<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by 雍晨 Chen Yung on Medium]]></title>
        <description><![CDATA[Stories by 雍晨 Chen Yung on Medium]]></description>
        <link>https://medium.com/@cyung.tw?source=rss-626888384c2e------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*v_AioIdusW8L507RvQHy_g.jpeg</url>
            <title>Stories by 雍晨 Chen Yung on Medium</title>
            <link>https://medium.com/@cyung.tw?source=rss-626888384c2e------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 09:58:34 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@cyung.tw/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[My first vibe coding — Map cleaner]]></title>
            <link>https://medium.com/@cyung.tw/my-first-vibe-coding-map-cleaner-5c0a1f6e932b?source=rss-626888384c2e------2</link>
            <guid isPermaLink="false">https://medium.com/p/5c0a1f6e932b</guid>
            <dc:creator><![CDATA[雍晨 Chen Yung]]></dc:creator>
            <pubDate>Fri, 09 Jan 2026 16:23:34 GMT</pubDate>
            <atom:updated>2026-01-10T11:13:02.432Z</atom:updated>
            <content:encoded><![CDATA[<h3>My first vibe coding — Map cleaner</h3><p><a href="#cee3">中文請點擊或往下滑</a></p><p>My google maps is full of pins, and most of them are restaurants or bistros that I want to visit at some point in the future. For the past few years, more and more of these pinned places have shown the red-marked “permanently closed”, and some even become pins with mere coordinates. (I don’t like how they are often called “ghost pins”.)</p><p>Old business disappearing and new ones emerging is natural. Sadly, I think most of these closed business may have gone through such difficult times as pandemic or inflation. While looking at these pins, I thought it may be a good idea to commemorate how many places and which places I had only the luck to know their names, but never got the chance to know them better. I turned to my Google Gemini.</p><p>Before this vibe coding practice, I have had only limited programming skills. I had used Matlab intensively for a few months around 10 years ago when I was working on my thesis. And only last year, 2025, I joined the Code in Place program provided by Standford University that covers the fundamentals of Python. I am curious to know how much I can achieve by using vibe coding without knowing much more than these past experiences.</p><p>While I am aware that there are many better ways to prompt, or even ask better questions so that the entire process can be automated, I used extremely plain English to start my project with: “I am looking at my google maps with all the pins with ‘want to go’, and some places are already ‘permanently closed’. How to write a script to list down all the permanently closed places that I pinned, so that I can unpin them on my google maps?”</p><p>Gemini showed me platforms or products that I had never heard of before, and the goal can be achieved in three main steps.</p><ol><li>Export the “want to go” list from Google Takeout, and save in .json format.</li><li>Set up a Google Cloud API key, and restrict it.</li><li>Run the Python Script that it provides, which exports the results into a CSV file, then open the CSV file with Excel or Google Sheets.</li></ol><p>Whenever I needed an elaboration on certain jargon or steps, Gemini broke down the process, and used analogies to help me understand.</p><p>Through back and forth debugging and discussion, using up two sessions of Gemini 3 thinking mode, I achieved the goal. The results showed that out of 858 pinned places, 30 are permanently closed. And I am still pondering if I want to unpin them as if they never existed. :(</p><p>Below shows my lessons to take away.