Hot take: the "vibe coding" trend is creating a generation of developers who can't debug their own code.
What I'm Seeing
I mentor junior developers at a startup. In the last 6 months, I've noticed a pattern:
- Junior dev gets stuck on a feature
- They paste the requirement into Claude/GPT
- They get working code back
- They ship it without understanding it
- It breaks in production
- They can't fix it because they don't understand what they shipped
This isn't a hypothetical. I've seen it happen 4 times in the last month.
The Problem Isn't AI
AI coding tools are incredible. I use them every day. The problem is using them as a replacement for understanding instead of an accelerator.
When I use an AI agent, I:
- Read every line it generates
- Understand why it made each choice
- Modify the output to fit our patterns
- Write the tests myself (usually)
When vibe coders use AI:
- Copy entire output
- Run it
- If it works, ship it
- If it doesn't, paste the error back into AI
The Real Cost
Vibe-coded systems are expensive to maintain. The code works but nobody on the team truly understands it. When it breaks at 2am, you can't "vibe" your way to a fix.
I've seen a 2000-line module that no human on the team could explain. It worked perfectly for 3 months. Then an API changed and it took 4 days to fix something that should have taken 2 hours.
What We Should Do Instead
- Use AI as a tutor, not a ghost writer. Ask it to explain its code before you use it.
- Write the hard parts yourself. Use AI for boilerplate, not for core logic.
- If you can't explain it, don't ship it. Simple rule that prevents 90% of the problems.
AI will make great developers better. But it won't make non-developers into great developers. There's no shortcut to understanding.
Agree? Disagree? I want to hear from both sides.
Top comments (0)