Vibecoding: revolution or risk?

The phrase "vibecoding" didn't emerge from a marketing team or academic paper. According to Wikipedia, it was coined by Andrej Karpathy, co-founder of OpenAI and former AI leader at Tesla, in February 2025. He described it as "a new kind of coding where you fully give in to the vibes, embrace exponentials, and forget that the code even exists" - a post that was viewed over 4.5 million times and sparked widespread discussion in the tech community.

What started as a casual description has become a defining trend of 2025 - one that's fundamentally changing how we think about web development.

The Numbers Are Staggering

The adoption statistics are remarkable. Second Talent's research reveals:

92% of US developers use AI coding tools daily
82% of global developers use them at least weekly
41% of all global code is now AI-generated
256 billion lines of AI-generated code were written in 2024 alone
87% of Fortune 500 companies have adopted at least one vibe coding platform

Perhaps most striking: among Y Combinator's Winter 2025 cohort, 21% of companies have codebases that are 91% AI-generated. This extreme adoption rate among well-funded startups suggests vibecoding may become the default approach for new ventures.

What Exactly Is Vibecoding?

Digiscorp explains that vibe coding is a method of software development that utilizes AI to convert human input into code. The underlying technology enables users to express their programming needs in natural language, and the AI translates these instructions into fully functional code.

The core idea is to make coding more accessible and efficient by reducing the technical barriers typically associated with writing software. It abstracts away many complex technical details, enabling both non-technical users and experienced developers to create software by focusing on higher-level goals.

In practice, it looks like this:

Developer describes what they want in natural language
AI generates code based on the description
Developer reviews, tests, and iterates
Repeat until the feature is complete

The Productivity Promise

The productivity gains are real. Second Talent's research shows that 74% of developers report increased productivity with vibe coding approaches. According to Y Combinator data, teams of 10 engineers are now delivering output previously requiring 50 to 100 engineers, fundamentally changing the economics of software development.

For web development specifically, vibecoding excels at:

Boilerplate generation - Components, API routes, configuration files
Pattern implementation - Common UI patterns, CRUD operations, authentication flows
Refactoring - Converting between frameworks, updating deprecated APIs
Documentation - Generating docs, comments, and README files
Testing - Creating test cases and fixtures

The Honeymoon Is Over

Six months into widespread adoption, reality has set in. Initial enthusiasm has given way to a more nuanced understanding of AI-assisted development.

Research on AI in software development reveals a concerning perception gap: 68% of junior developers report quality improvements with AI tools, while 48% of senior developers report quality degradation. This suggests less experienced developers may lack the pattern recognition to identify AI-generated code issues. Perhaps most telling: 89% of AI suggestions remain unchanged during code review, indicating developers often accept suggestions without thorough comprehension.

InfoQ's analysis of the Ox Security report echoes these concerns, finding that AI-generated code is "highly functional but systematically lacking in architectural judgment". API evangelist Kin Lane noted: "I don't think I have ever seen so much technical debt being created in such a short period of time during my 35-year career in technology."

From Vibes to Context Engineering

MIT Technology Review provides perhaps the most important insight: 2025 has seen a significant shift - a loose, vibes-based approach has given way to a systematic approach to managing how AI systems process context.

This evolution is called context engineering: the discipline of carefully crafting the prompts, examples, and constraints that guide AI code generation.

The article notes: "Although 2025 may have started with AI looking strong, the transition from vibe coding to what's being termed context engineering shows that while the work of human developers is evolving, they nevertheless remain absolutely critical."

The best results come not from blindly accepting AI output, but from:

Providing rich context about the codebase and requirements
Constraining outputs with specific patterns and conventions
Reviewing critically rather than assuming correctness
Iterating deliberately rather than accepting first responses

Impact on Employment and Hiring

Contrary to fears of widespread developer job losses, Datapro News reports a different trend:

Recent job postings and hiring trends in Silicon Valley clearly indicate that proficiency with AI coding tools has become a crucial asset. Employers are increasingly looking for candidates who not only write code but also integrate AI into their development workflows.

Key changes in the job market:

Job descriptions now frequently list experience with GitHub Copilot and ChatGPT as essential skills
Organisations are creating hybrid roles requiring AI process management and prompt engineering expertise
Economic analyses suggest companies employing AI tools are hiring more engineers, not fewer

The message is clear: AI augments rather than replaces developers - but developers need new skills to remain relevant.

Practical Recommendations for Web Developers

1. Learn AI Tools, But Understand the Fundamentals

Vibecoding works best when you can evaluate the output. If you can't read and understand the generated code, you can't:

Identify bugs and security vulnerabilities
Adapt code to your specific requirements
Debug issues when they arise
Optimise for performance

2. Adopt Context Engineering Practices

Move beyond simple prompts. Invest in:

Well-documented codebases that provide context
Style guides and patterns that constrain AI output
Code review processes specifically for AI-generated code
Testing strategies that catch AI-introduced bugs

3. Focus on Architecture and Design

AI excels at implementation but struggles with system design. Double down on:

Understanding architectural patterns and trade-offs
Designing systems for maintainability and scale
Making technology selection decisions
Defining interfaces and contracts between components

4. Maintain Healthy Skepticism

Every AI-generated line should be treated as code written by an extremely confident but occasionally wrong junior developer. Trust but verify - always.

The Future of Web Development

Vibecoding isn't going away. The productivity benefits are too significant to ignore. But the industry is maturing rapidly from the early "vibes" phase to a more disciplined, context-engineering approach.

The developers who will thrive are those who:

Use AI as a tool, not a crutch
Maintain deep technical understanding
Develop strong code review and quality practices
Continuously learn and adapt

The web development industry has always evolved. From jQuery to React, from REST to GraphQL, from servers to serverless - we've navigated major shifts before. Vibecoding is the latest, but it won't be the last.

Building with Modern Web Technologies?

For related reading on AI's impact on software development, see AI is eating software and vibe coding security risks.

Whether you're adopting AI-assisted development or modernising your tech stack, the fundamentals matter. My Web Development services deliver high-performance applications using React, Next.js, and TypeScript - with the code quality and architecture that AI tools alone cannot provide.

Get in touch to discuss your web development project.