The comic is my silly take on putting vibe-based-work everywhere. What if payroll ran with vibes as engineers are.

Vibe Coding—the practice of using AI to generate code based on natural language prompts—has been a hot topic in tech circles lately. While it offers rapid prototyping capabilities, it’s essential to understand both its advantages and limitations. Personally, I agree with Ben South’s post on X: “Vibe coding is all fun and games until you have to vibe debug”.

Here are a few thoughts 👇

  • 🧑‍💻 Treat the AI like a junior dev: Let it draft, but insist on code reviews, tests, and refactors—otherwise, technical debt snowballs fast.
  • 🔐 Security is not a vibe: Your LLM can be confidently wrong. Generated code can hide nasty vulnerabilities—automate static analysis, secret scanning, and dependency checks from Day 1.
  • ⚡ Creative speed ≠ long-term maintainability: Rapid MVPs look great in demos, yet lack longitudinal data on bug costs and total cost of ownership. Refactoring entire codebases presents its own challenges of dropped logic and needle-in-a-haystack issues.
  • 📉📈 Career impact is polarized: Entry-level postings are shrinking, yet senior engineers mastering AI-assisted workflows are reporting real productivity gains. (Refer to the Jedi chart.)
  • 🧠 Upskill in prompt design + architecture: The safest seats combine system thinking and AI fluency—be the person who can steer and debug the machine.

🐑 “AI Shepherd” might be a job requirement.