IsVibeCoding > Coding?

Welcome back to TheComplexWorldOfAI. Today we're moving away from Complexity SImplified for a bit and talking about something that's been all over tech Twitter, startup blogs, and even now in the dictionary! Collins English Dictionary named it Word of the Year for 2025. Merriam-Webster listed it as slang and trending. It's called vibe coding, and it's one of the most interesting shifts happening in tech right now.

So what is it, why does it matter, and should you care? Let's break it down.

FIRST, WHAT EVEN IS VIBE CODING?

The term was coined in early 2025 by Andrej Karpathy, a co-founder of OpenAI and former AI lead at Tesla, one of the most respected names in the field. He described it as a style of building software where you "fully give in to the vibes, embrace exponentials, and forget that the code even exists."

In plain English? Instead of sitting down and writing code line by line, you just describe what you want to an AI, in normal human language, and the AI writes the code for you. You guide, test, and tweak. The AI does the heavy lifting.

"The hottest new programming language is English." — Andrej Karpathy, 2023 (and he was right)

THE TECHY TALK

Vibe coding leverages large language models that have been trained on billions of lines of code. These models don't just understand syntax — they understand software architecture and design patterns, which means they can generate entire application structures from a single natural-language prompt. The developer's role shifts from implementation to intent-setting and verification...

(that wasn't written by AI btw)

THE NUMBERS ARE RIDICULOUS

92%

of US developers use AI coding tools daily

41%

of all code written globally is now AI-generated

63%

of vibe coding users are non-developers

That last number is the one that blows my mind. The majority of people using vibe coding tools didn't even come from a coding background. They're entrepreneurs, designers, students, and regular people who had an idea for an app and just... built it. That would have been impossible two years ago.

A REAL EXAMPLE THAT ACTUALLY HAPPENED FROM $20K TO $1.8 BILLION

An entrepreneur named Matthew Gallagher used AI tools to launch a telehealth company called Medvi with just $20,000 and essentially no staff. He used AI for the coding, marketing, customer service, and analytics. In its first year, the company scaled to $401 million in revenue, and is projected to hit $1.8 billion this year. His only full-time employee? His brother.

Now, that's an extreme case. But it shows where this is heading. The barrier between "I have an idea" and "I built the thing" has never been lower in the history of software.

SO WHAT'S THE CATCH?

THE GOOD

• Anyone can build, not just developers

• Prototypes are done in hours, not weeks

• Repetitive work gets automated

• 74% of developers report higher productivity

THE BAD

• AI-generated code has 1.7x more major bugs

• Security flaws can slip through unnoticed

• Over 40% of junior devs deploy code they don't understand

• AI can't handle things like payment processing safely yet

One vibe-coded app suffered a data breach in early 2026 that exposed 1.5 million API keys and 35,000 user email addresses. The owner admitted they hadn't written a single line of code manually. Speed is great. But speed without oversight is how things break. Badly.

THE TECHY TALK

A 2025 study by VeraCode found that while AI models have dramatically improved at writing functional code over the last three years, the security quality of that code has not improved at the same rate. Larger models weren't better than smaller ones at generating secure code. This gap between "it works" and "it's safe" is the central challenge of the vibe coding era.

THE BIGGER PICTURE

Vibe coding is doing to software what smartphones did to photography. Before smartphones, taking a good photo required a camera, skill, and equipment. After? Everyone became a photographer. Before vibe coding, building software required years of training. Now? You need a good idea and a clear description of what you want.

That democratization is genuinely exciting. But it also means we need to be smarter about when to trust the AI and when to bring in a human who understands what's actually going on under the hood. The best vibe coders aren't the ones who trust the AI blindly, they're the ones who know enough to know when to double-check.

Even the person who coined the term, Karpathy, said by early 2026 that pure vibe coding is already becoming "passé," the next evolution is something called agentic engineering, where AI doesn't just write code but plans, builds, and verifies it in structured phases. So this is moving fast. As usual.

If you've been following this series since Blog #1, you've watched AI go from a chatbot that summarizes your emails to a co-pilot that builds software for you. Vibe coding is just the latest chapter. The question isn't whether AI will keep reshaping how we build things, it's whether you understand it well enough to use it wisely. That's exactly what we're here for. See you in the next one.