What the Hell is Vibe Coding?

In this episode of How I Met Your Data: The Prompt, Anjali and Karen dig into one of the fastest-emerging patterns in development today: vibe coding - the practice of describing what you want and letting an LLM generate the code. It’s new. It’s evolving. And right now, it’s causing as much frustration as it is excitement.

Karen breaks down what vibe coding actually looks like in practice: developers prompting AI to produce entire features or files, navigating the wildly different “personalities” of today’s LLMs, and learning how to guide systems that might generate brilliant structure… or unintended chaos. Together, they talk through the real friction points - overly eager model behavior, unexpected file changes, incomplete suggestions, and the creeping loss of hands-on debugging skills that used to tie engineers closer to their code.

But underneath the surface is a bigger enterprise theme. The rise of vibe coding speaks to deeper issues: end users who still aren’t getting what they need, bottlenecks in IT and data teams, and the rapid expansion of citizen development as people search for faster paths to outcomes. Anjali and Karen unpack the operational and governance implications, from maintainability and handoff challenges to compliance blind spots and the need for standards that can coexist with AI-assisted creation.

They also dive into where AI does shine today - those repetitive, operational workflows that quietly save teams hours - and why focusing on value, ownership, and workflow design matters far more than chasing the next flashy LLM demo.

This episode is an honest, grounded look at how AI-assisted development is taking shape: what’s promising, what’s painful, and what it means for teams trying to build responsibly, collaboratively, and at scale.