The Journey • Vibe Coding

Week 1: The Halloween Spark, Paper Dolls, and Sounding Like a Scammer

This whole project started because of a Halloween costume.

I was browsing Taobao, endlessly scrolling, unable to picture how different pieces would actually look together. I was struck by a random, chaotic thought: Can I just use Claude to code an app that tests these outfits for me?

When I pitched this to Gigi, a Gen-Z coworker who lives on Taobao, she laughed. Building software usually starts with a grand master plan. Mine started because I was frustrated by a skirt.

Dressing for "The Scene"

Here is my philosophy on fashion: it's not just about whether a piece of clothing fits your body or matches a specific "style." I think, we should all "Dress for occasion". So, it is about The Scene.

With my background in landscape architecture, I know that you don't just pick beautiful outdoor furniture and drop it anywhere; you pick the right lounge chair that works within a specific resort pool; you pick the right bench that fits in a specific park. Clothes are exactly the same. They must complement the holistic environment. If you want to know if an outfit works, you don't test it against a plain white background. You test it against the bamboo forest path you are going to walk down.

Outfit Moodboard
My mood board when I need to choose the outfit for the Arashiyama bamboo forest. Testing the "vibe" against the actual scene
Final Outfit Result
The final result: How to stand out in a photo while still blending perfectly into the environment

David vs. Goliath (The Paper Doll Pivot)

As I started "vibe coding" with Claude, we hit a wall, many times. Especially when I looked at Taobao's built-in AI fitting room, and it was terrifyingly good. It maps clothes onto 3D models in seconds. As a solo vibe-coder, I cannot out-engineer a billion-dollar tech giant on hyper-realism.

So, I didn't try. Instead of fighting a losing battle, I pivoted. I realized that in a world flooded with flawless, sterile AI generation, people actually crave nostalgia. I decided to lean into the "flaw" and aim for a handmade, scrapbook, "paper doll" aesthetic. It doesn't need to fit perfectly; it needs to feel personal.

The "Scammer" Reality Check

Once I had the prototype, I needed testers. I messaged my girls group in WhatsApp, asking for their height and swimsuit photos to generate their avatars. Their response? Absolute silence.

Suddenly, I had a massive realization. I sounded exactly like a creepy scammer, or like my brother, a tech founder, when he randomly asks me to test something out of the blue. I was so high on the project I am working on that I forgot about human psychology. This was a masterclass in UX and Friction Empathy: you cannot just ask users for sensitive data without warming them up and building trust first.

Next week, I am moving on to my real target audience: Gigi and my students. (Beta testers who actually need the product are far more valuable than friends who are just being polite).

P.S. Claude told me "Fitting Room" is too generic a name for this app. Damn it. Now I have to think of a better name!

第一週:萬聖節、紙娃娃、詐騙集團

這整個專案的起點,其實是因為一套萬聖節服裝。

萬聖節是我每年最重視的節日,因為我每年都會辦萬聖節派對,且會有個特別主題,大家得針對這主題準備。我通常都得提早兩個月開始準備,因為我得上淘寶找到適合的衣物,自己搭配組合。當時我正在淘寶上無止盡地滑,完全無法想像這些不同的單品搭在一起會是什麼樣子。突然間,我腦中閃過一個瘋狂的念頭:「我能不能乾脆叫 Claude 幫我寫一個可以測試這些穿搭的 App?」

當我把這個想法告訴我的同事 Gigi(一個什麼都網購的Z世代)時,她大笑了出來。一般來說,開發軟體都需要一個宏大的計畫,而我的起點只是因為一條裙子讓我傷透腦筋。

為「場景」穿搭

我對時尚的哲學是這樣的:這不只是關於這件衣服合不合身,或是它符不符合某種「風格」。我覺得,我們都應該要「針對場合穿搭」,也就是說,穿搭最重要的是「場景」(The Scene)。

因為我有景觀建築的背景,所以我很清楚,你不能只是隨便挑一張漂亮的躺椅到某個高級渡假村泳池池畔,你也不能隨便放一張長椅到某個公園;你必須挑一張能完美融入那個場景的家具。衣服也是一樣的道理。它們必須與整體的環境相輔相成。如果你想知道一套衣服適不適合,你不能只把它放在純白背景前測試,你必須把它放在你即將走過的那條竹林小徑上來測試。

Outfit Moodboard
當我需要為嵐山竹林挑選穿搭時製作的 Mood Board(情緒板)。直接在真實場景中測試服裝的「氛圍」
Final Outfit Result
最終成果:如何讓自己在照片中脫穎而出,卻又完美地融入環境之中

小蝦米對抗大鯨魚(紙娃娃的轉向)

當我開始和 Claude 進行「語感寫程式(vibe coding)」時,我們撞牆了。且撞超多次的。尤其在我看了淘寶內建的 AI 試衣間的時候。它的技術強大到令人毛骨悚然,能在幾秒鐘內將衣服完美貼合在 3D 模特兒身上。作為一個單打獨鬥,沒資本,沒經驗的vibe-coder,我根本不可能在「超寫實」這條賽道上贏過那些砸了幾十億的科技巨頭。

所以,我放棄這個路線。與其打一場必輸的仗或是放棄不做了,我選擇了轉向(Pivot)。我意識到,在這個充斥著完美、冷冰冰 AI 生成圖像的世界裡,人們其實渴望某種復古與懷舊感。我決定把這個「缺點」放大,將風格轉向一種手工的、剪貼簿式的「紙娃娃」美學。它不需要完美貼合,它需要的是溫度與個人特色。

「詐騙集團」的震撼教育

原型做出來後,我需要測試者。我在好姊妹的 WhatsApp 群組裡傳了訊息,要她們給我身高和泳裝照,好讓我生成她們的數位分身。結果呢?一片死寂。

那一瞬間我突然恍然大悟。我聽起來完全就像個變態詐騙集團,或是像我哥 (某個科技創辦人) 突然傳訊息叫我幫他測試什麼奇怪東西一樣。我對這個設計專案太興奮,以至於完全忽略了人類心理學。這是一堂最真實的 UX(使用者體驗)與「阻力同理心」(Friction Empathy) 的專業課程:你不能在沒有建立信任、沒有任何鋪陳的情況下,就直接跟使用者討要敏感資料。

下週,我將把目標轉向我真正的受眾:Gigi 和我的學生們 (如果我能說服她們的話)(真正需要這項產品的 Beta 測試者,絕對比只是出於禮貌幫忙的朋友有價值多了)。

P.S. Claude 跟我說「Fitting Room(試衣間)」這個名字太普通了。可惡。現在我得想個更好的名字了!