29 janvier 2026","externalUrl":null,"permalink":"/fr/categories/latelier/","section":"Categories","summary":"","title":"L'Atelier","type":"categories"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/tags/literary-simulation/","section":"Tags","summary":"","title":"Literary Simulation","type":"tags"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/tags/llm/","section":"Tags","summary":"","title":"LLM","type":"tags"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/id/categories/lokakarya/","section":"Categories","summary":"","title":"Lokakarya","type":"categories"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/id/tags/mesin-aturan/","section":"Tags","summary":"","title":"Mesin Aturan","type":"tags"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/id/tags/mimpi-kamar-merah/","section":"Tags","summary":"","title":"Mimpi Kamar Merah","type":"tags"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/vi/tags/m%C3%B4-ph%E1%BB%8Fng-v%C4%83n-h%E1%BB%8Dc/","section":"Tags","summary":"","title":"Mô Phỏng Văn Học","type":"tags"},{"content":"","date":"29 janvier 2026","externalUrl":null,"permalink":"/fr/tags/moteur-de-r%C3%A8gles/","section":"Tags","summary":"","title":"Moteur De Règles","type":"tags"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/vi/tags/nhi%E1%BB%87t-%C4%91%E1%BB%99ng-l%E1%BB%B1c-h%E1%BB%8Dc/","section":"Tags","summary":"","title":"Nhiệt Động Lực Học","type":"tags"},{"content":"","date":"29 janvier 2026","externalUrl":null,"permalink":"/fr/tags/r%C3%AAve-de-la-chambre-rouge/","section":"Tags","summary":"","title":"Rêve De La Chambre Rouge","type":"tags"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/tags/rule-engine/","section":"Tags","summary":"","title":"Rule Engine","type":"tags"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/id/tags/simulasi-sastra/","section":"Tags","summary":"","title":"Simulasi Sastra","type":"tags"},{"content":"","date":"29 janvier 2026","externalUrl":null,"permalink":"/fr/tags/simulation-litt%C3%A9raire/","section":"Tags","summary":"","title":"Simulation Littéraire","type":"tags"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/id/tags/termodinamika/","section":"Tags","summary":"","title":"Termodinamika","type":"tags"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/categories/the-workshop/","section":"Categories","summary":"","title":"The Workshop","type":"categories"},{"content":"","date":"29 March 2026","externalUrl":null,"permalink":"/tags/thermodynamics/","section":"Tags","summary":"","title":"Thermodynamics","type":"tags"},{"content":"","date":"29 janvier 2026","externalUrl":null,"permalink":"/fr/tags/thermodynamique/","section":"Tags","summary":"","title":"Thermodynamique","type":"tags"},{"content":"","date":"2026년 3월 29일","externalUrl":null,"permalink":"/ko/tags/%EA%B7%9C%EC%B9%99-%EC%97%94%EC%A7%84/","section":"Tags","summary":"","title":"규칙 엔진","type":"tags"},{"content":"","date":"2026년 3월 29일","externalUrl":null,"permalink":"/ko/tags/%EB%AC%B8%ED%95%99-%EC%8B%9C%EB%AE%AC%EB%A0%88%EC%9D%B4%EC%85%98/","section":"Tags","summary":"","title":"문학 시뮬레이션","type":"tags"},{"content":"","date":"2026년 3월 29일","externalUrl":null,"permalink":"/ko/tags/%EB%B2%95%ED%95%99%EB%8C%80%ED%95%99%EC%9B%90/","section":"Tags","summary":"","title":"법학대학원","type":"tags"},{"content":"","date":"2026년 3월 29일","externalUrl":null,"permalink":"/ko/tags/%EB%B6%89%EC%9D%80-%EB%B0%A9%EC%9D%98-%EA%BF%88/","section":"Tags","summary":"","title":"붉은 방의 꿈","type":"tags"},{"content":"","date":"2026년 3월 29일","externalUrl":null,"permalink":"/ko/tags/%EC%97%B4%EC%97%AD%ED%95%99/","section":"Tags","summary":"","title":"열역학","type":"tags"},{"content":"","date":"2026년 3월 29일","externalUrl":null,"permalink":"/ko/categories/%EC%9B%8C%ED%81%AC%EC%88%8D/","section":"Categories","summary":"","title":"워크숍","type":"categories"},{"content":"","date":"2026年3月29日","externalUrl":null,"permalink":"/ja/tags/%E3%83%AB%E3%83%BC%E3%83%AB%E3%82%A8%E3%83%B3%E3%82%B8%E3%83%B3/","section":"Tags","summary":"","title":"ルールエンジン","type":"tags"},{"content":"","date":"2026年3月29日","externalUrl":null,"permalink":"/zh-hans/tags/%E7%83%AD%E5%8A%9B%E5%AD%A6/","section":"Tags","summary":"","title":"热力学","type":"tags"},{"content":"","date":"2026年3月29日","externalUrl":null,"permalink":"/ja/tags/%E7%86%B1%E5%8A%9B%E5%AD%A6/","section":"Tags","summary":"","title":"熱力学","type":"tags"},{"content":"","date":"2026年3月29日","externalUrl":null,"permalink":"/zh-tw/tags/%E7%86%B1%E5%8A%9B%E5%AD%B8/","section":"Tags","summary":"","title":"熱力學","type":"tags"},{"content":"","date":"2026年3月29日","externalUrl":null,"permalink":"/ja/tags/%E7%B4%85%E6%A5%BC%E5%A4%A2/","section":"Tags","summary":"","title":"紅楼夢","type":"tags"},{"content":"","date":"2026年3月29日","externalUrl":null,"permalink":"/zh-tw/tags/%E7%B4%85%E6%A8%93%E5%A4%A2/","section":"Tags","summary":"","title":"紅樓夢","type":"tags"},{"content":"","date":"2026年3月29日","externalUrl":null,"permalink":"/zh-hans/tags/%E7%BA%A2%E6%A5%BC%E6%A2%A6/","section":"Tags","summary":"","title":"红楼梦","type":"tags"},{"content":"","date":"2026年3月29日","externalUrl":null,"permalink":"/zh-tw/tags/%E8%A6%8F%E5%89%87%E5%BC%95%E6%93%8E/","section":"Tags","summary":"","title":"規則引擎","type":"tags"},{"content":"","date":"2026年3月29日","externalUrl":null,"permalink":"/zh-hans/tags/%E8%A7%84%E5%88%99%E5%BC%95%E6%93%8E/","section":"Tags","summary":"","title":"规则引擎","type":"tags"},{"content":"這是「2026 台股前十大下跌日」系列第 10 名。排名口徑是 2026/01/01 至 2026/07/29 的加權指數單日收盤報酬率，不是盤中最大跌幅。\n開低、破底、回收一小段；跌幅不如前幾名巨大，卻很適合觀察人如何把一次經驗直接複製到明天。\n這天發生了什麼\r#\r數值 前收 33,543.88 開盤 33,334.00 最高 33,334.00 最低 32,461.09 收盤 32,722.50 漲跌 ▼ 821.38（-2.45%） 2026 跌幅排名 第 10 名 BBS 全量留言\r#\rPTT 網頁版遇到推爆長文會省略中段；這裡走 BBS 層抓到 7,141 則，不是網頁殘片。\n推文型態 數量 推 4,199 噓 540 → 2,402 合計 7,141 留言最密集的時段是 09:00（2,053 則）。\n時段 留言數 相對量 08:00 1,109 █████████ 09:00 2,053 ████████████████ 10:00 1,199 █████████ 11:00 1,064 ████████ 12:00 838 ███████ 13:00 856 ███████ 14:00 18 █ PTT 原文：AID 1fm8cBpD 匿名化原始檔：純文字 · JSONL 8 種謬誤與偏誤\r#\r以下是語句模式的分類，不是對留言者作人格診斷。盤中閒聊包含反串、迷因與情緒宣洩；引文只能證明這句話出現過，不能證明作者真的照著交易。\n1. 單因謬誤：把市場縮成一個黑手\r#\r市場同時包含外資、內資、避險、被動資金與個別公司消息；把整段價格路徑只歸因於一個有意志的角色，故事會很順，證據卻通常不夠。\n#2241 · 03/23 09:25 · user7033：黑手撐完了 量不見 換小兒表演了\n#4782 · 03/23 11:20 · user7034：股票是現在政府唯一說嘴的 跌不下去啦 大家別怕 政\n2. 賭徒謬誤：跌多了就「該」彈\r#\r先前已經跌多少，不會自動提高下一分鐘上漲的機率。價格可以很便宜，也可以在缺乏新資訊時繼續變便宜。\n#1635 · 03/23 09:10 · user7035：特價5分鐘而已 現在誰在啦?\n#4507 · 03/23 11:06 · user7036：當年700gg特價350\n3. 錨定效應：月線、季線與整數都是答案\r#\r技術位置可以是風險管理參考，但把單一價位當成必然反轉點，會忽略波動、成交量與事件條件。\n#2318 · 03/23 09:29 · user7037：可以殺到季線來聞香一下嗎\n#5262 · 03/23 11:47 · user7038：力積月線就彈起來\n4. 直線外推：今天的斜率一路畫到明天\r#\r人腦很容易把眼前最強烈的方向延伸出去；市場真正困難之處，正是斜率會在新資訊出現時改變。\n#3656 · 03/23 10:25 · user7039：少年股神明天跑外送\n#6405 · 03/23 13:08 · user7040：台股就沒有在一次跌完的拉 要買更低的 明天請早\n5. 從眾效應：多蛙、空蛙與「大家都知道」\r#\r群體標籤能快速製造安全感，也會把不同持倉、期限與風險承受度的人壓成同一種對手。\n#2052 · 03/23 09:19 · user7041：應該沒人早上掛市價單砍在阿呆股的吧?\n#4981 · 03/23 11:28 · user7042：中午賣壓即將跌2000 大家抓緊了\n6. 後見之明偏誤：收盤後每個人都早知道\r#\r結果出現後，原先的多種可能性會從記憶裡消失。真正可檢驗的不是「早說」，而是事前是否留下方向、期限與失效條件。\n#2395 · 03/23 09:31 · user7043：幹我就知道是人道走廊\n#5615 · 03/23 12:09 · user7044：我們還怕套嗎？ 早就麻了\n7. 類比謬誤：把日韓、美股或夜盤直接翻譯\r#\r跨市場確實相關，但交易時段、權重、匯率與事件曝險不同；相關不等於可以一比一複製漲跌。\n#2076 · 03/23 09:20 · user7045：大媽畢竟不看夜盤 也不看日韓＊．＊\n#5614 · 03/23 12:09 · user7046：看看日韓 台股世界強\n8. 災難化與全有全無：不是 V，就是歸零\r#\r這比較接近認知扭曲而非形式邏輯謬誤：把連續的風險壓成生或死兩個選項，會讓部位管理退化成情緒口號。\n#1255 · 03/23 09:02 · user7031：救命啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊\n#4335 · 03/23 10:58 · user7047：救命啊！！！每一秒錢都在蒸發！！\n這份考古不能證明什麼\r#\r它不是情緒指標回測，也沒有證明某類留言能預測下一根 K 線。 關鍵字分類是可重現的抽樣入口，不是對全部 7,141 則留言做唯一正解標註。 同一句話可能同時包含多種偏誤；為了可讀性，每則引文只放在一個小節。 帳號已置換為合成代號（user0001 之類），與原帳號無對應關係、不可回推；保留樓層與時間，是為了讓讀者能回到匿名化原始檔核對。 2026 年截至 07/29 的前十大下跌日\r#\r名次 日期 收盤跌幅 跌點 文章 1 2026-07-17 -6.47% -2,953.71 「跌到收盤才停」 2 2026-07-28 -4.65% -2,030.83 「這次一定會 V？」 3 2026-03-09 -4.43% -1,489.12 「千點低開之後」 4 2026-03-04 -4.35% -1,494.77 「月線會救嗎」 5 2026-07-29 -3.76% -1,564.18 「四萬點保衛戰」 6 2026-06-26 -3.64% -1,683.50 「指數在跌，嘴在做什麼」 7 2026-06-08 -3.48% -1,568.16 「夜盤只是誤會？」 8 2026-06-10 -3.31% -1,478.90 「反彈第二天又殺」 9 2026-07-24 -2.67% -1,195.97 「日韓跌，台股就會？」 10 2026-03-23 -2.45% -821.38 「明天會噴回來？」（本篇） 資料口徑\r#\r指數排名與 OHLC：Yahoo Finance ^TWII 日線快照，抓取日 2026/07/29；排名用相鄰交易日收盤價計算。臺灣證券交易所亦提供發行量加權股價指數歷史資料供核對。 留言：PTT Stock 板當日「盤中閒聊」，以 BBS 層完整抓取；本篇產生前已核對 AID、標題與留言總數。 隱私：公開原始檔與本文引文的帳號一律置換為合成代號，不保留原字元、不可回推。同一個代號在所有檔案裡指同一人，跨日比對仍然成立。 ","date":"2026年3月23日","externalUrl":null,"permalink":"/zh-tw/stocks/taiex-chat-0323/","section":"股票","summary":"","title":"「明天會噴回來？」：7,141 則盤中留言的謬誤考古（2026/03/23）","type":"stocks"},{"content":"","date":"22 March 2026","externalUrl":null,"permalink":"/id/tags/ayat-nubuatan/","section":"Tags","summary":"","title":"Ayat Nubuatan","type":"tags"},{"content":"","date":"22 March 2026","externalUrl":null,"permalink":"/vi/tags/b%E1%BA%A3n-th%E1%BB%83-h%E1%BB%8Dc/","section":"Tags","summary":"","title":"Bản Thể Học","type":"tags"},{"content":"","date":"22 March 2026","externalUrl":null,"permalink":"/vi/tags/c%C3%A2u-th%C6%A1-ti%C3%AAn-tri/","section":"Tags","summary":"","title":"Câu Thơ Tiên Tri","type":"tags"},{"content":" Preface\r#\rPredicting the future — from fortune and misfortune to the fate of humanity — has been one of the grand challenges of human civilization since antiquity. Large Language Models (LLMs) now offer us a glimpse of hope for tackling this problem.\nThis article explores the use of LLMs as the latest tool, with Dream of the Red Chamber (紅樓夢) serving as a sandbox, to find methods for predicting the novel\u0026rsquo;s lost final forty chapters.\nLet me state upfront: I have not succeeded. Perhaps one day when someone does, this article will surface in their search results.\nThis piece is more of a meditation on the nature of text itself. While text lacks the precision of physical formulas,\nas a tool for humanity to grasp reality and speculate about the future, it is far more important than we imagine.\nText is not merely an \u0026ldquo;imagined\u0026rdquo; reality — it is not inherently subjective. It simply mirrors objective reality in the most cost-effective way possible.\nAnd LLMs, as automated mechanisms for predicting text, will radically compress the cost of extracting, generating, and mirroring objective reality.\nThe latest implementation will be updated in the iOS app: Dream of the Red Chamber Simulator.\nAPP: Link\nThe three great regrets in life: First, that the shad has too many bones; second, that the crabapple blossom has no fragrance; third, that Dream of the Red Chamber was never finished.\n— Eileen Chang (張愛玲)\nCelestial Patterns: More Than Just Word Prediction\r#\rPredicting the future has always been a matter of great importance in human societies. Every ancient civilization had priests or officials dedicated to observing the stars.\nSymbol systems such as astronomy and hydrology textualized natural phenomena and physical laws. The most quintessential example is the coordinate system of latitude and longitude — text became a crucial tool for humanity to understand and influence the objective world.\nThe practical power of this mapping between text and reality has been validated in these recent years of explosive LLM capability.\nIn the past, language as a tool was not sufficiently deterministic. After the Industrial Revolution, when science became the primary driver of productivity, language was perpetually relegated to the bottom of the prestige hierarchy.\nThe age of LLMs has finally brought the digestion and production of text into the millisecond domain, freeing it from the bottlenecks of human reading speed, typing speed, and typographical errors.\nWork that once consumed enormous mental energy and time now has the potential to be assembled and configured like a production line.\nBut what does this production line produce? The essence of an LLM is \u0026ldquo;predicting\u0026rdquo; the next token. Is this actually productive? Does the model \u0026ldquo;sort of\u0026rdquo; \u0026ldquo;understand\u0026rdquo; what it is saying?\nIlya Sutskever (former co-founder and chief scientist of OpenAI) once gave this example:\nSay you read a detective novel, and on the last page, the detective says \u0026ldquo;I am going to reveal the identity of the criminal, and that person\u0026rsquo;s name is\u0026hellip;\u0026rdquo;\nIf an LLM can consistently and correctly guess the identity of the culprit, then we can tentatively say it \u0026ldquo;understands\u0026rdquo; the novel — at least surpassing the many readers who guessed wrong.\nAnd we must properly appreciate what \u0026ldquo;understanding\u0026rdquo; means. Understanding is ultimately for predicting the future. Every ancient civilization, without exception, studied astronomy and hydrology\nprecisely to forecast upcoming climate patterns, changes in river courses, droughts and floods — to survive better in the objective environment.\nOne could even argue that predicting correctly matters more than understanding.\nThe Humanities: Both People and Agents Remain Black Boxes\r#\rPredicting the future is the pursuit and prerequisite (reproducibility) of the natural sciences, and the holy grail of the social sciences.\nThis admittedly sounds like science fiction. In Isaac Asimov\u0026rsquo;s Foundation series, such a discipline for predicting the future was fictionalized as \u0026ldquo;psychohistory\u0026rdquo; (心理史學).\nEconomists, historians, psychologists, social scientists — all want to know how individuals and societies will react to specific events.\nFinance, in particular, is probably the field outside of software where AI is being applied most aggressively.\nAlthough we cannot yet see the finish line, the feasibility of this endeavor has improved significantly.\nThe improvement — and its limitation — is that we now have a remarkable black box (the LLM agent).\nFor tasks at a level comparable to human performance, it is blazingly fast and extremely cheap, making it suitable for replacing human labor.\nThe limitation is that its current mode of use resembles a slot machine. We can use certain techniques (prompt/context engineering) to improve the hit rate, but that is about it.\nWe struggle to open the black box. Chaining multiple black boxes together (multi-agent) yields only limited improvement.\nCurrently, tasks that a single agent can handle are done quickly and well, but more abstract tasks are difficult to improve linearly.\nApplied to social science: a single agent cannot adequately simulate even one individual\u0026rsquo;s memory and emotions, let alone having multi-agent systems simulate an entire community.\nOn the optimistic side, this feels more like a performance problem — and performance within this paradigm will continue to improve.\nThe Sandbox: Don\u0026rsquo;t Aim for a One-Hit Kill\r#\rSince we are dealing with a black box, the intuitive approach is to find a smaller box to attempt to crack.\nAssume the current baseline model capability is what was described earlier: throw any detective novel into the LLM slot machine, and it can directly (one-shot) and correctly output who the culprit is.\nBuilding on this baseline, if we put in extra effort — erecting scaffolding, going back and forth with the LLM in discussion, finding ways to linearly accumulate results across each exchange — we should theoretically be able to make predictions of higher difficulty.\nDream of the Red Chamber is the perfect target. Based on the content of the first eighty chapters, we ask the model to predict, to some degree, the final forty chapters.\nThis prediction is extremely difficult, but it is just right for my working objectives. Theoretically the probability is not zero; practically it is highly unlikely. This makes it an ideal benchmark for observing LLM capability growth over the coming years.\nHaving written this far, I can finally articulate two working objectives:\nHow can we put in additional effort so that answers unattainable through one-shot prompting can be progressively approached? How should we choose our battleground so that our results are not immediately rendered obsolete by stronger models — and ideally, so that our framework also benefits when future models improve? Below, I begin considering research methods based on the characteristics of Dream of the Red Chamber and LLMs.\nAssumptions\r#\rWe assume that the ending of Dream of the Red Chamber did once exist, and that the first eighty chapters and the subsequent conclusion were written as an organic, intentional, continuous work — exhibiting the same internal coherence found within the first eighty chapters themselves.\nIf the ending never actually existed, the prediction difficulty is even higher — approaching the prediction of a parallel universe. The question becomes: if Cao Xueqin had written the ending, what would it necessarily have been?\nThis word \u0026ldquo;necessarily\u0026rdquo; is the crux. One must reach this level of confidence for generating something from nothing to be meaningful.\nThe Writing of Dream of the Red Chamber\r#\rThe novel was composed around the 1750s. At that time it circulated mostly among friends and relatives. It was not until 1791, when Cheng Weiyuan published it using movable wooden type, that it became widely known.\nRedology and AI-Assisted Research\r#\rWang Guowei and Hu Shi were pioneers of Redology (紅學 — the scholarly study of Dream of the Red Chamber). The field has continued to develop, and in recent years has trended toward popularization and entertainment. The attention given to textual archaeology (探佚學) and the controversial Guiyou manuscript (癸酉本) reflects the public\u0026rsquo;s curiosity about the ending.\nKey research achievements incorporating the latest technology include:\nMachine learning once again confirming that the final forty chapters were not written by the original author Using LLMs for more nuanced semantic vectorization of text (Word Embedding) Using LLMs to build domain-specific knowledge graphs Models trained specifically on the first eighty chapters and Qing dynasty historical texts as input data LLM Characteristics\r#\rThe LLM characteristic most relevant to this task is: it has been trained on all data available on the internet, plus all valuable materials these frontier AI labs could obtain.\nFor information already in its training data, the model\u0026rsquo;s predictive capability and tendency are very high. For instance, if you input a passage from Harry Potter, it can recite the subsequent paragraphs from memory.\nBut the final forty chapters of Dream of the Red Chamber were never transmitted to posterity. They are not in the model\u0026rsquo;s training data. It cannot recite them.\nProblem 1: Context Window Limitations\r#\rCan we simply input chapters one through eighty and ask the LLM to output the remaining forty?