0xarthur|11月 15, 2026 13:18
In the Web 1 era, liberal arts graduates were initially in high demand.
Back then, many portal websites needed people constantly writing numerous articles.
The AI era is similar; people from all industries can find their place here.
However, being able to write prompts doesn't necessarily mean you can vibe and create a good product.
I've seen many people's vibe creations, some of which are so impressive I'd want to move to Bengbu (a famous Chinese electronics market in China).
I've seen someone define the ID field of a log database as an integer. This is a log database! Normally, it should be defined as a bigint field with an index, and historical data should be backed up and cleared periodically.
I suspect he was just telling the AI: "Help me with the log entries so I can query past records." Then the AI rattled off a bunch of technical jargon, which he didn't understand, and just said: "You figure it out, handle the rest yourself, and give me the product."
These vibe-coded products are mostly suitable for personal use and difficult to scale up.
Developers, on the other hand, use AI primarily for processing within specific tasks.
We break down a project into numerous tasks, write specifications, and then review them using a combination of human and AI. Indeed, the number of coding steps decreases, allowing more time to be spent on the technology stack and business logic.
For example, consider a common file upload function. An average user might tell the AI, "Write me a file upload function."
But we would say, "I need you to write a FastAPI backend for file uploads. Files are uploaded to Pinata, and my Pinata key is XXXXX. During upload, the files need to be AES encrypted. Create a 'file' table in the database to record the file's gateway, upload timestamp, uploader, hash value, etc. Please provide a detailed implementation plan first, then I will modify it, and only after my confirmation will it be executed."
Share To
Timeline
HotFlash
APP
X
Telegram
CopyLink