Vibe Coding: What It Gets Right and Where It Bites

Vibe coding is the practice of building software by describing what you want in plain language and accepting what the model gives back — without necessarily reading or fully understanding every line. It's exhilarating, it's genuinely productive, and it's responsible for a wave of apps that would never have existed otherwise. It's also responsible for a wave of apps nobody can safely maintain.

Where vibe coding shines

For prototypes, throwaway tools, and validating an idea before you commit, vibe coding is a superpower. You can go from "what if" to a clickable thing in an afternoon. When the cost of being wrong is low and the goal is to learn fast, not reading every line is a feature, not a bug. I vibe code internal scripts and demos constantly and feel zero guilt.

Where it bites

The trouble starts when the prototype quietly becomes the product. Code you don't understand is code you can't debug at 2am, can't secure with confidence, and can't extend without breaking. Vibe-coded systems tend to accumulate invisible assumptions — an auth check that only mostly works, a query that's fine until you have 10,000 rows. The bill comes due exactly when the stakes are highest.

The rule I follow

The moment something is going to be maintained, secured, or scaled, the vibe ends and engineering begins. That doesn't mean throwing the AI out — it means reading every line, writing real tests, and understanding the system well enough to defend it. Vibe to discover; engineer to ship. The skill isn't choosing one camp. It's knowing exactly which line you're standing on.