\nOn the input side, the current top-tier models (Gemini 3.1 / GPT-5.4 / Opus 4.6) using API mode can support up to 1M tokens, which is sufficient.\nHowever, under the current paradigm, the output token window is far smaller than the input. Output is limited to roughly four to eight thousand Chinese characters at most — approximately one chapter\u0026rsquo;s worth of content.\nProblem 2: Listless Prose and Quality Degradation\r#\rWhat if we modify the prompt to ask the LLM to output only the content of chapter eighty-one?\nThe model gets \u0026ldquo;contaminated\u0026rdquo; by the massive text input. Its writing style closely resembles Cao Xueqin\u0026rsquo;s, and it can reasonably continue the known plot — but the result reads like a flat chronicle of events.\nThen, repeating the process for chapters eighty-two, eighty-three, and so on, the quality drops precipitously.\nProblem 3: Prior Contamination in the Model\r#\rAnother issue is that during training, the model has already seen Gao E\u0026rsquo;s continuation (高鶚續書), various scholarly speculations, and other secondary sources. If these materials diverge from the original ending, the output will be biased.\nTo Be Continued\r#\rDue to the length of this piece, I will wrap up here with a preview of what comes next.\nWe cannot simply have the LLM directly produce unknown information.\nSo we still need more traditional, mechanical, or programmatic methods.\nThe good news is: for the tireless researchers of literature, history, and philosophy — we now have a tractor for the field!\nDream of the Red Chamber possesses a highly structured nature. Important characters have their own 判詞 (prophetic verses, known as \u0026ldquo;pànCí\u0026rdquo;) — poetic passages that cryptically foreshadow each character\u0026rsquo;s fate.\nMoreover, the first eighty chapters can be cross-validated against one another, making the novel more amenable to prediction than many other works of fiction.\nAlthough the cast of characters is large and their backgrounds complex, what we are ultimately predicting is Cao Xueqin\u0026rsquo;s artistic vision — his creative will permeates the entire work. This is a tremendous aid for predicting the ending.\nNext: The Thermodynamics of Dream of the Red Chamber\r#\rThe next article will introduce the experimental approach: structurally extracting content from the text, iteratively experimenting to extract the rules embedded in the novel, and using code to run repeated experiments.\nThe idealized scenario is something akin to a thermodynamic system: given initial conditions (premises — e.g., characters, family wealth, socioeconomic status, interpersonal networks\u0026hellip;) plus the system\u0026rsquo;s operating mechanisms (human psychology, social hierarchy, economic dynamics, cultural norms, karmic retribution, etc.), one could predict the system\u0026rsquo;s state at any subsequent point in time.\n","date":"22 March 2026","externalUrl":null,"permalink":"/posts/stonestory_fate/","section":"Blog","summary":"","title":"Dream of the Red Chamber Simulator: The Holy Grail of Social Science, and LLMs as Prophetic Verse","type":"posts"},{"content":"","date":"22 March 2026","externalUrl":null,"permalink":"/tags/eileen-chang/","section":"Tags","summary":"","title":"Eileen Chang","type":"tags"},{"content":"","date":"22 March 2026","externalUrl":null,"permalink":"/id/tags/ontologi/","section":"Tags","summary":"","title":"Ontologi","type":"tags"},{"content":"","date":"22 janvier 2026","externalUrl":null,"permalink":"/fr/tags/ontologie/","section":"Tags","summary":"","title":"Ontologie","type":"tags"},{"content":"","date":"22 March 2026","externalUrl":null,"permalink":"/tags/ontology/","section":"Tags","summary":"","title":"Ontology","type":"tags"},{"content":"","date":"22 March 2026","externalUrl":null,"permalink":"/tags/prophetic-verse/","section":"Tags","summary":"","title":"Prophetic Verse","type":"tags"},{"content":"","date":"22 janvier 2026","externalUrl":null,"permalink":"/fr/tags/verset-proph%C3%A9tique/","section":"Tags","summary":"","title":"Verset Prophétique","type":"tags"},{"content":"","date":"2026년 3월 22일","externalUrl":null,"permalink":"/ko/tags/%EC%97%90%EC%9D%BC%EB%A6%B0-%EC%9E%A5/","section":"Tags","summary":"","title":"에일린 장","type":"tags"},{"content":"","date":"2026년 3월 22일","externalUrl":null,"permalink":"/ko/tags/%EC%98%88%EC%96%B8-%EA%B5%AC%EC%A0%88/","section":"Tags","summary":"","title":"예언 구절","type":"tags"},{"content":"","date":"2026년 3월 22일","externalUrl":null,"permalink":"/ko/tags/%EC%A1%B4%EC%9E%AC%EB%A1%A0/","section":"Tags","summary":"","title":"존재론","type":"tags"},{"content":"","date":"2026年3月22日","externalUrl":null,"permalink":"/ja/tags/%E5%AD%98%E5%9C%A8%E8%AB%96/","section":"Tags","summary":"","title":"存在論","type":"tags"},{"content":"","date":"2026年3月22日","externalUrl":null,"permalink":"/zh-hans/tags/%E5%BC%A0%E7%88%B1%E7%8E%B2/","section":"Tags","summary":"","title":"张爱玲","type":"tags"},{"content":"","date":"2026年3月22日","externalUrl":null,"permalink":"/zh-tw/tags/%E5%BC%B5%E6%84%9B%E7%8E%B2/","section":"Tags","summary":"","title":"張愛玲","type":"tags"},{"content":"","date":"2026年3月22日","externalUrl":null,"permalink":"/zh-hans/tags/%E6%9C%AC%E4%BD%93%E8%AE%BA/","section":"Tags","summary":"","title":"本体论","type":"tags"},{"content":"","date":"2026年3月22日","externalUrl":null,"permalink":"/zh-tw/tags/%E6%9C%AC%E9%AB%94%E8%AB%96/","section":"Tags","summary":"","title":"本體論","type":"tags"},{"content":"","date":"2026年3月22日","externalUrl":null,"permalink":"/zh-tw/tags/%E8%A9%A9%E8%AE%96/","section":"Tags","summary":"","title":"詩讖","type":"tags"},{"content":"","date":"2026年3月22日","externalUrl":null,"permalink":"/zh-hans/tags/%E8%AF%97%E8%B0%B6/","section":"Tags","summary":"","title":"诗谶","type":"tags"},{"content":"這是「2026 台股前十大下跌日」系列第 3 名。排名口徑是 2026/01/01 至 2026/07/29 的加權指數單日收盤報酬率，不是盤中最大跌幅。\n開盤先吞下巨大缺口，盤中再破底，尾盤才收回一小段。恐慌、抄底與「夜盤是假的」同時出現。\n這天發生了什麼\r#\r數值 前收 33,599.54 開盤 32,354.61 最高 32,354.61 最低 31,529.36 收盤 32,110.42 漲跌 ▼ 1,489.12（-4.43%） 2026 跌幅排名 第 3 名 BBS 全量留言\r#\rPTT 網頁版遇到推爆長文會省略中段；這裡走 BBS 層抓到 9,644 則，不是網頁殘片。\n推文型態 數量 推 5,563 噓 804 → 3,277 合計 9,644 留言最密集的時段是 09:00（2,886 則）。\n時段 留言數 相對量 08:00 1,811 ██████████ 09:00 2,886 ████████████████ 10:00 1,590 █████████ 11:00 1,113 ██████ 12:00 1,046 ██████ 13:00 1,140 ██████ 14:00 57 █ PTT 原文：AID 1fhXIVJR 匿名化原始檔：純文字 · JSONL 8 種謬誤與偏誤\r#\r以下是語句模式的分類，不是對留言者作人格診斷。盤中閒聊包含反串、迷因與情緒宣洩；引文只能證明這句話出現過，不能證明作者真的照著交易。\n1. 單因謬誤：把市場縮成一個黑手\r#\r市場同時包含外資、內資、避險、被動資金與個別公司消息；把整段價格路徑只歸因於一個有意志的角色，故事會很順，證據卻通常不夠。\n#2963 · 03/09 09:15 · user7017：鈺創主力敢在崩7%大盤拉嗎？\n#7255 · 03/09 11:50 · user7018：債蛙軍團：小兒科，你先買\n2. 賭徒謬誤：跌多了就「該」彈\r#\r先前已經跌多少，不會自動提高下一分鐘上漲的機率。價格可以很便宜，也可以在缺乏新資訊時繼續變便宜。\n#2666 · 03/09 09:11 · user7019：反彈真的要逃 其他股是都在破底 你還以為要抄底喔？\n#6488 · 03/09 11:08 · user7020：本債蛙抄底現在變身多頭大將軍\n3. 錨定效應：月線、季線與整數都是答案\r#\r技術位置可以是風險管理參考，但把單一價位當成必然反轉點，會忽略波動、成交量與事件條件。\n#2325 · 03/09 09:06 · user7021：31000肛底!y\n#6713 · 03/09 11:17 · user7022：拉超快還沒上車欸 欸欸欸季線勒\n4. 直線外推：今天的斜率一路畫到明天\r#\r人腦很容易把眼前最強烈的方向延伸出去；市場真正困難之處，正是斜率會在新資訊出現時改變。\n#3943 · 03/09 09:37 · user7023：今天買 明天賣 蒿吐露絲\n#8237 · 03/09 12:48 · user7024：說真的這一波一定會回12682 要買股票到時候再買\n5. 從眾效應：多蛙、空蛙與「大家都知道」\r#\r群體標籤能快速製造安全感，也會把不同持倉、期限與風險承受度的人壓成同一種對手。\n#3139 · 03/09 09:18 · user7025：沒人發現群創快v回平盤了XDDD\n#7570 · 03/09 12:08 · user7026：現在公司的廁所是不是都滿的 我需要個訊號 多蛙們\n6. 後見之明偏誤：收盤後每個人都早知道\r#\r結果出現後，原先的多種可能性會從記憶裡消失。真正可檢驗的不是「早說」，而是事前是否留下方向、期限與失效條件。\n#1834 · 03/09 09:00 · user7027：早說會死 記憶卡娃死盧~系賀\n#7726 · 03/09 12:19 · user7028：美股早就不知道什麼叫月線了\u0026hellip;\n7. 類比謬誤：把日韓、美股或夜盤直接翻譯\r#\r跨市場確實相關，但交易時段、權重、匯率與事件曝險不同；相關不等於可以一比一複製漲跌。\n#4136 · 03/09 09:43 · user7029：但台日韓都抵制俄羅斯 根本沒來源了\n#8287 · 03/09 12:51 · user7030：其實美股盤前才跌兩趴而已 怎麼就恐慌成這樣XD\n8. 災難化與全有全無：不是 V，就是歸零\r#\r這比較接近認知扭曲而非形式邏輯謬誤：把連續的風險壓成生或死兩個選項，會讓部位管理退化成情緒口號。\n#1159 · 03/09 08:51 · user7031：救命啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊\n#4715 · 03/09 10:00 · user7032：2026賺的錢歸零了 幹\n這份考古不能證明什麼\r#\r它不是情緒指標回測，也沒有證明某類留言能預測下一根 K 線。 關鍵字分類是可重現的抽樣入口，不是對全部 9,644 則留言做唯一正解標註。 同一句話可能同時包含多種偏誤；為了可讀性，每則引文只放在一個小節。 帳號已置換為合成代號（user0001 之類），與原帳號無對應關係、不可回推；保留樓層與時間，是為了讓讀者能回到匿名化原始檔核對。 2026 年截至 07/29 的前十大下跌日\r#\r名次 日期 收盤跌幅 跌點 文章 1 2026-07-17 -6.47% -2,953.71 「跌到收盤才停」 2 2026-07-28 -4.65% -2,030.83 「這次一定會 V？」 3 2026-03-09 -4.43% -1,489.12 「千點低開之後」（本篇） 4 2026-03-04 -4.35% -1,494.77 「月線會救嗎」 5 2026-07-29 -3.76% -1,564.18 「四萬點保衛戰」 6 2026-06-26 -3.64% -1,683.50 「指數在跌，嘴在做什麼」 7 2026-06-08 -3.48% -1,568.16 「夜盤只是誤會？」 8 2026-06-10 -3.31% -1,478.90 「反彈第二天又殺」 9 2026-07-24 -2.67% -1,195.97 「日韓跌，台股就會？」 10 2026-03-23 -2.45% -821.38 「明天會噴回來？」 資料口徑\r#\r指數排名與 OHLC：Yahoo Finance ^TWII 日線快照，抓取日 2026/07/29；排名用相鄰交易日收盤價計算。臺灣證券交易所亦提供發行量加權股價指數歷史資料供核對。 留言：PTT Stock 板當日「盤中閒聊」，以 BBS 層完整抓取；本篇產生前已核對 AID、標題與留言總數。 隱私：公開原始檔與本文引文的帳號一律置換為合成代號，不保留原字元、不可回推。同一個代號在所有檔案裡指同一人，跨日比對仍然成立。 ","date":"2026年3月9日","externalUrl":null,"permalink":"/zh-tw/stocks/taiex-chat-0309/","section":"股票","summary":"","title":"「千點低開之後」：9,644 則盤中留言的謬誤考古（2026/03/09）","type":"stocks"},{"content":"這是「2026 台股前十大下跌日」系列第 4 名。排名口徑是 2026/01/01 至 2026/07/29 的加權指數單日收盤報酬率，不是盤中最大跌幅。\n開盤接近前收，之後一路殺到收盤最低。每跌破一條線，討論就換一個新錨點。\n這天發生了什麼\r#\r數值 前收 34,323.65 開盤 34,228.75 最高 34,228.75 最低 32,828.88 收盤 32,828.88 漲跌 ▼ 1,494.77（-4.35%） 2026 跌幅排名 第 4 名 BBS 全量留言\r#\rPTT 網頁版遇到推爆長文會省略中段；這裡走 BBS 層抓到 10,200 則，不是網頁殘片。\n推文型態 數量 推 5,970 噓 836 → 3,394 合計 10,200 留言最密集的時段是 09:00（3,061 則）。\n時段 留言數 相對量 08:00 794 ████ 09:00 3,061 ████████████████ 10:00 1,581 ████████ 11:00 1,656 █████████ 12:00 1,655 █████████ 13:00 1,436 ████████ 14:00 15 █ PTT 原文：AID 1fftqCQW 匿名化原始檔：純文字 · JSONL 8 種謬誤與偏誤\r#\r以下是語句模式的分類，不是對留言者作人格診斷。盤中閒聊包含反串、迷因與情緒宣洩；引文只能證明這句話出現過，不能證明作者真的照著交易。\n1. 單因謬誤：把市場縮成一個黑手\r#\r市場同時包含外資、內資、避險、被動資金與個別公司消息；把整段價格路徑只歸因於一個有意志的角色，故事會很順，證據卻通常不夠。\n#2554 · 03/04 09:29 · user7001：國安基金 怎麼還不護盤啊==\n#6894 · 03/04 11:51 · user7002：等等撿隔日沖 無腦多 主力就是提款機\n2. 賭徒謬誤：跌多了就「該」彈\r#\r先前已經跌多少，不會自動提高下一分鐘上漲的機率。價格可以很便宜，也可以在缺乏新資訊時繼續變便宜。\n#3192 · 03/04 09:40 · user7003：這麼快V 特價又沒了？\n#7785 · 03/04 12:33 · user7004：記憶體各買了幾百股，繼續等特價\n3. 錨定效應：月線、季線與整數都是答案\r#\r技術位置可以是風險管理參考，但把單一價位當成必然反轉點，會忽略波動、成交量與事件條件。\n#4150 · 03/04 10:12 · user7005：今天就是要守月線就對了\n#7539 · 03/04 12:25 · user7006：日韓都破月線 很好 之前不是都看亞股嗎\n4. 直線外推：今天的斜率一路畫到明天\r#\r人腦很容易把眼前最強烈的方向延伸出去；市場真正困難之處，正是斜率會在新資訊出現時改變。\n#5410 · 03/04 10:59 · user7007：今天周選結算,台股一定要撐在這裡\n#8952 · 03/04 13:07 · user7008：破線不一定空頭確立 歐印！！\n5. 從眾效應：多蛙、空蛙與「大家都知道」\r#\r群體標籤能快速製造安全感，也會把不同持倉、期限與風險承受度的人壓成同一種對手。\n#2483 · 03/04 09:28 · user7009：一堆大媽撿鑽石。開心\n#7917 · 03/04 12:37 · user7010：放心！一堆融資還在成本低，沒事沒事\n6. 後見之明偏誤：收盤後每個人都早知道\r#\r結果出現後，原先的多種可能性會從記憶裡消失。真正可檢驗的不是「早說」，而是事前是否留下方向、期限與失效條件。\n#3123 · 03/04 09:39 · user7011：這麼早就開拉了嗎\n#7379 · 03/04 12:15 · user7012：小丑上周不是說看錯 本週要做多\n7. 類比謬誤：把日韓、美股或夜盤直接翻譯\r#\r跨市場確實相關，但交易時段、權重、匯率與事件曝險不同；相關不等於可以一比一複製漲跌。\n#3308 · 03/04 09:42 · user7013：阿跟昨晚美股一摸摸一樣樣 操 那個大財團\n#8764 · 03/04 13:00 · user7014：多蛙這兩天都看夜盤爆V自嗨 結果早盤都被沙爆…\n8. 災難化與全有全無：不是 V，就是歸零\r#\r這比較接近認知扭曲而非形式邏輯謬誤：把連續的風險壓成生或死兩個選項，會讓部位管理退化成情緒口號。\n#2144 · 03/04 09:23 · user7015：我想這次真的完蛋了\n#8098 · 03/04 12:40 · user7016：人踩人 融資仔塊陶\n這份考古不能證明什麼\r#\r它不是情緒指標回測，也沒有證明某類留言能預測下一根 K 線。 關鍵字分類是可重現的抽樣入口，不是對全部 10,200 則留言做唯一正解標註。 同一句話可能同時包含多種偏誤；為了可讀性，每則引文只放在一個小節。 帳號已置換為合成代號（user0001 之類），與原帳號無對應關係、不可回推；保留樓層與時間，是為了讓讀者能回到匿名化原始檔核對。 2026 年截至 07/29 的前十大下跌日\r#\r名次 日期 收盤跌幅 跌點 文章 1 2026-07-17 -6.47% -2,953.71 「跌到收盤才停」 2 2026-07-28 -4.65% -2,030.83 「這次一定會 V？」 3 2026-03-09 -4.43% -1,489.12 「千點低開之後」 4 2026-03-04 -4.35% -1,494.77 「月線會救嗎」（本篇） 5 2026-07-29 -3.76% -1,564.18 「四萬點保衛戰」 6 2026-06-26 -3.64% -1,683.50 「指數在跌，嘴在做什麼」 7 2026-06-08 -3.48% -1,568.16 「夜盤只是誤會？」 8 2026-06-10 -3.31% -1,478.90 「反彈第二天又殺」 9 2026-07-24 -2.67% -1,195.97 「日韓跌，台股就會？」 10 2026-03-23 -2.45% -821.38 「明天會噴回來？」 資料口徑\r#\r指數排名與 OHLC：Yahoo Finance ^TWII 日線快照，抓取日 2026/07/29；排名用相鄰交易日收盤價計算。臺灣證券交易所亦提供發行量加權股價指數歷史資料供核對。 留言：PTT Stock 板當日「盤中閒聊」，以 BBS 層完整抓取；本篇產生前已核對 AID、標題與留言總數。 隱私：公開原始檔與本文引文的帳號一律置換為合成代號，不保留原字元、不可回推。同一個代號在所有檔案裡指同一人，跨日比對仍然成立。 ","date":"2026年3月4日","externalUrl":null,"permalink":"/zh-tw/stocks/taiex-chat-0304/","section":"股票","summary":"","title":"「月線會救嗎」：10,200 則盤中留言的謬誤考古（2026/03/04）","type":"stocks"},{"content":"","date":"2026년 2월 25일","externalUrl":null,"permalink":"/ko/tags/ai-%EC%A7%80%EC%9B%90-%EA%B0%9C%EB%B0%9C/","section":"Tags","summary":"","title":"AI 지원 개발","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/tags/ai-assisted-development/","section":"Tags","summary":"","title":"AI-Assisted Development","type":"tags"},{"content":"","date":"2026年2月25日","externalUrl":null,"permalink":"/ja/tags/ai%E6%94%AF%E6%8F%B4%E9%96%8B%E7%99%BA/","section":"Tags","summary":"","title":"AI支援開発","type":"tags"},{"content":"","date":"2026年2月25日","externalUrl":null,"permalink":"/zh-tw/tags/ai%E8%BC%94%E5%8A%A9%E9%96%8B%E7%99%BC/","section":"Tags","summary":"","title":"AI輔助開發","type":"tags"},{"content":"","date":"2026年2月25日","externalUrl":null,"permalink":"/zh-hans/tags/ai%E8%BE%85%E5%8A%A9%E5%BC%80%E5%8F%91/","section":"Tags","summary":"","title":"AI辅助开发","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/tags/app-store/","section":"Tags","summary":"","title":"App Store","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/id/tags/bagan-gantt/","section":"Tags","summary":"","title":"Bagan Gantt","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/vi/tags/bi%E1%BB%83u-%C4%91%E1%BB%93-gantt/","section":"Tags","summary":"","title":"Biểu Đồ Gantt","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/vi/tags/c%E1%BB%ADa-h%C3%A0ng-%E1%BB%A9ng-d%E1%BB%A5ng/","section":"Tags","summary":"","title":"Cửa Hàng Ứng Dụng","type":"tags"},{"content":"","date":"25 janvier 2026","externalUrl":null,"permalink":"/fr/tags/d%C3%A9veloppement-assist%C3%A9-par-lia/","section":"Tags","summary":"","title":"Développement Assisté Par L'IA","type":"tags"},{"content":"","date":"25 janvier 2026","externalUrl":null,"permalink":"/fr/tags/d%C3%A9veloppement-de-produits/","section":"Tags","summary":"","title":"Développement De Produits","type":"tags"},{"content":"","date":"25 janvier 2026","externalUrl":null,"permalink":"/fr/tags/d%C3%A9veloppeur-ind%C3%A9pendant/","section":"Tags","summary":"","title":"Développeur Indépendant","type":"tags"},{"content":"","date":"25 janvier 2026","externalUrl":null,"permalink":"/fr/tags/diagramme-de-gantt/","section":"Tags","summary":"","title":"Diagramme De Gantt","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/vi/tags/d%E1%BB%B1-%C3%A1n-ph%E1%BB%A5/","section":"Tags","summary":"","title":"Dự Án Phụ","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/tags/gantt-chart/","section":"Tags","summary":"","title":"Gantt Chart","type":"tags"},{"content":"\rPreface\r#\rIn this post, I\u0026rsquo;ll talk about the market, resources, ecosystem, and development process from the perspective of an indie developer. As a shameless plug, I\u0026rsquo;m using Gantt Planet as my running example: URL. I\u0026rsquo;ll admit upfront that these are just my side projects — the pressure is very different from someone who makes a living off this — so I\u0026rsquo;m only discussing the research approach here.\nThe Spark and the Stall\r#\rThe idea behind Gantt Planet was simple: free Gantt chart tools — whether desktop software, mobile apps, or web apps — are all pretty terrible to use. The ones that actually seem decent all charge money, so I figured I\u0026rsquo;d just build my own Gantt chart app.\nIt didn\u0026rsquo;t take long before I realized things weren\u0026rsquo;t that simple:\nViewing a spreadsheet-style Gantt chart on a phone screen is way too cramped A proper Gantt chart needs to connect to a ton of resources — email, contacts, meeting rooms, and so on Solving either of these problems is expensive. You\u0026rsquo;d need to spend a huge amount of time fine-tuning the UI and designing ideal usage flows, while accepting that some workflows simply can\u0026rsquo;t be integrated and have to be dropped.\nAs for resource integration, you\u0026rsquo;d need to handle sign-ins for all major platforms, deal with countless APIs and authentication protocols, and maintain all of it going forward.\nAt this point, I hit a wall — and when you\u0026rsquo;re working at a scale that doesn\u0026rsquo;t benefit from economies of scale, that\u0026rsquo;s pretty much inevitable.\nPivot After Pivot\r#\rIn moments like this, I like to take each factor and extend it a step or two outward, looking for a viable intersection where things might actually work.\nAs a developer driven by personal interest, \u0026ldquo;viable\u0026rdquo; means extremely low cost, plus a value proposition that\u0026rsquo;s small but clearly defined.\nAI helped me achieve the first part — extremely low cost.\nAs for the value part, it\u0026rsquo;s mostly self-defined, though bouncing ideas off AI can help crystallize things too.\nFor me, it mainly comes down to building something I\u0026rsquo;d actually want to use — something I\u0026rsquo;d enjoy looking at, at the very least. Beyond that, if nobody else has done it, there\u0026rsquo;s no free version, or there\u0026rsquo;s a clear differentiator, that counts as value too.\nAt this point, I started wondering: is there something that\u0026rsquo;s like a Gantt chart, but not really a Gantt chart?\nAnd then a picture formed in my mind.\nI remembered that when I use Gantt charts, I tend to put the more important items further down.\nThe bottommost item is usually the big-picture condition for completing the entire project — or it represents the project itself.\nBut what if there were items even below that bottom row — items even more important? What would those be?\nWell, there are plenty of things more important — they just have nothing to do with work. They\u0026rsquo;re about me. About life.\nAnd so it clicked: I wasn\u0026rsquo;t going to build a regular business Gantt chart. I was going to build a life Gantt chart.\nThe Next Step\r#\rSo I decided to build a Gantt chart that departs from the typical business use case.\nThis conveniently meant I no longer needed to integrate with online services,\nbecause now it was all about the user — just them, and nothing else.\nWith that, I\u0026rsquo;d taken one more step forward and kept the project alive for the time being. But could it lead to enough substance to be complete?\nI thought about self-management and the important-but-not-urgent things in life — they all have rhythms and frequencies.\nHealth matters, so companies do annual check-ups. Family matters, so you make sure to see your loved ones before too much time passes.\nCombined with the nature of Gantt charts, within any given time window, items overlap on the current day.\nAnd if you consider the span of an entire lifetime, every item is potentially relevant today. That meant I could collapse everything onto the center line of the UI.\nThis solved the cramped UI problem while expressing a set of values I found genuinely meaningful.