An approach to building software where you describe what you want in plain language and let an AI model write, iterate, and debug the code — operating on intent and feel rather than technical specification.
Also known as natural language programming, AI-assisted development
Vibe coding is not traditional software development. It is closer to collaborative prototyping with an AI that happens to speak code. You describe what you want to build in plain language, iteratively, the way you would explain it to a capable colleague — and the AI writes, adjusts, and debugs the code on your behalf. You don’t need to understand the code. You need to understand what you want it to do.
The term was coined by AI researcher Andrej Karpathy in early 2025 and spread quickly as AI coding tools made it genuinely viable for non-engineers to produce working software. It is most useful for internal tools, rapid prototypes, and one-off solutions where the output needs to work rather than needs to be maintained by a large engineering team.
The walls between “creative” and “technical” inside agencies are dissolving, and vibe coding is part of why. Strategists are building data dashboards. Copywriters are prototyping interactive content. Operations teams are spinning up internal tools without filing tickets to engineering. The constraint used to be access to developer time. Vibe coding changes that equation.
The governance question is real. What do you do with vibe-coded tools that are running in production but that no one on the team fully understands? What is the agency’s policy on AI-generated code in client deliverables? These are operational questions that agencies need answers to before they are in the middle of an incident.
The literacy requirement doesn’t disappear. Vibe coding removes the need to write code. It does not remove the need to understand what you are building, recognize when something is wrong, and know when a problem has outgrown what vibe coding can safely handle.
An account director needs a simple tool that takes a client’s weekly ad spend data, compares it to the prior week, and sends an email summary with flagged variances. The task would typically require a developer and a two-week queue. Instead, she spends an afternoon describing the tool she wants to an AI coding assistant, iterating on the output, and testing it against sample data. By end of day she has a working version. The agency’s IT team reviews it before it touches live client data — the right governance checkpoint for this kind of workflow. The developer queue is reserved for things that actually need a developer.
The automations and agents module of the workshop teaches you how to build AI workflows that compress the busywork without taking the craft out of the studio.