</p><ol><li>Source data</li></ol><p>In Google’s data system, want-to-go lists are treated differently than starred or labelled places. Thus, in Google Takeout,</p><ul><li><em>want-to-go</em> list is derived from the category named “Saved”, and is saved directly in a CSV file.</li><li><em>starred</em> spots are derived from the category “Maps (Your Places)”, and are saved in a JSON file.</li><li><em>labelled </em>places, such as my home and work address, are derived from the category “Map”, and are saved in a JSON file.</li></ul><p>2. Place ID</p><p>Initially, there weren’t any places being found as permanently closed. I figured that the code starts with searching by coordinates. But this ran into one problem: a new restaurant is open exactly where an old one closed. I told Gemini about my concern, and it provided new code that starts with searching by Place ID.</p><p>3. Language</p><ul><li>Perhaps because I chat typically with Gemini in English, Gemini initially didn’t take into account those pins in my native tongue, Mandarin Chinese. And Taiwanese bistros sometimes can have weird names that is a mix of Chinese characters with number digits or with English. And in this case, the standard text formats can break. Thus, the encoding of <em>utf-8-sig</em>, instead of <em>utf-8 </em>is crucial in global data, as the “sig” renders the special characters for the software.</li><li>I noticed that the CSV files downloaded from my Google Maps are named as “<em>labelled</em> places.csv” as in British English, whereas Gemini chatted with me and provided code using the American “<em>labeled</em> places”. Funny, yet of course, Gemini fixed it in no time.</li></ul><p>I had a lot of fun! And more importantly, I re-experienced moments of flow that had disappeared for a long time. I am really happy. :)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EgJ6iV2WARjumLEfkDN68g.jpeg" /></figure><p>我的 Google 地圖上佈滿了圖釘，其中大多數是我希望某天能造訪的餐廳或小館子。過去幾年，這些標記的地點中，越來越多顯示著大大的紅字「永久停業」，有些甚至僅剩下一個沒有名字的座標。（不是很喜歡稱它們為「幽靈圖釘」。）</p><p>舊店歇業、新店崛起是滿自然的。但令我有些難過的是，這些歇業的店家大多可能經歷了疫情或通膨這些非常艱難的時刻。看到這些圖釘，我突然覺得或許可以找個方式來紀念一下那些我有幸得知其名，卻遺憾沒能造訪的地方。於是，我打開我的 Google Gemini。</p><p>在開始這次 vibe coding 之前，我的寫碼技能非常有限。頂多就大約 10 年前為了寫論文，大量寫過幾個月的 Matlab；以及去年 2025，我才參加史丹佛大學辦的 Code in Place 計畫，學習 Python 基礎。我很好奇，憑這點經驗，不具備更多專業知識的情況下，光靠 vibe coding 能達到什麼境界。</p><p>雖然我知道有很多更好的下 prompt 指令的技巧，甚至能問更精準的問題來讓整個過程全自動化，但我選擇用最直白、通俗的英文開始：</p><blockquote><em>「我正在看我 Google 地圖上標記「想去」的地點，有些地方已經「永久停業」了。該如何寫個程式，列出所有我標記過但已永久停業的地點，好讓我能取消標記？」</em></blockquote><p>Gemini 寫了些我從沒用過的平台與產品，並透過三個主要步驟達成目標：</p><ol><li>從 Google Takeout 匯出「想去的地點」清單，並儲存為 .json 格式。</li><li>設定 Google Cloud API 金鑰，並將其設定為限制金鑰。</li><li>執行它提供的 Python 程式，將結果匯出為 CSV 檔，最後用 Excel 或 Google 試算表打開。</li></ol><p>每當我需要某些專有名詞或步驟的詳細說明，Gemini 都能拆解流程，並用類比法來幫我理解。</p><p>經過來回的除錯（debug）和討論，以及耗盡兩次 Gemini 3 思考模式的額度，最終達到了目標。結果顯示在 858 個標記地點中，有 30 個地方已經永久停業了。我仍在想，是否真的要取消標記，讓它們像從未存在過一樣。:(</p><p>以下是我學到的幾件事：</p><p><strong>1. 原始資料來源 (Source data)</strong></p><p>在 Google 數據系統中，「想去的地點」與「已加星號」或「已加上標籤」的地點處理方式不同。因此，在 Google Takeout 中：</p><ul><li>「想去的地點」來自名為「已儲存」的類別，直接存成 CSV 檔。</li><li>「已加星號的地點」來自「地圖（您的地點）」類別，存成 JSON 檔。</li><li>「已加上標籤的地點」（例如我的住家與公司地址）來自「地圖」類別，存成 JSON 檔。</li></ul><p><strong>2. 地點 ID (Place ID)</strong></p><p>最初程式完全找不到任何永久停業的地點。我發現是因為程式優先以「座標」搜尋，但這遇到一個問題：有些新餐廳正好開在舊店歇業的原址。我向 Gemini 說了這個疑慮，它提供了新的程式後，改以「地點 ID (Place ID)」作為搜尋基礎。</p><p><strong>3. 語言</strong></p><ul><li>可能是因為我平常是用英文跟 Gemini 對話，它最初沒有考慮那些以我的母語 — — 繁體中文標記的圖釘。台灣有些餐館名稱很特別，會混合中文、數字或英文，在這種情況下，標準文字格式很可能會出錯，變成亂碼。因此，在處理全局 (?)(global) 數據時，要使用 utf-8-sig 而非一般的 utf-8 編碼，因為 sig 能讓軟體正確輸入特殊字元。</li><li>我也注意到，從 Google 地圖下載的 CSV 檔名是英式拼法 <em>labelled</em> places.csv，而 Gemini 跟我聊天和寫碼則用美式拼法 <em>labeled</em> places。有點好笑，但Gemini 當然很快就修正掉。</li></ul><p>我很開心！更重要的是，我從中回到消失已久的心流。我真的很開心。:)</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5c0a1f6e932b" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>