\nThe actual timeline view: all life items converge on the calendar centerline — see everything that matters today at a glance\nCompleteness\r#\rOne of the App Store review guidelines is that your app can\u0026rsquo;t just replicate what a plain text webpage could do.\nFor example, a simple to-do list might not pass muster. So I had to make sure this app was more than just a spreadsheet — otherwise, Google Sheets could do the same thing.\nThe top-to-bottom visual flow of the spreadsheet reminded me of digging downward — like each day you only do the bare minimum surface-level tasks. There\u0026rsquo;s a Chinese idiom, \u0026ldquo;people floating above their work,\u0026rdquo; that captures this state perfectly.\nThe metaphor of more important items sitting at deeper layers made me want to make it more visual, more tangible. The immediate association was excavation — digging through geological strata, mining.\nThen came the question of implementation. Should I slightly curve each row of the spreadsheet? Add some perspective distortion?\nI thought about the context of this life Gantt chart — solitary and introspective.\nThe image that came to mind was: on the surface of a planet\u0026rsquo;s crust, one person digging alone. And then it hit me — isn\u0026rsquo;t that the golden-haired boy who waters his rose and tames a fox?\nSo I built a 3D version of the Gantt chart, using a mine shaft and gemstones as the visual representation of to-do items.\nAn even more radical approach would have been to keep only the planet version, but considering usability, review difficulty, and how intuitive it would be to understand, I decided to keep both views.\nThe 3D planet Gantt chart — mine shafts and gemstones as visual representations of life goals\nStill Missing a Desk\r#\rBack when I was still in school, I spent a lot of time sitting properly at my desk, alone — either studying or writing.\nUsing and thinking about this life Gantt chart felt like it was bringing me back to that desk — the one that\u0026rsquo;s long been thrown away.\nIf I completed something I only do once every three months or once a year — or even a long-term goal —\nI think I\u0026rsquo;d really want to write in a journal, or maybe write a letter to a close friend.\nI realized this Gantt chart was still missing a final emotional outlet. But if I added social media sharing, users wouldn\u0026rsquo;t be able to be fully honest.\nAnother option was in-app messaging between users, but there would never — now or in the future — be enough installs to support that, or at least an Android version would need to be available too. Either way, it wasn\u0026rsquo;t necessary for the first version.\nThe most self-consistent solution I landed on was the most versatile one: a chatbot.\nFeed the chatbot a bunch of literary classics and let it play the role of a \u0026ldquo;tree hollow\u0026rdquo; — a confidant — offering users some thoughtful feedback.\nFinal Thoughts\r#\rSo that\u0026rsquo;s the product development and decision-making behind this app.\nIt might look like I just kept changing direction until it was done, but in reality, there were tons of scrapped ideas and rejected features that I haven\u0026rsquo;t even mentioned.\nBeyond giving curious friends a window into the kinds of considerations that go into product development,\nthe last thing I want to emphasize — and the answer to the title — is that the indie developer\u0026rsquo;s niche and consideration is: doing whatever the hell makes you happy!\nI\u0026rsquo;m sure plenty of people will think this is too niche, or that it doesn\u0026rsquo;t match their taste or values.\nBut even so, with a bit of time and the help of AI, you can build the thing you want that doesn\u0026rsquo;t exist yet.\nYou get to be the boss — deciding what\u0026rsquo;s valuable and what\u0026rsquo;s worth building.\nYou get to be the designer — choosing your favorite layouts, colors, fonts, and images.\nYou get to be the PM — deciding how to write it and how complete the features need to be.\nAI will only get stronger. Even if it can\u0026rsquo;t do everything today, in the foreseeable future, you\u0026rsquo;ll be able to enjoy all of this too.\nThe App Store is now the new-era personal homepage — everyone can publish their own story.\nIf you\u0026rsquo;re interested, follow this blog. I\u0026rsquo;ll keep sharing real experiences and reflections from publishing on the App Store.\n","date":"25 February 2026","externalUrl":null,"permalink":"/posts/gantt-planet-intro/","section":"Blog","summary":"","title":"Gantt Planet: An Indie Developer's Niche and Considerations","type":"posts"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/tags/indie-developer/","section":"Tags","summary":"","title":"Indie Developer","type":"tags"},{"content":"","date":"25 janvier 2026","externalUrl":null,"permalink":"/fr/tags/magasin-dapplications/","section":"Tags","summary":"","title":"Magasin D'applications","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/vi/tags/nh%C3%A0-ph%C3%A1t-tri%E1%BB%83n-%C4%91%E1%BB%99c-l%E1%BA%ADp/","section":"Tags","summary":"","title":"Nhà Phát Triển Độc Lập","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/id/tags/pengembang-indie/","section":"Tags","summary":"","title":"Pengembang Indie","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/id/tags/pengembangan-dengan-bantuan-ai/","section":"Tags","summary":"","title":"Pengembangan Dengan Bantuan AI","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/id/tags/pengembangan-produk/","section":"Tags","summary":"","title":"Pengembangan Produk","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/vi/tags/ph%C3%A1t-tri%E1%BB%83n-%C4%91%C6%B0%E1%BB%A3c-h%E1%BB%97-tr%E1%BB%A3-b%E1%BB%9Fi-ai/","section":"Tags","summary":"","title":"Phát Triển Được Hỗ Trợ Bởi AI","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/vi/tags/ph%C3%A1t-tri%E1%BB%83n-s%E1%BA%A3n-ph%E1%BA%A9m/","section":"Tags","summary":"","title":"Phát Triển Sản Phẩm","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/tags/product-development/","section":"Tags","summary":"","title":"Product Development","type":"tags"},{"content":"","date":"25 janvier 2026","externalUrl":null,"permalink":"/fr/tags/projet-parall%C3%A8le/","section":"Tags","summary":"","title":"Projet Parallèle","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/id/tags/proyek-sampingan/","section":"Tags","summary":"","title":"Proyek Sampingan","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/tags/side-project/","section":"Tags","summary":"","title":"Side Project","type":"tags"},{"content":"","date":"25 February 2026","externalUrl":null,"permalink":"/id/tags/toko-aplikasi/","section":"Tags","summary":"","title":"Toko Aplikasi","type":"tags"},{"content":"","date":"2026년 2월 25일","externalUrl":null,"permalink":"/ko/tags/%EA%B0%84%ED%8A%B8-%EC%B0%A8%ED%8A%B8/","section":"Tags","summary":"","title":"간트 차트","type":"tags"},{"content":"","date":"2026년 2월 25일","externalUrl":null,"permalink":"/ko/tags/%EC%82%AC%EC%9D%B4%EB%93%9C-%ED%94%84%EB%A1%9C%EC%A0%9D%ED%8A%B8/","section":"Tags","summary":"","title":"사이드 프로젝트","type":"tags"},{"content":"","date":"2026년 2월 25일","externalUrl":null,"permalink":"/ko/tags/%EC%95%B1%EC%8A%A4%ED%86%A0%EC%96%B4/","section":"Tags","summary":"","title":"앱스토어","type":"tags"},{"content":"","date":"2026년 2월 25일","externalUrl":null,"permalink":"/ko/tags/%EC%9D%B8%EB%94%94-%EA%B0%9C%EB%B0%9C%EC%9E%90/","section":"Tags","summary":"","title":"인디 개발자","type":"tags"},{"content":"","date":"2026년 2월 25일","externalUrl":null,"permalink":"/ko/tags/%EC%A0%9C%ED%92%88-%EA%B0%9C%EB%B0%9C/","section":"Tags","summary":"","title":"제품 개발","type":"tags"},{"content":"","date":"2026年2月25日","externalUrl":null,"permalink":"/ja/tags/%E3%82%AC%E3%83%B3%E3%83%88%E3%83%81%E3%83%A3%E3%83%BC%E3%83%88/","section":"Tags","summary":"","title":"ガントチャート","type":"tags"},{"content":"","date":"2026年2月25日","externalUrl":null,"permalink":"/ja/tags/%E3%82%B5%E3%82%A4%E3%83%89%E3%83%97%E3%83%AD%E3%82%B8%E3%82%A7%E3%82%AF%E3%83%88/","section":"Tags","summary":"","title":"サイドプロジェクト","type":"tags"},{"content":"","date":"2026年2月25日","externalUrl":null,"permalink":"/ja/tags/%E3%83%97%E3%83%AD%E3%83%80%E3%82%AF%E3%83%88%E9%96%8B%E7%99%BA/","section":"Tags","summary":"","title":"プロダクト開発","type":"tags"},{"content":"","date":"2026年2月25日","externalUrl":null,"permalink":"/zh-hans/tags/%E4%BA%A7%E5%93%81%E5%BC%80%E5%8F%91/","section":"Tags","summary":"","title":"产品开发","type":"tags"},{"content":"","date":"2026年2月25日","externalUrl":null,"permalink":"/ja/tags/%E5%80%8B%E4%BA%BA%E9%96%8B%E7%99%BA%E8%80%85/","section":"Tags","summary":"","title":"個人開発者","type":"tags"},{"content":"","date":"2026年2月25日","externalUrl":null,"permalink":"/zh-hans/tags/%E7%8B%AC%E7%AB%8B%E5%BC%80%E5%8F%91%E8%80%85/","section":"Tags","summary":"","title":"独立开发者","type":"tags"},{"content":"","date":"2026年2月25日","externalUrl":null,"permalink":"/zh-tw/tags/%E7%8D%A8%E7%AB%8B%E9%96%8B%E7%99%BC%E8%80%85/","section":"Tags","summary":"","title":"獨立開發者","type":"tags"},{"content":"","date":"2026年2月25日","externalUrl":null,"permalink":"/zh-hans/tags/%E7%94%98%E7%89%B9%E5%9B%BE/","section":"Tags","summary":"","title":"甘特图","type":"tags"},{"content":"","date":"2026年2月25日","externalUrl":null,"permalink":"/zh-tw/tags/%E7%94%98%E7%89%B9%E5%9C%96/","section":"Tags","summary":"","title":"甘特圖","type":"tags"},{"content":"","date":"2026年2月25日","externalUrl":null,"permalink":"/zh-tw/tags/%E7%94%A2%E5%93%81%E9%96%8B%E7%99%BC/","section":"Tags","summary":"","title":"產品開發","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/id/tags/ai-di-perangkat/","section":"Tags","summary":"","title":"AI Di Perangkat","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/vi/tags/ai-tr%C3%AAn-thi%E1%BA%BFt-b%E1%BB%8B/","section":"Tags","summary":"","title":"AI Trên Thiết Bị","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/id/tags/aplikasi-ios/","section":"Tags","summary":"","title":"Aplikasi IOS","type":"tags"},{"content":"","date":"22 janvier 2026","externalUrl":null,"permalink":"/fr/tags/application-ios/","section":"Tags","summary":"","title":"Application IOS","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/tags/claude/","section":"Tags","summary":"","title":"Claude","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/tags/claude-code/","section":"Tags","summary":"","title":"Claude Code","type":"tags"},{"content":"","date":"22 janvier 2026","externalUrl":null,"permalink":"/fr/tags/code-claude/","section":"Tags","summary":"","title":"Code Claude","type":"tags"},{"content":"","date":"22 janvier 2026","externalUrl":null,"permalink":"/fr/tags/g%C3%A9meaux/","section":"Tags","summary":"","title":"Gémeaux","type":"tags"},{"content":"","date":"22 janvier 2026","externalUrl":null,"permalink":"/fr/tags/g%C3%A9meaux-cli/","section":"Tags","summary":"","title":"Gémeaux Cli","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/tags/gemini/","section":"Tags","summary":"","title":"Gemini","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/tags/gemini-cli/","section":"Tags","summary":"","title":"Gemini Cli","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/tags/handwriting-recognition/","section":"Tags","summary":"","title":"Handwriting Recognition","type":"tags"},{"content":"","date":"22 janvier 2026","externalUrl":null,"permalink":"/fr/tags/ia-sur-lappareil/","section":"Tags","summary":"","title":"IA Sur L'appareil","type":"tags"},{"content":"","date":"22 janvier 2026","externalUrl":null,"permalink":"/fr/tags/interface-utilisateur-rapide/","section":"Tags","summary":"","title":"Interface Utilisateur Rapide","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/tags/ios-app/","section":"Tags","summary":"","title":"IOS App","type":"tags"},{"content":"","date":"2026년 2월 22일","externalUrl":null,"permalink":"/ko/tags/ios-%EC%95%B1/","section":"Tags","summary":"","title":"IOS 앱","type":"tags"},{"content":"\rPreface\r#\rKana Juku is the first app I ever built and shipped to the App Store.\nSince it was my first, there\u0026rsquo;s a full story arc to share.\nThis series covers the development process, how I used AI assistance and how that evolved, working with public datasets and copyright considerations, and more.\nIf other apps have noteworthy stories, I\u0026rsquo;ll publish those separately.\nThis post focuses on the transition from chatbots to AI agents starting in Q4 2025.\nThings move fast in this space, so I\u0026rsquo;ve bluntly timestamped the key moments.\nAbout the App\r#\rIf you have an Apple device, feel free to download it and give it a try.\nSeveral upcoming posts will also use this app as a running example — topics like cleaning ETL datasets, Apple Create ML, PyTorch, VOICEVOX, on-device large language models, and more.\nKana Juku: URL\nDevelopment Timeline\r#\rMotivation\r#\rMy family and I are both interested in learning Japanese, and I\u0026rsquo;ve long wanted a Japanese-learning app that perfectly fits our needs.\nMy family\u0026rsquo;s pain point is that they don\u0026rsquo;t read English, so the romaji in most textbooks and apps is meaningless to them.\nFor me, I really wanted kana displayed alongside their kanji origins (e.g., \u0026ldquo;あ\u0026rdquo; derives from \u0026ldquo;安\u0026rdquo;).\nAnother annoyance: I installed the Japanese keyboard for occasional use, but switching input methods every day meant an extra tap to skip past the Japanese keyboard — a small friction that added up.\nEarly Preparation\r#\r[Q4 2024]\nI was between jobs at the time, so I had the bandwidth to take Udemy courses. Since I had some JavaScript experience, I started with React \u0026amp; Expo.\nAt this stage I was following along with course content — simple web-style pages, plus extras like GPS, camera control, and fetching remote data.\nBut since it wasn\u0026rsquo;t Apple\u0026rsquo;s native ecosystem, there was a lot of extra tooling to manage.\n[Q1 2025]\nAfter hesitating for a long time, I bought a Mac Mini and switched entirely to Apple\u0026rsquo;s own SwiftUI. Again, I learned from Udemy courses.\nMost of my time went into getting comfortable with basic UI components and layouts, plus all the fundamental features — data persistence, fetching data, embedding maps — and their SwiftUI equivalents.\nSwiftUI is more modern and isn\u0026rsquo;t as tightly coupled to Xcode as UIKit, but it\u0026rsquo;s also harder to predict how a SwiftUI layout will actually look. Early on I cared too much about that and burned a lot of time experimenting.\n[Q3 2025]\nSince I had a day job and could only code in the evenings — and not every evening at that — progress was slow. I was basically building out the basic skeleton and plugging in the Japanese language data.\nWith a first app, it\u0026rsquo;s hard to foresee the final shape, so I kept revising. Sometimes I\u0026rsquo;d circle back to rewatch course videos for features I now knew I needed. Essentially, I was paying tuition.\nUp to this point, starting from Q1 2024, plain chatbots like ChatGPT were already a big help for coding.\nBut the copy-paste cycle and having to explain mountains of context was incredibly time-consuming. The output often missed the mark on the first try or drifted off course, sending me right back to the copy-paste loop. It never reached a positive feedback cycle — it was only useful as a learning reference.\nAt the time, the hottest tool was actually the Cursor editor with its tab-autocomplete, but it required a subscription for meaningful usage, so I didn\u0026rsquo;t try it.\nMeanwhile, Claude was already gaining popularity as the best model for coding, and Anthropic had released Claude Code — an AI agent that runs on your local machine. But again, it required a subscription, so I didn\u0026rsquo;t try it.\nPivoting to AI Agents\r#\r[Q4 2025]\nAt this point I expected I\u0026rsquo;d only ever subscribe to one chatbot at a time, and I had just switched from ChatGPT to Google Gemini.\nSpec-Driven Development (SDD) was trending, and Google had launched Gemini CLI — their answer to Claude Code — so I finally gave it a shot.\nI discovered that agents eliminated the copy-paste step entirely, massively boosting efficiency. The step of pasting code back and hunting for which lines to change was also gone.\nBy then I was convinced: for coding, you should use an agent, not a chatbot. So I went ahead and subscribed to Claude to use Claude Code (CC from here on).\nCC\u0026rsquo;s underlying model was clearly stronger. Its comprehension of conversations and its ability to execute as expected were already remarkably reliable.\nControlling the Computer, and Opus 4.5\r#\rOne time my Mac Mini\u0026rsquo;s disk was completely full and the machine was unusable. I just asked CC what to do — the same way I\u0026rsquo;d ask a question on a chatbot\u0026rsquo;s web page.\nCC came back with a concrete plan: which directories could be cleared, what could be moved to an external drive, and so on.\nI was worried it might brick my computer, so I approved each step one at a time. In the end, everything went smoothly.\nI wasn\u0026rsquo;t very familiar with macOS or the Xcode build environment. That\u0026rsquo;s when I realized AI has at least an 80% understanding of everything — including things I don\u0026rsquo;t know — and that being able to write code is roughly equivalent to being able to operate a computer.\nBecause CC could directly control the machine, it moved freely between directories, wrote code, saw its own errors, and fixed them — a fully self-sustaining positive feedback loop.\nThe development speed with an agent was on a completely different level, and the fact that I\u0026rsquo;d waited three extra months before switching to CC made me feel pretty foolish.\nThe time wasted was staggering, both subjectively and objectively.\nSubjectively: if I had adopted the latest tools earlier, the previous three months of work could have been done in two to three weeks.\nObjectively: other people using the latest tools were more productive than me and shipping their products sooner.\nMy earlier refusal to try — saving maybe half an hour of setup time and a few hundred dollars in subscription fees — ended up wasting vast stretches of my life.\nThis might also explain why so many people are obsessed with chasing the latest AI product news.\nAt least that\u0026rsquo;s how it is for me — I can\u0026rsquo;t afford not to stay on top of the latest releases. It\u0026rsquo;s a form of time-management risk hedging.\n[November 24, 2025]\nOpus 4.5 was released. Opus is Claude\u0026rsquo;s highest-tier flagship model, and version 4.5 had just dropped.\nBeyond significant performance improvements across the board compared to its predecessor, the biggest difference was its understanding of intent.