因為ai社課的原因，我想要用v0這個vercel開發的ai coding平台，來建立一個可以部署上vercel的服務。如果我可以成功，學生也一定能成功，我原本是這樣想的。沒想到，因為我前幾個版本，因為我的prompt寫得不夠好，加上v0本身的智商限制（例如他常常漏東漏西或者是根本沒看懂我給他的spec導致出現幻覺亂作一通），所以一直都沒有辦法在v0給的5美金quota裡面搞定。
最後，靠我的好夥伴 Chuan-Chi Hsu 花了重本，探索v0的邊界，最後才寫出一個可以一步（喔不，最後還是靠了4步但至少做出來了）到位的prompt，讓v0 ai可以在quota限制內寫出一個想要的rag 平台，給學生參考練習。
很多人說現在是 Vibe Coding 的時代，只要有手、有 AI，什麼都能蓋出來。但作為一個正在裡面掙扎的開發者，我的 Hot Take 可能會讓你冷靜一點：
你對資料工程（Data Engineering）越一無所知，你就要花越多錢。 AI 可以幫你寫 Code，但它沒辦法幫你補齊底層邏輯的空缺。如果你不懂 Data Pipeline、不懂向量檢索的原理，你只會在錯誤的邏輯裡反覆消耗 Token。
你的需求 Spec 寫得越不清楚，你就要花越多時間 Debug。 「幫我做一個 RAG 平台」這句話在 v0 面前價值趨近於零。模糊的指令只會換來模糊的實作。最後你會發現，花在精煉 Prompt 的時間，其實就是以前在寫 Spec 的時間。
v0 是個還不錯的工具，但務必要寫非常清楚的 Prompt。 它確實能讓你無痛接上 Vercel 快速部署，這點非常強大。但如果你沒辦法給出極度精確的約束條件，你不是在開發，你是在跟你的信用卡過不去（或者像我一樣，在跟 Gmail 帳號的數量過不去）。
🛠️ 實作成果： 在嘗試到第六個免費帳號後，我終於在額度內成功搭建了這個 RAG 平台的 Demo App。這是一個從慘痛經驗中萃取出來的產物。
因為這個平台問問題還是會花我的錢，所以如果大家想要玩玩看的，只能夠到下週的AI社社課（5/4, 19:00-21:00, 博雅505） 來試用！
⚠️ 重要提醒： 由於這是使用免費帳號搭建的 Demo，你上傳的資料其他人都會看到！ 請當作測試場域就好，千萬、絕對、務必不要上傳任何敏感或機密資料！
🚀 The Price of #VibeCoding: It Cost Me 6 Gmail Accounts to Build This RAG Platform
For our AI club course, I wanted to use v0 (the AI coding platform by Vercel) to build a service that could be deployed seamlessly. My initial thought was: “If I can do it, my students definitely can too.”
I was wrong.
Because my early prompts weren’t sharp enough—and because of v0’s inherent “intelligence gaps” (it often misses details or completely fails to grasp the spec, leading to hallucinations)—I burned through account after account without ever finishing within the $5 free quota.
Eventually, thanks to my partner Chuan-Chi Hsu, who “invested” heavily to explore the boundaries of v0, we finally nailed a prompt that worked in one shot (well, technically 4 steps, but close enough). This allowed the AI to build a functional RAG platform within the quota limits for student practice.
Everyone says we are in the era of Vibe Coding—that as long as you have hands and an AI, you can build anything. But as a developer currently in the trenches, here is a Hot Take to bring you back to reality:
1️⃣ The less you know about Data Engineering, the more you’ll pay. AI can write code, but it cannot bridge the gaps in your underlying logic. If you don’t understand Data Pipelines or the principles of Vector Retrieval, you’ll just end up burning tokens on broken logic.
2️⃣ The vaguer your Spec, the more time you’ll waste debugging. The phrase “Build me a RAG platform” is worth exactly zero to v0. Vague instructions lead to vague implementations. Eventually, you’ll realize that the time spent refining your prompt is exactly the time you used to spend writing technical specs.
3️⃣ v0 is a great tool, but your prompts MUST be crystal clear. Its ability to deploy seamlessly to Vercel is a superpower. However, if you can’t provide extremely precise constraints, you aren’t “developing”—you’re just picking a fight with your credit card (or, in my case, the number of Gmail accounts you own).
🛠️ The Result: After my sixth Gmail account, I finally successfully built this RAG platform demo within the free tier. This is a product forged in the fires of trial and error.
💸 Limited Testing: Since running queries on this platform costs actual money, it will only be available for live testing during next week’s AI Club session (May 4, 19:00-21:00, Boya 505)!
⚠️ CRITICAL WARNING: Because this is a demo built on a free account, any data you upload will be visible to others! Please use it for testing purposes only. DO NOT upload any sensitive or confidential information!
#DataEngineering #RAG #v0 #Vercel #LLM #AIPrompting #EngineeringMindset #VibeCoding #NTUAI