框架這一週,從一句話到可以開始寫
一週把站蓋起來。真正花時間的不是版面,是想清楚內容要怎麼產出。
AbstractA week of scaffolding. The layout was the easy part — the real work was designing how the content gets produced.
一個人的開發現場紀錄。依時間切片,寫下每個專案當時在想什麼、哪些概念實驗留了下來、哪些做到一半被自己否決。
一週把站蓋起來。真正花時間的不是版面,是想清楚內容要怎麼產出。
AbstractA week of scaffolding. The layout was the easy part — the real work was designing how the content gets produced.
從 commit 生成日誌這件事我試了,結果是一坨流水帳。改成讓工具生成「該問的問題」。
AbstractAuto-generating posts from commits produced filler. The tool now generates recall questions instead, and the writing happens in a Q&A pass.
色票整組從 super-reversi2 的宣紙主題搬過來,實算對比度之後發現有三顆在長文站上不能直接用。
AbstractReused the ink-wash palette from super-reversi2, but three values failed WCAG contrast once measured against every paper surface.
一個 repo 寫一篇總結,聽起來很乾淨,實際上會變成一篇沒人讀得完、也沒人寫得出來的東西。
AbstractThe "one post per repo" idea sounded clean but collapses under real history — replaced by calendar-based slicing.
想把「為什麼這樣改」寫下來,可是那些理由早就忘了,剩下的只有一堆寫著 fix 的 commit。
AbstractStarting a devlog because the reasoning behind past decisions lives nowhere — only in commit messages that mostly say "fix".
截了一張 TikTok 直播打禮物 PK 的圖丟給 AI 說照著做,它做出來了。但我要的是規則上的二重判斷,拿到的是把棋子數量畫成一條 bar。
AbstractHanded the AI a screenshot of a TikTok gift-PK bar and asked it to clone the thing. It did. But what I wanted was a second win condition in the rules, and what came back was piece count drawn as a bar.
第一次用 spec-kit,一個 commit 進來 87 個檔案兩萬三千行。接下來四天就是住進那包程式碼裡,決定哪些要留。
AbstractFirst time using spec-kit. One commit dropped 87 files and 23,366 lines, and the next four days were spent living inside that generated code deciding what to keep.