\nThe old version essentially did exactly what you pointed at (which was already quite good, honestly). Starting with 4.5, after receiving your request, it would first summarize and plan to some degree. In human terms: it became sharper, more experienced.\nYou no longer needed to spell out which file to modify and how. You could describe the end goal like a manager or executive, and it would break it down and plan the next couple of steps on its own.\nThis planning capability boosted efficiency even further. As I mentioned, AI already knows at least 80% of everything — now it was proactively doing the next steps of work, and doing them well.\nCombined with this, I was able to operate at a much higher level of abstraction. More and more was delegated to CC. Gradually, I stopped needing to read or edit code myself.\nAfter Opus 4.5 came out, the debate on social media about whether AI can write code essentially ended.\nFor full-time software engineers and seasoned pros, I can\u0026rsquo;t speak to their experience.\nBut compared to myself: things that would have taken me one to two years could now be done in two to three months.\nThe output settled at just beyond the edges of my own knowledge — I was actually the biggest bottleneck.\nEnd of Part 1\n","date":"22 February 2026","externalUrl":null,"permalink":"/posts/kana_juku_dev_1/","section":"Blog","summary":"","title":"Kana Juku Dev Log (Part 1): From Chatbots to AI Agents","type":"posts"},{"content":"","date":"22 janvier 2026","externalUrl":null,"permalink":"/fr/tags/kitui/","section":"Tags","summary":"","title":"KitUI","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/id/tags/kode-claude/","section":"Tags","summary":"","title":"Kode Claude","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/vi/tags/m%C3%A3-claude/","section":"Tags","summary":"","title":"Mã Claude","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/vi/tags/nh%E1%BA%ADn-d%E1%BA%A1ng-ch%E1%BB%AF-vi%E1%BA%BFt-tay/","section":"Tags","summary":"","title":"Nhận Dạng Chữ Viết Tay","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/tags/on-device-ai/","section":"Tags","summary":"","title":"On-Device AI","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/id/tags/pengenalan-tulisan-tangan/","section":"Tags","summary":"","title":"Pengenalan Tulisan Tangan","type":"tags"},{"content":"","date":"22 janvier 2026","externalUrl":null,"permalink":"/fr/tags/reconnaissance-de-l%C3%A9criture-manuscrite/","section":"Tags","summary":"","title":"Reconnaissance De L'écriture Manuscrite","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/vi/tags/song-t%E1%BB%AD/","section":"Tags","summary":"","title":"Song Tử","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/vi/tags/song-t%E1%BB%AD-cli/","section":"Tags","summary":"","title":"Song Tử Cli","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/tags/swiftui/","section":"Tags","summary":"","title":"SwiftUI","type":"tags"},{"content":"","date":"22 janvier 2026","externalUrl":null,"permalink":"/fr/tags/udemie/","section":"Tags","summary":"","title":"Udemie","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/tags/udemy/","section":"Tags","summary":"","title":"Udemy","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/id/tags/ui-cepat/","section":"Tags","summary":"","title":"UI Cepat","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/tags/uikit/","section":"Tags","summary":"","title":"UIKit","type":"tags"},{"content":"","date":"22 February 2026","externalUrl":null,"permalink":"/vi/tags/%E1%BB%A9ng-d%E1%BB%A5ng-ios/","section":"Tags","summary":"","title":"Ứng Dụng IOS","type":"tags"},{"content":"","date":"2026년 2월 22일","externalUrl":null,"permalink":"/ko/tags/%EC%8A%A4%EC%9C%84%ED%94%84%ED%8A%B8ui/","section":"Tags","summary":"","title":"스위프트UI","type":"tags"},{"content":"","date":"2026년 2월 22일","externalUrl":null,"permalink":"/ko/tags/%EC%8C%8D%EB%91%A5%EC%9D%B4-%EC%9E%90%EB%A6%AC-cli/","section":"Tags","summary":"","title":"쌍둥이 자리 Cli","type":"tags"},{"content":"","date":"2026년 2월 22일","externalUrl":null,"permalink":"/ko/tags/%EC%8C%8D%EB%91%A5%EC%9D%B4%EC%9E%90%EB%A6%AC/","section":"Tags","summary":"","title":"쌍둥이자리","type":"tags"},{"content":"","date":"2026년 2월 22일","externalUrl":null,"permalink":"/ko/tags/%EC%98%A8%EB%94%94%EB%B0%94%EC%9D%B4%EC%8A%A4-ai/","section":"Tags","summary":"","title":"온디바이스 AI","type":"tags"},{"content":"","date":"2026년 2월 22일","externalUrl":null,"permalink":"/ko/tags/%EC%9C%A0%EB%8D%B0%EB%AF%B8/","section":"Tags","summary":"","title":"유데미","type":"tags"},{"content":"","date":"2026년 2월 22일","externalUrl":null,"permalink":"/ko/tags/%ED%81%B4%EB%A1%9C%EB%93%9C/","section":"Tags","summary":"","title":"클로드","type":"tags"},{"content":"","date":"2026년 2월 22일","externalUrl":null,"permalink":"/ko/tags/%ED%81%B4%EB%A1%9C%EB%93%9C-%EC%BD%94%EB%93%9C/","section":"Tags","summary":"","title":"클로드 코드","type":"tags"},{"content":"","date":"2026년 2월 22일","externalUrl":null,"permalink":"/ko/tags/%ED%95%84%EA%B8%B0-%EC%9D%B8%EC%8B%9D/","section":"Tags","summary":"","title":"필기 인식","type":"tags"},{"content":"","date":"2026年2月22日","externalUrl":null,"permalink":"/zh-hans/tags/%E6%89%8B%E5%86%99%E8%AF%86%E5%88%AB/","section":"Tags","summary":"","title":"手写识别","type":"tags"},{"content":"","date":"2026年2月22日","externalUrl":null,"permalink":"/zh-tw/tags/%E6%89%8B%E5%AF%AB%E8%BE%A8%E8%AD%98/","section":"Tags","summary":"","title":"手寫辨識","type":"tags"},{"content":"","date":"2026年2月22日","externalUrl":null,"permalink":"/ja/tags/%E6%89%8B%E6%9B%B8%E3%81%8D%E8%AA%8D%E8%AD%98/","section":"Tags","summary":"","title":"手書き認識","type":"tags"},{"content":"\rAbout This Site\r#\rThe title refers to \u0026ldquo;The Miniature Boat\u0026rdquo; (核舟記) — a classical Chinese text about an impossibly detailed carving on a tiny boat. It means this site can\u0026rsquo;t carry much, just small crafts and the joy of making things, leaving behind a trace of information. Here I share real examples of using AI to help build apps, make small tools, and improve everyday efficiency — along with ideas, reflections, and setbacks. I won\u0026rsquo;t rehash the hot takes everyone\u0026rsquo;s already discussing. In short, the focus is on the process of mining, not repeatedly introducing the shovel. About Me\r#\rQQder — Taiwan-based sysadmin \u0026amp; indie developer · qqder339@gmail.com Until age 24, I identified as a humanities person. After that, I became a system administrator (the \u0026ldquo;admin\u0026rdquo; that error messages tell you to contact). I use AI in a punk rock way — simple chords, rough technique, but genuine expression. The blog posts here are written by me — the ideas and the first draft are mine, and the editing is done by me together with AI. Other language versions are machine-translated. The fun parts should be enjoyed firsthand. Philosophy\r#\rExperience is ownership — experiencing something new takes priority over whether it can make money. Success doesn\u0026rsquo;t have to be mine — if someone else is willing to do the same thing and does it better, I\u0026rsquo;ll find something else to work on. After the invention of cinema, human life has been extended by at least three times. — Yi Yi: A One and a Two\nAI is this era\u0026rsquo;s new medium for extending human life. ","externalUrl":null,"permalink":"/about/","section":"QQder · The Miniature Boat","summary":"The Miniature Boat","title":"About","type":"page"},{"content":"Currently in pre-submission with the App Store. This page is a preview — the listing will go live once it ships.\nWhy this exists\r#\rPhoto libraries quietly grow into a weight you stop wanting to look at. The screenshots, the duplicates, the failed shots, the once-bright moments that now feel heavy — they pile up. Deleting them one by one feels both tedious and slightly sad, so most people don\u0026rsquo;t. Afterglow was built for that hesitation. It turns letting go into a calm, deliberate act, then quietly turns the act of release into something that grows.\nThe app has a single thesis: you don\u0026rsquo;t have to fight forgetting. Let it become a garden you can walk through.\nThree Acts: Triage, Grace, Garden\r#\rAfterglow is shaped as three tabs, mirrored after a three-act narrative. You don\u0026rsquo;t have to use them in order, but most sessions naturally pass through all three.\nTriage\r#\rA card-stack swipe. Left to forget, right to keep, up to favorite. The Vision framework quietly skips blurs and burst duplicates so the rhythm stays calm — you\u0026rsquo;re not punished by a wall of near-identical RAW shots from a single afternoon. Smart Sort surfaces similar, blurry, or overexposed shots first, so the early swipes feel productive without feeling brutal.\nGrace\r#\rA 7-day grace period. Photos you swipe away don\u0026rsquo;t disappear right away — recall any of them with a long press. After the 7 days, iOS\u0026rsquo;s Recently Deleted album then gives you another 30 days, so 37 days of room to change your mind. Grace exists because letting go shouldn\u0026rsquo;t be irreversible. It is the breath between \u0026ldquo;I think I\u0026rsquo;m done with this\u0026rdquo; and \u0026ldquo;yes, really.\u0026rdquo;\nGarden\r#\rWhat you truly release blooms as a 3D abandoned-amusement-park ecosystem. Cool light becomes moss; warm light becomes flowers; high-energy frames become fireflies. Each plant is a quiet farewell. You can walk the garden, watch the seasons shift, and share a composed image of it to keep — without keeping the original photos.\nWhat the ML actually does (and doesn\u0026rsquo;t)\r#\rThe ✦ icon on the Triage tab is Smart Sort: a local pass over the oldest 200 unreviewed photos, ranked by \u0026ldquo;likely to delete\u0026rdquo; descending. Settings exposes two more entries — Similar Photos (perceptual clusters, e.g. 127 burst frames of the same newborn pose) and Browse by Scene (coarse buckets: people, food, animal, landscape, document, other).\nAll three run on Apple\u0026rsquo;s Vision framework: image classification, face capture quality, text detection, and feature print embeddings — the same neural nets Photos.app uses internally. On top, Afterglow layers a small, transparent weighted formula: screenshots, blur, age, and burst non-heroes add points; favorites and sharp human faces subtract them; photos belonging to a near-duplicate cluster get an extra nudge if they\u0026rsquo;re not the cluster\u0026rsquo;s best frame. Simple, auditable, no black magic.\nIn all honesty:\nFor raw classification quality, Apple wins. Photos.app has facial identity clustering, natural-language search, places, and curated memory montages. Our six scene buckets are not in the same league. What Apple doesn\u0026rsquo;t do is hand you a curated deletion queue. Their Duplicates album and Screenshots album exist, but you have to go look. Afterglow surfaces deletion candidates proactively as a swipeable queue, wrapped in a 7-day grace period, redeemed by the garden. You decide. ML only ranks — the swipe is always your finger. Anything you send away enters the 7-day grace; after that, iOS Recently Deleted keeps it for 30 more days. Three layers of recall. Curious how a single photo gets scored? A developer toggle in Settings — \u0026ldquo;Show analysis info\u0026rdquo; — surfaces the formula breakdown under each thumbnail (screen +0.30, blur ×0.18, face quality ×0.92, …). It\u0026rsquo;s the layer Apple never lets you see, and Afterglow is quietly fond of letting you peek.\nWhy on-device\r#\rAfterglow is 100% on-device. Every Vision analysis, SwiftData write, and 3D render happens on your iPhone. There is no network, no account, no third-party SDK, no analytics, no advertising ID, no telemetry. Your photos never leave your device. We can\u0026rsquo;t see what you tossed; iCloud doesn\u0026rsquo;t see it either.\nThis isn\u0026rsquo;t a feature. It is the architectural posture of the app. A cleanup tool earns trust by handling sensitive content with as few external moving parts as possible — so we removed all of them.\nWhat grows in the garden\r#\rThe garden has 11 archetypes, unlocking gradually as the diversity of your released photos grows:\nMoss — cool, blue-shifted scenes Mushroom — low-light or shadowy frames Flower — warm, saturated palettes Vine — long-form sequences and panoramas Firefly — high-energy, motion-rich shots Butterfly — colorful, cheerful frames Fruit tree — densely populated scenes Paper — text-heavy screenshots Scrap — failed exposures, technical misfires Statue — portrait-shaped compositions Wild grass — anything else Each is a tiny codex entry, a reminder that what you released wasn\u0026rsquo;t worthless — it was just done.\nQuiet farewell\r#\rAfterglow is hobbyist software, written for one person at a time. There is no leaderboard, no streak, no notification asking you to come back. The seasonal recap will tell you what colors your spring or autumn had, and that\u0026rsquo;s it. You don\u0026rsquo;t have to fight forgetting. Let it be a garden you can walk through.\n","externalUrl":null,"permalink":"/apps/afterglow/","section":"Apps","summary":"","title":"Afterglow","type":"apps"},{"content":"Last Updated: 2026-05-07\n1. Overview\r#\rAfterglow, developed by ChengChe Lee, is a card-stack photo cleanup app with a 3D abandoned-amusement-park \u0026ldquo;garden\u0026rdquo; that grows from the photos you release. Every part of the app — Vision analysis, SwiftData persistence, and 3D rendering — runs 100% on-device.\nIn short: We do NOT collect, store, or transmit any of your personal data to external servers. There is no account, no telemetry, no analytics.\n2. Data We Do NOT Collect\r#\rThis app does not collect, transmit, or share any of the following:\nPhotos, thumbnails, or photo embeddings Face data, biometric data, or face embeddings Location data (GPS / geo-coordinates from EXIF or system services) EXIF metadata beyond the local capture date used for sorting Contacts, calendars, or other personal data Usage analytics, crash reports, or behavioral tracking Advertising identifiers (IDFA) Persistent device identifiers transmitted off-device 3. Photo Library Access\r#\rAfterglow requests NSPhotoLibraryUsageDescription in order to read your photos for triage and to compute on-device classification (cool light → moss, warm light → flower, etc.).\nPhotos remain on-device at every step. Limited Photo Access (selecting only specific photos) is fully supported. You can change this any time in Settings → Privacy \u0026amp; Security → Photos → Afterglow. The app never writes photo content out of the device, and never sends image data over the network. 4. Third-Party SDKs\r#\rAfterglow uses zero third-party SDKs for analytics, advertising, crash reporting, or attribution. Specifically:\nNo Google Analytics, no Firebase No Facebook SDK, no AppsFlyer, no Adjust, no Branch No advertising networks No crash reporting services (Sentry, Crashlytics, etc.) The only Swift Package Manager dependencies are first-party Apple frameworks plus narrowly-scoped open-source utilities used for on-device rendering.\n5. Required Reason API Declarations\r#\rIn line with Apple\u0026rsquo;s Required Reason API rules, Afterglow declares the following reasons in PrivacyInfo.xcprivacy:\nUserDefaults (CA92.1) — to store user preferences (e.g. last-used tab, garden camera position) on-device. All declared APIs are used only for app-internal purposes. No values are transmitted off-device.\n6. Children\u0026rsquo;s Privacy\r#\rAfterglow is rated 4+. The app does not collect data from anyone, including children. We do not engage in targeted advertising or profiling of any kind.\n7. Your Rights\r#\rBecause Afterglow does not collect or transmit personal data, there is no server-side data to request, export, or delete. To remove all data associated with the app, simply uninstall it — this deletes the local SwiftData store, garden state, and any cached thumbnails immediately.\n8. Changes to This Policy\r#\rWe may update this policy as the app evolves. Any changes will be reflected by an updated Last Updated date at the top of this page. Material changes (e.g. introducing a backend or SDK) would be announced in the app\u0026rsquo;s release notes.\n9. Contact\r#\rqqder339@gmail.com\nSubject: [Afterglow] Privacy Policy Inquiry\n","externalUrl":null,"permalink":"/privacy/afterglow/","section":"Privacy Policies","summary":"","title":"Afterglow — Privacy Policy","type":"privacy"},{"content":"Privacy Policy\nFAQ\r#\rQ: I granted limited photo access — can Afterglow still triage all my photos?\nA: Only the photos you selected are visible to Afterglow. Tap the in-app banner that says \u0026ldquo;Limited access\u0026rdquo; to grant full access if you want Afterglow to triage your entire library. You can also change this any time in Settings → Privacy \u0026amp; Security → Photos → Afterglow.\nQ: I swiped a photo away by accident — how do I recover it?\nA: Open the Grace tab and long-press the thumbnail to recall it. Photos stay there for 7 days. After that, iOS\u0026rsquo;s Recently Deleted album gives you another 30 days, totaling 37 days of recovery time.\nQ: When does a photo \u0026ldquo;bloom\u0026rdquo; in the Garden?\nA: When you explicitly send it from Grace to the Garden, or when the 7-day grace period ends. Then Vision quietly classifies it (cool light → moss, warm light → flower, high energy → firefly) and a plant sprouts.\nQ: Why do I see only some archetypes?\nA: There are 11 archetypes total: moss, mushroom, flower, vine, firefly, butterfly, fruit tree, paper, scrap, statue, and wild grass. They unlock as the diversity of your released photos grows. If your releases all look similar (e.g. all screenshots), expect mostly papers for a while.\nQ: Does Afterglow upload my photos anywhere?\nA: No. Every Vision analysis, SwiftData write, and 3D render happens on your iPhone. There is zero network access for the core flow. The app does not call any backend, third-party SDK, or analytics service.\nQ: How do I share my garden?\nA: Tap Share garden in the Garden tab. A composed image of your garden is rendered locally and offered through the system share sheet — save to Photos, send via Messages, AirDrop, anywhere you like. No cloud round-trip.\nQ: Smart Sort surfaces a screenshot I want to keep — am I supposed to delete it?\nA: Smart Sort is just a suggestion. Right-swipe to keep anything you want; the suggestion ranking adapts. You\u0026rsquo;re never forced to act on a Smart Sort suggestion.\nContact\r#\r📧 qqder339@gmail.com\nPlease include: device model, iOS version, app version, and a brief description of what you were doing when the issue happened.\nAfterglow collects no user data. All Vision analysis, SwiftData writes, and 3D rendering run on-device. We have no access to your photos.\n","externalUrl":null,"permalink":"/support/afterglow/","section":"Support","summary":"Support and contact for Afterglow","title":"Afterglow Support","type":"support"},{"content":"This is the entry page for all apps currently released and actively maintained. The portfolio is organised into two product lines:\nOffline Growth — tools for long-term learning, reflection, and personal growth. Covers two sub-categories: Language Learning and Self-Reflection. Digital Citizen — apps focused on authenticity, memory, and personal digital agency. The current entry point is Democracy EDC (EveryDay Carry). Jump straight into any product from the cards below. Every entry includes its App Store link, support page, and privacy policy.\n","externalUrl":null,"permalink":"/apps/","section":"Apps","summary":"The current catalog of released and maintained apps","title":"Apps","type":"apps"},{"content":"\rWhen recordings get challenged, verifiable authenticity is what\u0026rsquo;s scarce\r#\rLowering the barrier to video and audio production is mostly a good thing. The problem is that fabrication, re-editing, and context-stripping get cheaper at the same time, so \u0026ldquo;I recorded it\u0026rdquo; has drifted away from \u0026ldquo;I can prove this is how it really happened.\u0026rdquo; Atomic Presence steps into that gap: it lets you start building a verifiable evidence chain the moment you hit record.\nThe situations it\u0026rsquo;s designed for are the ones where you\u0026rsquo;d worry about the footage being challenged later: interviews, witness accounts, whistleblowing, contested scenes, any context where a recording might get disputed, re-cut, or forged. Casual everyday capture sits outside its target.\nHow it differs from ordinary recording tools\r#\rMost recording tools think \u0026ldquo;record the file first, worry about preservation later.\u0026rdquo; Atomic Presence weaves hash chains, dynamic QR codes, and digital signatures into the capture flow while it\u0026rsquo;s happening. Verifiability is the core of the product from the start, not a patch bolted on afterwards.\nThat makes it feel more like a technical defense tool for risk scenarios. You may not reach for it every day, but when you do need it, you\u0026rsquo;ll want it already installed, with a workflow you already know, rather than scrambling to assemble tools in the moment.\nWhy four protection levels: different risks, different costs\r#\rThe protection levels correspond to real scenario differences. Sometimes you only need to signal \u0026ldquo;this is being recorded\u0026rdquo; to the other party; sometimes you need something closer to evidence-grade integrity verification. Having intermediate steps lets the tool sit inside actual workflows instead of offering only a crude on/off switch.\nFor journalists, legal professionals, citizen journalists, and anyone who regularly needs a clean record of a conversation, this is valuable. What helps them is a recording tool that leaves fewer ambiguous zones in the file itself.\nPrivacy and offline are part of credibility\r#\rA tool that claims to stand for authenticity loses credibility if its core data handling depends heavily on external servers. Atomic Presence keeps the critical computation on-device, partly for privacy and partly to reduce external dependencies inside the evidence chain itself. The fewer third parties your material passes through, the easier it is to explain later what did and didn\u0026rsquo;t happen to it.\nIf you want a recording tool you\u0026rsquo;ll already be familiar with when a dispute breaks out, Atomic Presence is worth getting installed and tried before you need it.\n","externalUrl":null,"permalink":"/apps/atomic-presence/","section":"Apps","summary":"","title":"Atomic Presence","type":"apps"},{"content":"Last Updated: 2026-04-15\n1. Overview\r#\rAtomic Presence, developed by ChengChe Lee, is an anti-deepfake tool that uses cryptographic hash chains, digital signatures, and audio watermarking to help users self-verify the integrity of their recordings.\nIn short: We do NOT collect, store, or transmit any of your personal data to external servers. All cryptographic operations and verification are performed on-device.\n2. Data We Do NOT Collect\r#\rThis app does not collect:\nPersonally Identifiable Information (name, email, phone number) Location data Device identifiers Usage analytics or tracking data 3. Locally Stored Data\r#\rThe following data is stored strictly on your device and never transmitted externally:\nAudio/Video Files: All recorded content stored in your device\u0026rsquo;s local storage Hash Chain Records: SHA-256 hash sequences and corresponding verification data Digital Signatures: Signature data generated by on-device Curve25519 algorithm Verification Reports: Integrity reports and metadata records Anonymized Device Identifier: Each .evidence.json embeds a 16-character hex prefix of SHA-256(identifierForVendor), used only to correlate recordings from the same device during verification. This identifier lives only inside evidence files on your device, is never transmitted to any server, and cannot be reversed back to the original device information 4. Cryptographic Features (Fully Offline)\r#\rAll core features are completed on-device without network connection:\nHash Chain Generation: Real-time SHA-256 hash sequences; all computation runs locally Digital Signing: Uses Curve25519 algorithm to sign recordings on-device Audio Watermarking: Embeds FSK signals in recordings; all signal processing runs on-device Verification: Integrity verification computed locally 5. Important Note\r#\rThe content processed by this app (audio, video) may contain sensitive information. All processing occurs on your device, and we cannot and will never access any of your recorded content.\n6. Third-Party Services\r#\rThis app does NOT use any third-party analytics or advertising frameworks (No Google Analytics, No Facebook SDK, No Ads).\n7. Network Access\r#\rThis app requires no network connection to use all features. The only network access is:\nExternal Links: Opens browser when tapping relevant links 8. Contact Us\r#\r📧 qqder339@gmail.com\nSubject: Atomic Presence Privacy Policy Inquiry\n","externalUrl":null,"permalink":"/privacy/atomic-presence/","section":"Privacy Policies","summary":"","title":"Atomic Presence — Privacy Policy","type":"privacy"},{"content":" FAQ\r#\rQ: The QR code is unclear in the video and can\u0026rsquo;t be scanned during verification?\nA: Ensure sufficient screen brightness during recording, and keep the camera 30–50 cm from the screen. The QR code updates once per second — the camera needs to be able to focus clearly. If the problem persists, try reducing the recording resolution.\nQ: Audio watermark verification fails?\nA: Watermark verification may fail if: the audio was heavily compressed (e.g., forwarded via WhatsApp), the audio was truncated, or there was excessive background noise. Record in a quiet environment and use the original audio file for verification.\nQ: The digital signature is invalid on a new device?\nA: Each device\u0026rsquo;s signing key is stored in the iOS Keychain, and a new device generates a different key. You do NOT need to manually export the public key — every .evidence.json written by the app already embeds the public key used for that recording\u0026rsquo;s signature, so any verifier who holds the evidence file can verify it regardless of which device they\u0026rsquo;re on.\nQ: The app crashed during recording — is the file still there?\nA: When the app crashes unexpectedly, partial recordings may remain in the Documents directory. Reopen the app, tap the VERIFY button at the top of the main screen, and check the three tabs (Level 1 / Level 2 / Level 3) for any recoverable files.\nQ: Hash chain verification shows \u0026ldquo;integrity broken\u0026rdquo; but I didn\u0026rsquo;t edit the recording?\nA: Possible causes include: the app was interrupted by the system during recording, low battery, or a write error due to insufficient storage. Ensure sufficient battery and storage before recording.\nTroubleshooting\r#\rEnsure the device has sufficient storage (recommend at least 2 GB available) Keep the screen on during recording to avoid system interruptions Force quit and relaunch the app Check iOS version ≥ 17.0 If a specific scenario consistently causes issues, screenshot the error message and email us Contact Support\r#\r📧 qqder339@gmail.com\nSubject: [Atomic Presence] Issue Description\nPlease include: device model, iOS version, app version, recording mode (video/audio), steps to reproduce.\nThis app collects no user data. All cryptographic operations run entirely on-device. We have no access to your recordings.\n","externalUrl":null,"permalink":"/support/atomic-presence/","section":"Support","summary":"Support and contact for Atomic Presence","title":"Atomic Presence Support","type":"support"},{"content":"\rSound as something you design, not just play in the background\r#\rMost white noise apps eventually converge on the same conclusion: play rain, waves, or wind, and hope it helps you focus or sleep. Auditory Companion aims further. Instead of bundling a handful of ambient samples, it treats \u0026ldquo;how sound forms an inhabitable space\u0026rdquo; as the product\u0026rsquo;s core.\nSo what you see is three distinct systems rather than a playlist: a noise synthesizer, a scene mixer, and a reader. You can shape your acoustic environment from several directions: a stable, emotionally flat spectrum when you need it; the layered spatial feel of a rainy night, a fireplace, or a café at other times; a frame that lets reading voice and background sound coexist when the task calls for it.\nWhen it\u0026rsquo;s most valuable\r#\rFor long stretches of reading, writing, or deep work, \u0026ldquo;play some nature sounds\u0026rdquo; usually falls short; what\u0026rsquo;s actually useful is an adjustable soundscape. Auditory Companion fits here. Lay down a noise floor with the synthesizer, layer in event sounds and loops in the scene mixer, then let the reader speak the text aloud. The fit to your current state is much closer than simply opening a Spotify playlist.\nRelaxation and pre-sleep are another common scenario. Many people want silence that isn\u0026rsquo;t actually silent: a sound that masks the outside world without demanding attention. That\u0026rsquo;s where adjustable noise and scene mixing earn their place: you\u0026rsquo;re not forced to pick between a handful of canned presets.\nThe real story is sound-control granularity\r#\rThink of it as a small personal sound workstation. The synthesizer handles spectrum and texture; the scene mixer handles atmosphere and spatial feel; the reader handles content input. Each module stands on its own; together they form a complete focus system.\nThat\u0026rsquo;s also why it resonates with \u0026ldquo;people who need background sound.\u0026rdquo; You get to find the configuration you can actually sit with for hours, rather than accepting whatever soundfield someone else prepared.\nThe detail work shows up clearly: the synthesizer puts four noise colors (white, pink, brown, green) and multiple parameters directly in your hands, going well past an on/off switch. The scene mixer lets you layer over a hundred audio samples into stackable scenes instead of playing one file. The reader wraps on-device TTS, automatic background-audio ducking, and a full-screen player into a single flow.\nKeeping sound and reading data on-device is practically significant\r#\rWhat you read, listen to, and paste in tends to be private, especially when an app supports clipboard reading or local TTS. If that content travels to a server, the experience sours immediately. Auditory Companion keeps synthesis, mixing, and reading on-device, and that shapes whether you\u0026rsquo;ll comfortably use it as a daily tool, not just whether you\u0026rsquo;ll try it once.\nIf you\u0026rsquo;re looking for a sound engine that can keep you company through work, reading, downtime, and immersive listening, rather than an app measured by \u0026ldquo;how many ambient tracks,\u0026rdquo; it\u0026rsquo;s worth trying this one firsthand.\n","externalUrl":null,"permalink":"/apps/auditory-companion/","section":"Apps","summary":"","title":"Auditory Companion","type":"apps"},{"content":"Last Updated: 2026-04-15\n1. Overview\r#\rAuditory Companion, developed by ChengChe Lee, is a sophisticated audio engine combining real-time DSP noise synthesis, 108 ambient sound samples, and AI-powered text-to-speech reading.\nIn short: We do NOT collect, store, or transmit any of your personal data to external servers.\n2. Data We Do NOT Collect\r#\rThis app does not collect:\nPersonally Identifiable Information (name, email, phone number) Location data Device identifiers Usage analytics or tracking data 3. Locally Stored Data\r#\rThe following data is stored strictly on your device and never transmitted externally:\nSoundscape Settings: Your saved mixing configurations and favorite scenes Reading Content: Articles, clipboard text, and other reading materials (processed locally only) User Settings: Volume levels, sound preferences, auto-ducking settings, etc. 4. On-Device AI Features\r#\rText-to-Speech (TTS) and audio processing run on-device:\nAI Voice Reading: Uses iOS built-in TTS, or an optional downloadable MeloTTS on-device model; all speech synthesis runs on-device Auto-Ducking: DSP signal processing runs entirely locally, analyzing voice and background audio in real-time to automatically adjust volume 5. Third-Party Services\r#\rThis app does NOT use any third-party analytics or advertising frameworks (No Google Analytics, No Facebook SDK, No Ads).\n6. Network Access\r#\rCore features (noise synthesis, scene mixing, iOS built-in TTS reading) operate fully offline. Network access occurs only when you explicitly trigger it:\nDownloading the MeloTTS model (Optional): When you choose to download the on-device TTS model in Settings, the app fetches the model files from a public source and caches them locally External Links: Opens the system browser when tapping relevant links These requests transmit only the URL of the file you chose to download; no personally identifiable information is attached.\n7. Contact Us\r#\r📧 qqder339@gmail.com\nSubject: Auditory Companion Privacy Policy Inquiry\n","externalUrl":null,"permalink":"/privacy/auditory-companion/","section":"Privacy Policies","summary":"","title":"Auditory Companion — Privacy Policy","type":"privacy"},{"content":" FAQ\r#\rQ: There\u0026rsquo;s static or crackling during audio playback?\nA: Some noise may come from Bluetooth headphone connection issues. Try switching to wired headphones to test. If the issue persists with wired headphones, try adjusting the sample rate in Settings or restarting the audio engine.\nQ: The TTS reading sounds very unnatural?\nA: The app uses iOS\u0026rsquo;s built-in TTS engine. You can switch between different voice packs and speech rates in Settings. TTS quality for some languages (like Traditional Chinese) depends on your iOS version — updating to the latest iOS typically improves quality.\nQ: Auto-Ducking sometimes doesn\u0026rsquo;t work?\nA: Auto-Ducking requires both background sound and TTS reading to be playing simultaneously. If only one audio source is active, ducking won\u0026rsquo;t trigger. Make sure both sources are playing and that Auto-Ducking is enabled in Settings.\nQ: Saved soundscape settings disappear on next launch?\nA: This may happen if the app was force-closed before settings were saved. After adjusting settings, confirm the save before exiting the app.\nQ: Can playback continue after the screen locks?\nA: Yes. The app supports background audio playback and continues after screen lock. If playback stops automatically, check the app\u0026rsquo;s background refresh permission in iOS Settings.\nTroubleshooting\r#\rCheck that volume is not muted (physical mute switch + media volume) Try switching the audio output device (wired vs. Bluetooth) Force quit and relaunch the app Restart your device to clear potential audio routing conflicts Check iOS version ≥ 17.0 Contact Support\r#\r📧 qqder339@gmail.com\nSubject: [Auditory Companion] Issue Description\nPlease include: device model, iOS version, app version, headphone/speaker model, steps to reproduce.\nThis app collects no user data. All audio processing is performed entirely on-device.\n","externalUrl":null,"permalink":"/support/auditory-companion/","section":"Support","summary":"Support and contact for Auditory Companion","title":"Auditory Companion Support","type":"support"},{"content":"\rThe usual problem is poorly-fitted material, not lack of effort\r#\rThe hardest part of learning English is rarely vocabulary size. It\u0026rsquo;s opening an article and having no idea whether it\u0026rsquo;ll be comfortably challenging or discouragingly hard. Material that\u0026rsquo;s too easy gives no sense of progress; material that\u0026rsquo;s too hard simply erodes patience. English N+1 was built for that specific gap.\nIt turns Krashen\u0026rsquo;s i+1 theory into a working product: estimate your level first, then have AI generate content just above your current ability. The point isn\u0026rsquo;t \u0026ldquo;bolt AI onto a study app\u0026rdquo;; it\u0026rsquo;s automating the work of making material fit the person.\nHow you\u0026rsquo;d actually use it\r#\rA common flow: take the placement test to locate your rough CEFR range, pick a topic you actually want to read about, and let the app generate an article at your level. Save unfamiliar words as you go, and they enter a review cadence automatically. You don\u0026rsquo;t need to separately hunt for articles, run dictionary lookups, take notes, and then wire up flashcards; the app was designed around that whole habit from the start.\nThis matters especially for people anxious about English. Instead of opening by asking you to prove what you know, it lets you expand your boundaries while still understanding most of what\u0026rsquo;s in front of you. That \u0026ldquo;only slightly harder than where I am now\u0026rdquo; margin is where most people can actually sustain practice.\nAI here is an engine, not a performance\r#\rPlenty of AI English products highlight how natural their chat is. What actually keeps learners around is whether content generation quality is consistent and whether the review rhythm flows. English N+1 behaves like a curriculum engine that tracks your level, balancing text difficulty, vocabulary density, and topic interest so you don\u0026rsquo;t gamble on \u0026ldquo;maybe this one fits\u0026rdquo; every session.\nRunning the model on-device matters too. Learning records, level information, and reading preferences are private, and especially sensitive when you\u0026rsquo;re looking at your own weaknesses. Keeping this data local makes long-term use much more comfortable than a cloud service.\nModel selection isn\u0026rsquo;t pushed in your face either. The placement test establishes your rough CEFR range, and a local model takes over generation based on your device\u0026rsquo;s capability and memory state. What you experience is \u0026ldquo;open the app and start reading\u0026rdquo; rather than being blocked by setup. Hardware-adaptive under the hood, simple on the surface. That proves more useful than showcasing AI for its own sake.\nWho should download it\r#\rIf flashy English apps that never quite fit have left you worn out, this one works differently. It earns retention by being useful every time you open it, not through streaks or gamified loops. Students, self-learners, people returning to English, or anyone who wants a steady dose of reading during the commute will find this more structured than scraping the web for material.\nPeople usually give up on English not from lack of effort, but because the material doesn\u0026rsquo;t fit. English N+1 only tries to fix that one thing — so the next time you open an article, you don\u0026rsquo;t have to guess whether it\u0026rsquo;s about to break your patience.\n","externalUrl":null,"permalink":"/apps/english-n-plus-1/","section":"Apps","summary":"","title":"English N+1","type":"apps"},{"content":"Last Updated: 2026-05-03\n1. Overview\r#\rEnglish N+1, developed by ChengChe Lee, is an English learning app featuring CEFR-level assessment and on-device AI conversation technology.\nIn short: We do NOT collect, store, or transmit any of your personal data to external servers.\n2. Data We Do NOT Collect\r#\rThis app does not collect:\nPersonally Identifiable Information (name, email, phone number) Location data Device identifiers Usage analytics or tracking data 3. Locally Stored Data\r#\rThe following data is stored strictly on your device and never transmitted externally:\nLearning Progress \u0026amp; Level: CEFR assessment results and study records Reading Article History: Recent AI-generated reading articles, kept only in a local file Word Collections: Saved vocabulary and learning notes User Settings: Language preferences, difficulty settings, etc. 4. Offline AI Features\r#\rAll AI features run entirely on-device without network connection:\nAI Conversation Practice: Uses local Llama 3.2 or Qwen 2.5 models; all inference runs on-device Article Generation: Personalized learning articles generated locally based on your level Level Assessment: CEFR level evaluation computed on-device AI models require a one-time download before first use (user-initiated); all features work offline after download.\n5. Third-Party Services\r#\rThis app does NOT use any third-party analytics or advertising frameworks (No Google Analytics, No Facebook SDK, No Ads).\n6. Network Access\r#\rNetwork access is restricted to:\nDownloading AI Models (Optional, one-time): Only connects when you explicitly choose to download LLM model resources External Links: Opens browser when tapping relevant links Other than the above, the app does not initiate network connections.\n7. Contact Us\r#\r📧 qqder339@gmail.com\nSubject: English N+1 Privacy Policy Inquiry\n","externalUrl":null,"permalink":"/privacy/english-n-plus-1/","section":"Privacy Policies","summary":"","title":"English N+1 — Privacy Policy","type":"privacy"},{"content":" FAQ\r#\rQ: After the CEFR level test, the difficulty feels off. What can I do?\nA: The initial test is a quick vocabulary-based assessment and may not perfectly match your actual level. You can manually adjust your level from the home screen, or retake the test. After using the app for a while, it will automatically adapt based on your answer history.\nQ: What model does the AI conversation feature need? How large is it?\nA: Based on your device\u0026rsquo;s performance, the app recommends an appropriate model (Llama 3.2 or Qwen 2.5). Model size varies — from a few hundred megabytes to several gigabytes depending on the model. After downloading, conversations work completely offline — no internet required.\nQ: AI conversation responses are very slow or frozen?\nA: On-device AI inference speed depends on your device\u0026rsquo;s performance. Older iPhones will be noticeably slower. Try selecting a smaller model in Settings.\nQ: The generated articles are too difficult to understand?\nA: Articles are generated based on your CEFR level. If they feel too hard, lower your level on the home screen, or manually select a lower difficulty when generating articles.\nQ: Will my study records and saved vocabulary be backed up?\nA: All data is currently stored locally on your device only. iCloud backup is not supported. Uninstalling the app will erase all records.\nTroubleshooting\r#\rAI model fails to load: Ensure enough free storage for the model you\u0026rsquo;ve selected, and check that the download wasn\u0026rsquo;t interrupted App crashes during AI conversation: Try switching to a smaller model in Settings Force quit and relaunch the app Check iOS version ≥ 17.6 Contact Support\r#\r📧 qqder339@gmail.com\nSubject: [English N+1] Issue Description\nPlease include: device model, iOS version, app version, steps to reproduce.\nThis app collects no user data. All AI conversations are processed entirely on-device.\n","externalUrl":null,"permalink":"/support/english-n-plus-1/","section":"Support","summary":"Support and contact for English N+1","title":"English N+1 Support","type":"support"},{"content":"\rA place to anchor long-term life projects\r#\rMost task management tools handle today, this week, and this month well enough. They struggle with the important-but-not-urgent projects: reading, exercise, language learning, processing emotions, keeping up certain relationships. These aren\u0026rsquo;t unimportant. They just get crowded out by noisier, more immediate demands. Gantt Planet is designed for exactly those projects.\nThe goal isn\u0026rsquo;t to make you busier. It\u0026rsquo;s to help you see what was already worth your time. Timeline, 3D planet, AI tree hole, art collection: these surface as different modules, but all serve the same outcome. Self-discipline becomes a rhythm you can visualise, sense, and return to, rather than a willpower grind.\nWhy it tends to stay on your phone longer than a typical productivity tool\r#\rTraditional to-do tools think in list logic: done means check off, undone means accumulating pressure. Gantt Planet is closer to tending a small universe. You\u0026rsquo;re letting a planet grow its own terrain and memories, not wiping items off a list. That visual language makes long-term goals harder to abandon, because they\u0026rsquo;re no longer just cold rows of text.\nThe timeline view shows what today deserves attention; the 3D planet view shows the shape of overall progress. The former pulls you back to reality; the latter reminds you why you started. Together they keep \u0026ldquo;today\u0026rdquo; and \u0026ldquo;a lifetime\u0026rdquo; in the same frame.\nIts starting point is the personal \u0026ldquo;important but not urgent,\u0026rdquo; not an enterprise project manager shrunk down. You can track daily, weekly, monthly, or longer cadences on the same timeline, then turn completion into 3D planet terrain and collection entries. That\u0026rsquo;s where it diverges most from productivity software that mostly helps you pack tasks in more tightly.\nThe tree hole and the art collection: the reason to come back\r#\rLong-term habit building isn\u0026rsquo;t just a planning problem; a lot of it is emotional. The issue is often not that you don\u0026rsquo;t know what to do. You\u0026rsquo;re tired, annoyed, distracted, or simply sick of productivity tools that never respond. Gantt Planet\u0026rsquo;s AI tree hole exists to sit with that layer. When you just want to talk something out and reorient yourself, it doesn\u0026rsquo;t demand productivity from you.\nThe collection system turns \u0026ldquo;finishing things\u0026rdquo; from obligation into something that accumulates. Not everyone needs this, but for people who get worn down by monotony, it\u0026rsquo;s exactly how patience gets rebuilt. Completing tasks gradually unlocks illustrations and collectibles, closer to leaving marks on a long-running life project than to KPI pressure.\nPrivacy and offline matter more here than usual\r#\rGoals, journal entries, emotions, conversations: this content is more personal than what goes into a typical productivity app. Part of Gantt Planet\u0026rsquo;s value is that you don\u0026rsquo;t have to hand this data to an external service to get the companionship and visualisation. For many people, inner content only gets written down when the data really stays on their own device.\nGantt Planet won\u0026rsquo;t fill your calendar for you, and it won\u0026rsquo;t flash red when you fall behind. What it\u0026rsquo;s good at is letting the things you don\u0026rsquo;t want to abandon accumulate, quietly, into a planet you can actually see.\n","externalUrl":null,"permalink":"/apps/gantt-planet/","section":"Apps","summary":"","title":"Gantt Planet","type":"apps"},{"content":"Last Updated: 2026-04-15\n1. Overview\r#\rGantt Planet, developed by ChengChe Lee, is a life goal management app combining 3D visual habit tracking with an on-device AI companion.\nIn short: We do NOT collect, store, or transmit any of your personal data to external servers. Your habits, journals, and conversations belong only to you.\n2. Data We Do NOT Collect\r#\rThis app does not collect:\nPersonally Identifiable Information (name, email, phone number) Location data Device identifiers Usage analytics or tracking data 3. Locally Stored Data\r#\rThe following data is stored strictly on your device and never transmitted externally:\nHabits \u0026amp; Goals: All items, completion records, and timeline data AI Conversation Logs: All conversations with the built-in AI stored locally Journals \u0026amp; Mood Records: All journal content Art Collection: Unlocked stickers and illustration records 3D Planet State: Your planet\u0026rsquo;s terrain and growth data User Settings: All preference settings 4. Offline AI Features\r#\rThe AI companion feature runs entirely on-device:\nAI Conversations: Uses local Large Language Models (LLM); all inference runs on-device; conversation content is never transmitted to any server Model Download: AI models require a one-time download before first use (user-initiated); fully offline after download 5. Third-Party Services\r#\rThis app does NOT use any third-party analytics or advertising frameworks (No Google Analytics, No Facebook SDK, No Ads).\n6. Network Access\r#\rNetwork access is restricted to:\nDownloading AI Models (Optional, one-time): Only connects when you explicitly choose to download Downloading Art Collection Stickers (On-demand): When you unlock art rewards, the app fetches matching images from a public GitHub repository and caches them locally for offline use Weather Information (Optional): If you enable real weather, only minimal regional data is sent to retrieve weather External Links: Opens browser when tapping relevant links These requests transmit only the URL of the resource you chose or triggered; no personally identifiable information is attached.\n7. Contact Us\r#\r📧 qqder339@gmail.com\nSubject: Gantt Planet Privacy Policy Inquiry\n","externalUrl":null,"permalink":"/privacy/gantt-planet/","section":"Privacy Policies","summary":"","title":"Gantt Planet — Privacy Policy","type":"privacy"},{"content":" FAQ\r#\rQ: Does the AI companion chat require an internet connection?\nA: No. The AI companion uses an on-device local model. All conversations are processed completely offline and are never sent to any server. A one-time model download (typically a few gigabytes, depending on which model you pick) is required on first use, after which everything works offline.\nQ: The 3D planet is laggy?\nA: The 3D planet requires some GPU performance. If it\u0026rsquo;s lagging on an older device, try lowering the render quality or disabling particle effects in Settings. Recommended device: iPhone 12 or newer.\nQ: Habit items on the timeline have disappeared?\nA: All data is stored locally on your device. If data disappears unexpectedly, check if items were accidentally deleted (they may be restorable from the recycle bin). If not, please email us with your app version.\nQ: The weather feature shows incorrect information?\nA: The weather feature requires location permission. Please ensure this app is allowed location access in iOS Settings \u0026gt; Privacy \u0026gt; Location Services. If already allowed but still incorrect, try toggling the weather feature off and on again.\nQ: Unlocked art stickers have disappeared?\nA: Sticker unlock records are stored locally and will be erased if the app is uninstalled. If data disappears without uninstalling, please email us with details.\nTroubleshooting\r#\rForce quit and relaunch the app Check available storage (AI model + 3D assets need 2+ GB) Check iOS version ≥ 17.0 If 3D rendering is abnormal, try resetting the planet display settings in Settings Contact Support\r#\r📧 qqder339@gmail.com\nSubject: [Gantt Planet] Issue Description\nPlease include: device model, iOS version, app version, steps to reproduce (screenshots preferred).\nThis app collects no user data. All AI conversations and habit records are processed entirely on-device.\n","externalUrl":null,"permalink":"/support/gantt-planet/","section":"Support","summary":"Support and contact for Gantt Planet","title":"Gantt Planet Support","type":"support"},{"content":"\rLearning kana straight into your hands\r#\rMost beginner Japanese materials quietly assume you\u0026rsquo;re willing to sit through a long \u0026ldquo;romaji transition period.\u0026rdquo; For native Chinese speakers, that\u0026rsquo;s usually the less natural path. You already have a strong sense of character form and stroke order, and you\u0026rsquo;re used to learning visually and through writing. Kana Juku starts from that premise and designs around it.\nRather than prettifying the hiragana/katakana chart, it connects \u0026ldquo;seeing the form, writing the form, typing the form, recognising the form\u0026rdquo; into a single loop. The payoff: you reach direct kana recognition sooner and rely on romaji as a crutch for less time.\nWhy this approach suits Chinese speakers in particular\r#\rA Chinese speaker\u0026rsquo;s real advantage lies in a strong sensitivity to character structure and visual form, less so in pronunciation. Kana Juku amplifies that advantage. You memorise kana through handwriting, image recognition, a custom keyboard, and shape association, which makes learning feel like picking up a new script rather than grinding through rote repetition.\nPeople who\u0026rsquo;ve quit halfway often did so because the tool\u0026rsquo;s angle didn\u0026rsquo;t match them, not because they didn\u0026rsquo;t try hard. Kana Juku\u0026rsquo;s job is to correct that angle.\nA real doorway forward, not just a memorisation drill\r#\rMemorising kana is only the starting line. The real challenge is turning it into input, recognition, and comprehension ability. That\u0026rsquo;s why the app goes beyond static drills and includes handwriting recognition, a custom keyboard, and AI assistance. What you\u0026rsquo;re building is muscle memory closer to how kana actually gets used.\nThe design suits two kinds of learners especially: people starting Japanese who want a low-pressure entry point, and people who learned before, forgot, and now want familiarity back. The first group needs to skip detours; the second needs to rebuild recognition. The app works for both.\nPrivacy and offline have practical weight here\r#\rLanguage learning tools tend to drift toward content-platform behaviour over time. You feel like you\u0026rsquo;re learning, but you\u0026rsquo;re really being shuffled between recommendations. Kana Juku stays restrained. The focus sits on real input and recognition training, and both the local AI and data processing run on-device. You don\u0026rsquo;t have to trade your usage habits for a little learning convenience.\nYou don\u0026rsquo;t have to let romaji lead you through kana. Kana Juku shows you that entering through the shapes works too — and often gets you there faster.\n","externalUrl":null,"permalink":"/apps/kana-juku/","section":"Apps","summary":"","title":"Kana Juku","type":"apps"},{"content":"Last Updated: 2026-04-15\n1. Overview\r#\rKana Juku, developed by ChengChe Lee, is a Japanese kana learning app designed for native Chinese speakers.\nIn short: We do NOT collect, store, or transmit any of your personal data to external servers.\n2. Data We Do NOT Collect\r#\rThis app does not collect:\nPersonally Identifiable Information (name, email, phone number) Location data Device identifiers Usage analytics or tracking data 3. Locally Stored Data\r#\rThe following data is stored strictly on your device and never transmitted externally:\nLearning Progress: Tracks your kana learning status User Settings: Saves your preferences Handwriting Input: Processed in real-time memory and discarded immediately; no files are saved Widget Data: Uses iOS shared container mechanism to display kana on home screen widgets (local only) 4. Offline AI Features\r#\rAll AI features operate completely offline:\nHandwriting Recognition: Uses on-device machine learning models; all processing is local Text-to-Speech: Uses pre-downloaded audio assets AI Assistance: Uses local Large Language Models (LLM); inference is performed on-device without data upload 5. Third-Party Services\r#\rThis app does NOT use any third-party analytics or advertising frameworks (No Google Analytics, No Facebook SDK, No Ads).\n6. Network Access\r#\rNetwork access is restricted to:\nDownloading AI Models (Optional): Only connects when you explicitly choose to download local model resources External Links: Opens the browser when you tap \u0026ldquo;Rate on App Store\u0026rdquo; or \u0026ldquo;Privacy Policy\u0026rdquo;; opens the browser when you use the \u0026ldquo;Search Web\u0026rdquo; function after translation/recognition Other than the above, the app does not initiate network connections.\n7. Contact Us\r#\rIf you have questions about this Privacy Policy, please contact:\n📧 qqder339@gmail.com\nSubject: Kana Juku Privacy Policy Inquiry\n","externalUrl":null,"permalink":"/privacy/kana-juku/","section":"Privacy Policies","summary":"","title":"Kana Juku — Privacy Policy","type":"privacy"},{"content":" FAQ\r#\rQ: Handwriting recognition keeps making mistakes. What should I do?\nA: Make sure you\u0026rsquo;re not writing too fast — pause briefly after each stroke before lifting the pen. The recognition model needs complete stroke information. If the problem persists, try resetting the recognition calibration in Settings.\nQ: How do I download the local AI model? Do I still need internet after downloading?\nA: When you first use an AI feature, the app will prompt you to download the model (a few hundred MB). Once downloaded, all AI features work fully offline — no internet connection required.\nQ: The custom keyboard doesn\u0026rsquo;t appear in other apps?\nA: The built-in keyboard in Kana Juku is for in-app use only and is not a system-level keyboard extension. To type Japanese in other apps, please use the iOS system Japanese keyboard.\nQ: The home screen widget isn\u0026rsquo;t updating kana?\nA: Try long-pressing to remove the widget from the home screen, then re-adding it. If it still doesn\u0026rsquo;t update, force-quit and relaunch the app, or restart your device.\nQ: My learning progress has disappeared?\nA: Progress is stored locally on your device. Uninstalling the app will erase all data. iCloud backup is not currently supported. If progress disappears without uninstalling, please email us with details.\nTroubleshooting\r#\rForce quit and relaunch the app (swipe up on the app in the app switcher) Check iOS version ≥ 17.0 Check available storage (AI models require ~1–2 GB) If none of the above works, uninstall and reinstall (note: progress data will be erased) Contact Support\r#\r📧 qqder339@gmail.com\nSubject: [Kana Juku] Issue Description\nPlease include: device model, iOS version, app version, steps to reproduce (screenshots welcome).\nThis app collects no user data. All data is stored locally on your device.\n","externalUrl":null,"permalink":"/support/kana-juku/","section":"Support","summary":"Support and contact for Kana Juku","title":"Kana Juku Support","type":"support"},{"content":"","externalUrl":null,"permalink":"/privacy/","section":"Privacy Policies","summary":"Privacy policies for all apps","title":"Privacy Policies","type":"privacy"},{"content":"\rWhat matters is getting the learning rhythm right\r#\rMost Python learning tools get stuck at two extremes. One side gives you fragmented questions; answering them still leaves you in the dark about where you\u0026rsquo;re actually weak. The other side drops you into a full IDE that tends to scare beginners off. Python Dimensions bridges those extremes, helping you first build reading ability, grammatical sense, and logical sense before pushing toward more complete coding capability.\nThe core idea is less \u0026ldquo;do lots of questions\u0026rdquo; and more \u0026ldquo;break learning into three layers.\u0026rdquo; Points are vocabulary and concept recognition; Lines are syntax and local structure; Surfaces are complete program flow. This layering works for complete beginners and also for people who already know where they\u0026rsquo;re stuck and want an efficient way to patch gaps.\nWhat situations it\u0026rsquo;s most useful for\r#\rIf you\u0026rsquo;re preparing for PCEP, TQC+, or CPE, the app is well suited to daily maintenance. You don\u0026rsquo;t need to open a laptop to get started; in 10 to 20 minute windows you can run through multiple-choice questions, fill in a few blanks, or re-sequence a program flow. That low friction matters more over the long run than intense burst study.\nFor self-taught beginners, the app also doesn\u0026rsquo;t behave like a machine that only reports right and wrong. You can use the question types to sketch the basic outline, then move into the playground to actually run code and understand why one variant works and another doesn\u0026rsquo;t. Knowledge stops living purely in memory and starts becoming your own judgement.\nOn-device AI here isn\u0026rsquo;t a gimmick\r#\r\u0026ldquo;AI tutor\u0026rdquo; often triggers the question, \u0026ldquo;is this about to upload my content to the cloud?\u0026rdquo; Python Dimensions places AI in a useful role that doesn\u0026rsquo;t compromise privacy. When you answer incorrectly, it can hint based on the question\u0026rsquo;s context. When you want to confirm a syntax idea, you can just ask, instead of bouncing between search engines and forums.\nJust as importantly, none of this requires handing your learning history to an external server. For students, that lowers the barrier to use. For teachers, parents, or anyone wary of data leakage, it turns the app into something closer to a long-term learning tool rather than a casual demo.\nThe AI layer also goes beyond \u0026ldquo;a chat model stuffed in for show.\u0026rdquo; The question bank, error context, context-aware retrieval, and a directly executable Python playground operate inside the same loop. You answer, ask, then run code to verify; when needed, capability analytics let you see whether you\u0026rsquo;re stuck at syntax, concepts, or program flow.\nWhy this app deserves a permanent place on your phone\r#\rThe learning tools people actually keep opening are the ones that sense when you\u0026rsquo;re about to give up, more than the ones packed with features. Python Dimensions gathers question training, AI hints, and an executable environment onto a single device. The point is to let you push forward a little, even in the moments you\u0026rsquo;d otherwise scroll away.\nWhat actually moves the needle isn\u0026rsquo;t the rush of fifty problems in one sitting. It\u0026rsquo;s the three minutes you\u0026rsquo;re willing to open the app each day. Python Dimensions is built around those three minutes.\n","externalUrl":null,"permalink":"/apps/python-dimensions/","section":"Apps","summary":"","title":"Python Dimensions","type":"apps"},{"content":"Last Updated: 2026-05-26\n1. Overview\r#\rPython Dimensions, developed by ChengChe Lee, is a Python programming learning app featuring a built-in Python 3.13 runtime, an on-device AI tutor, and a system-wide Python coder keyboard.\nIn short: We do not collect any personally identifiable information; we only collect fully anonymous usage statistics to help us improve the app. All learning content, code, and AI conversations stay on your device.\n2. Data We Do NOT Collect\r#\rThis app does not collect:\nPersonally Identifiable Information (name, email, phone, Apple ID) Location data Advertising identifier (IDFA) The content you type into questions, code, AI conversations, or AI prompts Your raw IP address (TelemetryDeck briefly uses it at ingest to derive country, then discards it; raw IPs are never persisted) 3. Locally Stored Data\r#\rThe following data is stored strictly on your device and never transmitted externally:\nLearning Progress — answer records and error tracking across all question types (Points/MCQ, Lines/Cloze, Planes/Parsons) Code — programs you write in the built-in Playground and via the Python system keyboard AI Conversation Logs — conversations with the AI tutor stored locally User Settings — difficulty preferences, interface settings, keyboard preferences 4. Offline AI and Local Execution\r#\rAll AI features and code execution run entirely on-device without a network connection:\nAI Tutor — Uses a local Large Language Model (LLM) to provide hints and explanations; all inference runs on-device Python Runtime — The built-in Python 3.13 interpreter runs entirely on-device; your code is never sent to any server Python System Keyboard — The keyboard extension runs in its sandbox without Full Access; it never transmits any text you type AI models require a one-time download before first use (user-initiated). After download, all features work offline.\n5. Anonymous Usage Analytics (TelemetryDeck)\r#\rTo help us understand which features are useful, find where new users get stuck in onboarding, and catch performance issues, this app uses TelemetryDeck (provider headquartered in Germany, GDPR-compliant) to collect anonymous usage signals.\nWhy we collect\r#\rImprove the new-user experience. Step-by-step onboarding events (welcome shown, persona picked, completed, skipped) tell us which step loses people, so we can refine the copy and flow. Optimize performance. perf_* events measure cold start, tab switching, LLM load, and Python first-run timing, so we know which surfaces need work. Guide content direction. Which dimensions (Point/Line/Plane) and which certifications (PCEP / TQC+ / CPE) users actually engage with helps decide where the next question-bank expansion goes. Judge feature value and retention. Tab dwell, widget use, and session length tell us which features earn their keep and which to cut. What we collect\r#\rEach signal carries an event name and a small structured payload:\nUser journey (funnel events, 100% uploaded)\napp_launched — launch type, app version, build, locale, device model onboarding_welcome_shown / onboarding_persona_picked / onboarding_completed — onboarding progress, including your chosen persona and certification preference (cert_type) practice_first_question_shown — first question ID and dimension practice_first_answer_submitted — first answer correct/wrong, dimension, thinking time session_ended — session length, questions answered, tabs visited, top tab Performance (perf events, 25% sampled)\nperf_cold_start_complete — cold-start ms and phase breakdown perf_tab_switched — tab-switch latency perf_llm_load_complete / perf_llm_first_token — LLM load and first-token latency, success, model_id perf_python_first_run — Python first-run warmup, script size, success Engagement (engagement events, 10% sampled)\npractice_question_answered — correctness, dimension, source, thinking time chat_message_sent — AI response ms, success (no prompt content sent) widget_deep_link_used — widget-tapped concept ID tab_appeared — which tab the user opened TelemetryDeck SDK default body: app version, build, device model (e.g. iPhone17,1), iOS major version, region + language, a one-way hashed identifier generated on-device by TelemetryDeck (per-vendor; cannot be reversed to your Apple ID or device UUID), and a session UUID.\nWhat we do NOT collect\r#\rYour name, email, advertising identifier (IDFA) Your raw IP address (TelemetryDeck briefly uses it at ingest to derive country, then discards) The content of your answers, your Playground code, your AI conversations, or your AI prompts Any data that can be linked back to your personal identity Sampling\r#\rFunnel events upload at 100%; performance events at 25%; engagement events at 10%. Total bandwidth per session is typically \u0026lt; 1 KB.\nApp Privacy mapping\r#\rCategory Subtype Purpose Usage Data Product Interaction App Functionality Diagnostics Performance Data App Functionality Identifiers Device ID (TelemetryDeck anonymous hash) App Functionality All three are declared Not Linked to Identity, Not used for Tracking.\nHow to opt out\r#\rTurn off tracking under iOS Settings → Privacy \u0026amp; Security → Tracking — this reduces the resolution of any identifier signal. An in-app toggle is under consideration for a future release. 6. Third-Party Services\r#\rThis app uses the following third-party services:\nService Purpose Data collected Provider region TelemetryDeck Anonymous usage analytics Anonymous events, device model, language, anonymous hashed identifier Germany (EU/GDPR) This app does NOT use: Google Analytics, Facebook SDK, Firebase Analytics, any advertising SDK, or any third-party crash-reporting service.\n7. Network Access\r#\rNetwork access is limited to:\nDownloading AI models (optional, one-time) — only when you explicitly choose to download LLM model resources (typically Hugging Face or similar public model hosts) Anonymous analytics upload (background) — small anonymous event packets sent over HTTPS to TelemetryDeck (see section 5) External links — opens your browser when you tap relevant links Beyond the above, the app does not initiate network connections. Code execution runs entirely in the local Python environment.\n8. Children\u0026rsquo;s Privacy\r#\rThis app is suitable for all ages and does not intentionally target analytics events at children under 13. TelemetryDeck signals are fully anonymous and cannot identify any specific user (including minors).\n9. Policy Changes\r#\rThis policy may be updated from time to time. Significant changes will be announced inside the app or by updating the \u0026ldquo;Last Updated\u0026rdquo; date on this page.\n10. Contact Us\r#\r📧 qqder339@gmail.com Subject: Python Dimensions Privacy Policy Inquiry\n","externalUrl":null,"permalink":"/privacy/python-dimensions/","section":"Privacy Policies","summary":"","title":"Python Dimensions — Privacy Policy","type":"privacy"},{"content":" FAQ\r#\rQ: The built-in Python runtime throws an error or crashes the app?\nA: Complex code (infinite loops, excessive memory usage) may cause timeouts or crashes. Make sure your code has no infinite loops and avoids allocating very large amounts of memory. If a specific code snippet causes a crash, please email us with that code.\nQ: Does the AI tutor need to download a model? How large?\nA: Yes. The first time you use the AI tutor, you choose and download a local model. Sizes range from a few hundred megabytes for the smallest model up to several gigabytes for the largest — pick what fits your device\u0026rsquo;s storage and memory. After downloading, the AI works completely offline — all Q\u0026amp;A and explanations run on-device without internet.\nQ: I think there\u0026rsquo;s an error in the question bank?\nA: If you find an incorrect question or answer, please email us with: the question content, your proposed correct answer, and your reasoning. We\u0026rsquo;ll verify and update the question bank as soon as possible.\nQ: The error radar chart isn\u0026rsquo;t showing?\nA: The radar chart requires a minimum number of answer records to generate. Please complete at least 20 questions first.\nQ: How do I use the code templates?\nA: Open the Playground tab. On iPad you\u0026rsquo;ll see a left sidebar with Templates at the top — tap any item (Hello World, for-loop, If-Else, function definition, list operations, etc.) to load it into the editor. On iPhone, scroll the Playground view and tap the Templates card to expand and pick one.\nTroubleshooting\r#\rPython runtime crashes: Ensure no infinite loops in your code; ensure the device has sufficient available memory AI model fails to load: Ensure 3+ GB free storage; retry downloading on Wi-Fi Force quit and relaunch the app Check iOS version ≥ 17.0 Contact Support\r#\r📧 qqder339@gmail.com\nSubject: [Python Dimensions] Issue Description\nPlease include: device model, iOS version, app version, steps to reproduce (include code if it\u0026rsquo;s a code-related issue).\nThis app collects no user data. Python execution and AI inference run entirely on-device.\n","externalUrl":null,"permalink":"/support/python-dimensions/","section":"Support","summary":"Support and contact for Python Dimensions","title":"Python Dimensions Support","type":"support"},{"content":"\rWhat I Do\r#\rI turn repetitive, rule-following computer work into machinery that runs by itself.\nThis is not just what I sell — I live on it. Before the market opens, a written morning briefing appears in my inbox; nobody writes it. My trade records check themselves against the broker every day, and only call me when something does not match. A batch of 500 stickers — drawn, background removed, packaged into store-ready files — came out of one run. This site exists in 11 languages; I only wrote the Chinese one. One person cannot do all of that by hand, so I had machines do it. Now I build the same thing on top of your workflow.\nIf a task is done by a person, on a schedule, by moving things between screens — it is probably automatable, and probably cheaper than you think.\nWhat I Build\r#\rYou already have customers; I turn the repetitive parts of serving them into machinery — faster delivery, lower labor cost.\nCategory Typical scenarios You receive Running cost I use it myself LINE automation Customer-service replies, group summaries, order/shipping notifications LINE Official Account bot, deployed, with an operating runbook Usage-based AI billing (on your own API key, transparent) My own LINE bot, live in production, tests green Reconciliation \u0026amp; reporting Orders vs. payments matching, daily revenue/inventory reports auto-mailed Scheduled scripts + plain-text config + loud error alerts $0/mo (pure scripts + scheduler, zero AI dependency) Daily reconciliation against my broker\u0026rsquo;s API Data movement \u0026amp; scheduling Scraping sites/back-offices, cross-system import/export, timed sync Scheduled scripts + plain-text config $0/mo Morning briefing pipeline, cross-machine config sync Document \u0026amp; email processing Email triage, invoice/PDF extraction, auto-filing and summaries Scripts + your own AI API key (setup guide, budget caps, usage alerts included) Usage-based, typically a few dollars/mo My own inbox and document flow AIGC content pipelines Batch product-image cutouts/composites, multilingual copy, sticker/graphic mass production A re-runnable pipeline + operating runbook From $0, usage-based if AI is attached A 500-item-per-batch sticker pipeline, an 11-locale site iOS app development (commissioned) Anyone who wants their idea shipped as their own app; students who need a published app on their resume (minors contract through a parent) A complete app on the App Store + full code ownership + submission walkthrough (delivery is counted at submission; App Store review timing is not mine to control) US$99/yr Apple Developer account (your own) 9 iOS apps of my own — built, shipped, operating Not sure which row your problem fits? Just describe it in writing — categorizing it is my job, not yours.\nHow It Works\r#\rYou write to me. One email describing the workflow: what happens today, which tools are involved, where it hurts. I reply with a fixed quote. Scope, deliverables, price, timeline. If I don\u0026rsquo;t think agents fit your problem, I\u0026rsquo;ll say so — and you\u0026rsquo;ve spent nothing. Build → document → hand off. You get a working system, a runbook your team can operate, and full ownership of the code. Designed so you don\u0026rsquo;t need me afterwards. What I Don\u0026rsquo;t Do\r#\rNo retainers, no on-call, no maintenance contracts — the handoff documentation exists precisely so I can leave. No open-ended \u0026ldquo;AI transformation\u0026rdquo; consulting. I automate specific, nameable workflows. Contact\r#\rqqder339@gmail.com — subject starting with [Automation].\nDescribe the workflow in writing; asynchronous by design. I reply within a week.\n","externalUrl":null,"permalink":"/services/","section":"QQder · The Miniature Boat","summary":"Fixed-scope agent automation — designed, built, documented, handed off.","title":"Services","type":"page"},{"content":"\rLeaving a person behind, not just a diary\r#\rMost recording tools deal with what happened today. Sown Echoes addresses a different scale of question. If a person\u0026rsquo;s values, experiences, preferences, tone, and ways of making judgments are worth preserving, how do you keep them? And how do you keep them as a structure that can be understood and conversed with again in the future, rather than as a pile of scattered notes?\nSo the app is part journal, part personal knowledge base, and carries a trace of digital legacy system. What you leave here goes beyond events: it includes how you see events, how you explain yourself, what you care about, and what you don\u0026rsquo;t. That\u0026rsquo;s the material Sown Echoes is really collecting.\nWhy it deserves to exist separately from notes or voice journals\r#\rThe reason it warrants its own category: it helps you progressively organise content into an analysable structure, rather than only storing it. Text, voice, questionnaires, persona summaries, values radar charts, digital-twin conversation. These modules form a complete chain. First record, then organise, then understand, then eventually converse.\nThis suits people who feel a strong urge to record, but also know pure notes tend to pile up into chaos. It\u0026rsquo;s a container designed for organising life material, not just another blank page that only handles input.\nKeeping both private and public tracks matters\r#\rMany products force you to choose between \u0026ldquo;fully private\u0026rdquo; and \u0026ldquo;fully social.\u0026rdquo; Sown Echoes takes the more mature position that both needs are legitimate. You can keep everything entirely on your own device and iCloud, or contribute selected content under an open licence as part of the broader Human Wisdom Library.\nThis dual-track choice is the product\u0026rsquo;s philosophy, not an add-on. Some parts of your life belong only to you; others might be worth entering public knowledge. Whether to share should be yours to decide.\nOn-device AI makes this feel less like surrendering yourself\r#\rWhen a product\u0026rsquo;s core is your thoughts, values, and life experience, privacy stops being a feature and becomes the precondition for the product to work at all. Sown Echoes keeps analysis and conversation as on-device as it can, so you don\u0026rsquo;t have to hand yourself over wholesale just to get a tool that understands you.\nOne day you\u0026rsquo;ll try to recall a chapter of your life and find you can\u0026rsquo;t quite put it into words anymore. Sown Echoes exists to push that day as far into the future as you can.\n","externalUrl":null,"permalink":"/apps/sown-echoes/","section":"Apps","summary":"","title":"Sown Echoes","type":"apps"},{"content":"Last Updated: 2026-05-07\n1. Overview\r#\rSown Echoes, developed by ChengChe Lee, is an app that lets you actively capture your thoughts, values, and life experiences, building a digital legacy through a BIP-39 cryptographic identity.\nIn short: We do NOT collect personal data and we do NOT send any of your content to ChengChe Lee-operated servers. By default your records only flow between your own device and your own iCloud account; content leaves that boundary only when you actively choose to publish an open contribution, in which case it is de-identified and published under a pseudonym in the Human Wisdom Library under CC-BY-SA 4.0.\n2. Data We Do NOT Collect\r#\rThis app does not collect:\nPersonally Identifiable Information (name, email, phone number) Location data Device identifiers Usage analytics or tracking data 3. Where Your Data Lives\r#\rOn your device (the source of truth for everything):\nBIP-39 Mnemonic: your Meme ID identity key Voice and text records: all thoughts, values, and stories captured through entries and questionnaires Speech-to-text results: Whisper or on-device Apple speech recognition output Vector index: on-device cache used for digital-twin semantic retrieval User settings: preference values In your own iCloud account (end-to-end encrypted; neither ChengChe Lee nor any third party can access):\niCloud Keychain: your BIP-39 mnemonic syncs across Apple devices signed in to the same Apple ID, so you can recover your identity when migrating devices CloudKit Private Database (via SwiftData): all records and questionnaire responses are automatically backed up to your private iCloud database You can disable either of these at any time in iOS Settings → Apple ID → iCloud.\n4. On-Device AI Features\r#\rAll AI inference runs on-device; content is never transmitted to any AI server:\nSpeech-to-text: choose between a local Whisper model or Apple\u0026rsquo;s built-in speech recognition (forced to requiresOnDeviceRecognition = true, i.e. on-device) AI analysis, persona summaries, and digital-twin conversation: a three-tier fallback — Apple Foundation Models (iOS 26+, OS-built-in) when available; otherwise a local MLX language model (one-time download of Qwen3 / Gemma-3 / SmolLM3 etc., ~0.4–3 GB, user-initiated); otherwise a basic stub mode BIP-39 identity generation: mnemonic generated locally on device, no external service Model files are fetched from huggingface.co only when you explicitly choose to download. After the download all AI features work offline.\n5. Data Export and Open Contributions\r#\rThe app provides two user-initiated ways to share your content:\nPersonal export (files on your device): under \u0026ldquo;Data Management\u0026rdquo; you can export personal backup (JSON), digital-twin data (JSON), and SFT/KTO training data (JSONL). Exports land as files on your device — you decide whether to share them. Open contribution (CC-BY-SA 4.0): when you mark an entry as \u0026ldquo;Open\u0026rdquo; and confirm it under \u0026ldquo;My Contributions\u0026rdquo;, the content is de-identified (PII stripped), uploaded to Apple CloudKit Public Database, and subsequently mirrored by ChengChe Lee to the public Hugging Face dataset qqder/memelives-open-contributions under CC-BY-SA 4.0 for research and AI training. You can withdraw an entry from within the app at any time (already-published snapshots remain under their CC-BY-SA license). 6. Third-Party Services\r#\rThis app does NOT use any third-party analytics or advertising frameworks (no Google Analytics, no Facebook SDK, no ads). The app\u0026rsquo;s only outbound dependencies are huggingface.co (for model downloads) and Apple iCloud (for the user\u0026rsquo;s own backup) — both connect only when you trigger them or when iCloud is enabled in iOS.\n7. Network Access\r#\rThe app initiates network connections in the following situations:\niCloud automatic sync (background, controlled by your iOS settings): syncs your mnemonic via iCloud Keychain and your records via CloudKit Private Database to your own iCloud account. Data is end-to-end encrypted and flows only between your Apple devices and your iCloud. You can disable this in iOS Settings. Downloading AI models (user-initiated): when you choose to download Whisper or MLX language models, the app fetches the files from huggingface.co. Open contribution upload (user-initiated): when you confirm an open contribution, the de-identified content is uploaded to Apple CloudKit Public Database (and subsequently mirrored offline by us to Hugging Face). See section 5. External links: tapping links inside the app opens the system browser. Outside of the cases above, the app does not initiate network connections.\n8. Contact Us\r#\r📧 qqder339@gmail.com\nSubject: Sown Echoes Privacy Policy Inquiry\n","externalUrl":null,"permalink":"/privacy/sown-echoes/","section":"Privacy Policies","summary":"","title":"Sown Echoes — Privacy Policy","type":"privacy"},{"content":" FAQ\r#\rQ: I forgot my BIP-39 mnemonic (Meme ID). Can I recover it?\nA: In most cases no. The mnemonic is only shown once at generation, and is stored on your device and in iCloud Keychain (which syncs automatically across Apple devices signed in to the same Apple ID). We cannot access your mnemonic and have no server-side backup. Strongly recommended: write it down or screenshot it immediately and store it somewhere safe — if you disable iCloud Keychain or lose access to all your Apple devices, the mnemonic cannot be recovered.\nQ: Does voice recognition (Whisper) require internet?\nA: No. Voice recognition uses a local on-device Whisper model. All recognition is processed entirely offline. On first use you download the model you select (base ≈ 150 MB, small ≈ 500 MB, large-v3-turbo ≈ 800 MB); after the download it works fully offline.\nQ: Where are my records? Can I export them?\nA: Records are stored on your device by default, and sync automatically to your own iCloud account via SwiftData ↔ CloudKit (we never see them). From \u0026ldquo;Data Management\u0026rdquo; in the app you can export several formats: personal backup (JSON), digital-twin (JSON), SFT training data (JSONL), and KTO training data (JSONL). Exports land as files on your device — you decide whether to share them.\nQ: Voice input recognition accuracy is low?\nA: Recognition accuracy depends on: background noise, clarity of speech, and language selection. Use in a quiet environment and ensure the app has microphone permission. If accuracy is particularly poor for a specific language, please email us.\nQ: Records disappeared after an update?\nA: Normal updates should not erase data. If data has disappeared, it may be due to accidental deletion or abnormal storage behavior. Please email us immediately with your app version information so we can help diagnose.\nTroubleshooting\r#\rVoice recognition fails: Ensure microphone permission is enabled (iOS Settings \u0026gt; Privacy \u0026gt; Microphone) Model download fails: Ensure stable Wi-Fi and sufficient device storage Force quit and relaunch the app Check iOS version ≥ 17.0 Contact Support\r#\r📧 qqder339@gmail.com\nSubject: [Sown Echoes] Issue Description\nPlease include: device model, iOS version, app version, issue description.\n⚠️ Important: Please keep your mnemonic (Meme ID) safe. It cannot be recovered if lost.\nThis app collects no user data. All content is processed entirely on-device.\n","externalUrl":null,"permalink":"/support/sown-echoes/","section":"Support","summary":"Support and contact for Sown Echoes","title":"Sown Echoes Support","type":"support"},{"content":"\rTurning a classical novel back into a world that runs\r#\rMost literary apps put the original text into a prettier reader. StoneStory takes a different path. It disassembles the characters, scenes, relationships, and events of Dream of the Red Chamber into a running narrative system. What you see is more than passages. You see how characters pull at each other, how they reveal personality inside situations, and how a classical novel operates like a small society.\nThat\u0026rsquo;s why it\u0026rsquo;s called a simulator rather than a reader. The goal is to take you into the structure beneath the text, not to serve up the same words in a nicer shell.\nWho actually needs a product like this\r#\rIf you already love Dream of the Red Chamber, what you\u0026rsquo;ll get here is a new way in rather than simple nostalgia. Characters become entities you can compare, observe, and reinterpret, rather than reference points to memorise. You\u0026rsquo;ll see more easily who truly made choices in which scenes, how emotions accumulated, and which details were already foreshadowing what came later.\nIf classical literature has always felt distant, the app may actually be easier to approach. It doesn\u0026rsquo;t demand that you first swallow the thick original to qualify; it breaks a complex work into a system you can approach slowly and understand incrementally.\nAI here opens the door to understanding\r#\rStoneStory uses AI for the work of comprehension, not plot generation: character inner life, emotional tension, modern-perspective interpretation, structural connections between events. For the user, this adds a layer of guided commentary: an interactive interpretation that shifts with each scene, rather than a dogmatic gloss.\nThe design suits Dream of the Red Chamber\u0026rsquo;s particular shape: many characters, complex relationships, extremely high detail density. You can let the system open the door, then decide how deep to go, instead of rebuilding the structure from scratch every time.\nWhy on-device AI still matters here\r#\rThis kind of product is easy to ship as a cloud demo. The moment content comprehension, reading history, and interaction all depend on external services, though, the experience becomes fragile, and it stops feeling like something that can accompany you long-term. StoneStory pushes the core experience back to the device, so immersive reading and exploration can actually exist as everyday tools, not as a technical showcase.\nIf you want the deeper technical and methodological thinking behind this, the related reading below goes there. If you\u0026rsquo;d rather start with the experience, the App Store is the most direct entry.\n","externalUrl":null,"permalink":"/apps/stonestory/","section":"Apps","summary":"","title":"StoneStory","type":"apps"},{"content":"Last Updated: 2026-04-15\n1. Overview\r#\rStoneStory, developed by ChengChe Lee, is an immersive reading and character simulation app based on the classic novel \u0026ldquo;Dream of the Red Chamber.\u0026rdquo;\nIn short: We do NOT collect, store, or transmit any of your personal data to external servers.\n2. Data We Do NOT Collect\r#\rThis app does not collect:\nPersonally Identifiable Information (name, email, phone number) Location data Device identifiers Usage analytics or tracking data 3. Locally Stored Data\r#\rThe following data is stored strictly on your device and never transmitted externally:\nTraveler Profile: The name, personality traits, speech style, background, and optional avatar image you configure in Traveler Mode Interface Preferences: Your selected display language (Traditional Chinese / English / Japanese) and chosen offline AI model Downloaded Content: Character portraits and scene images cached after you view them in the app Offline AI Model: The Qwen 2.5 model file you\u0026rsquo;ve chosen to download (stored in an App Group container for in-app use only) 4. Third-Party Services\r#\rThis app does NOT use any third-party analytics or advertising frameworks (No Google Analytics, No Facebook SDK, No Ads).\n5. Network Access\r#\rCore reading and simulation features work fully offline and require no network connection. The following features initiate network requests only when you explicitly trigger them:\nDownloading Character Portraits / Scene Images: The first time you view a character or scene, the app fetches the corresponding image from a public CDN and caches it locally Downloading the Offline AI Model: When you choose to download the Qwen2.5 model in Settings, the app fetches the file from the model\u0026rsquo;s public release source External Links: Opens the system browser when you tap relevant links These network requests transmit only the URL of the file you\u0026rsquo;ve chosen. No personally identifiable information is attached, and no data is collected in return.\n6. Contact Us\r#\r📧 qqder339@gmail.com\nSubject: StoneStory Privacy Policy Inquiry\n","externalUrl":null,"permalink":"/privacy/stonestory/","section":"Privacy Policies","summary":"","title":"StoneStory — Privacy Policy","type":"privacy"},{"content":" FAQ\r#\rQ: The app launches slowly or stalls on the splash screen.\nA: If you\u0026rsquo;ve downloaded the offline AI model (Standard 1.9 GB or High-Quality 4 GB), the app loads it into memory on launch; this may take several seconds on older devices. The first time you enter a chapter, the bundled database (characters, events, poems) is loaded — this is normal. If launch is unusually slow, please email us with your device model and iOS version.\nQ: Poetry or passages show garbled text, missing glyphs, or blank boxes.\nA: Three fonts are bundled in the app (LXGW WenKai TC, Noto Serif TC, Iansui) — no download or switching is required. If anomalies persist, please force-quit the app and relaunch, then send us a screenshot so we can fix it in the next release.\nQ: Is reading progress saved?\nA: In the current version (v1.1.1), chapter playback progress is retained only within the current app session — you can return to a chapter during the same session. However, if you force-quit the app or restart your device, chapter playback will start from the beginning. Persistent cross-session bookmarks are planned for a future release.\nQ: Character portraits or scene images won\u0026rsquo;t load.\nA: Portraits and scene images are downloaded on-demand the first time you view them, and cached locally. If they won\u0026rsquo;t load:\nVerify your network connection Swipe away from the screen and return to trigger a retry Or go to Settings → Clear Art Cache and re-enter with a stable network Q: The offline AI chat doesn\u0026rsquo;t respond.\nA: First-time use requires downloading a Qwen 2.5 model under Settings → Model Management. Choose one based on your device:\nSmall 1.5B (~0.9 GB) — entry-level Apple Silicon devices Standard 3B (~1.9 GB, default) — Pro-class devices High-Quality 7B (~4.0 GB) — current-generation Pro / M-series devices Ensure sufficient free space on your device. Once downloaded, the chat runs fully offline.\nQ: Can the app be used offline?\nA: Yes. Chapter playback, True Endings, personality system, poem/object collections, and on-device AI chat (after model download) all work offline. Only character portraits, scene images, and the AI model file require an internet connection on first retrieval.\nTroubleshooting\r#\rForce-quit and relaunch the app Check iOS version ≥ 17.0 If a specific chapter misbehaves, note its name and email us Uninstall and reinstall (your Traveler profile and downloaded images will be cleared) Contact Support\r#\r📧 qqder339@gmail.com\nSubject: [StoneStory] Issue Description\nPlease include: device model, iOS version, app version, steps to reproduce (screenshots preferred).\nThis app collects no user data. All content is stored locally on your device.\n","externalUrl":null,"permalink":"/support/stonestory/","section":"Support","summary":"Support and contact for StoneStory","title":"StoneStory Support","type":"support"},{"content":"","externalUrl":null,"permalink":"/support/","section":"Support","summary":"Support pages for all apps","title":"Support","type":"support"}]