The Barrier Is Gone: What Happens When Everyone Can Build
Karpathy coined 'vibe coding.' A year later he moved on to 'agentic engineering.' 25% of YC startups have 95% AI-generated codebases. The casual phase was a phase. What replaced it is a discipline.
Andrej Karpathy coined "vibe coding" on February 2, 2025:
"There's a new kind of coding I call 'vibe coding,' where you fully give in to the vibes, embrace exponentials, and forget that the code even exists."
A shower-thoughts throwaway tweet that became the defining software term of 2025.
One year later, Karpathy himself moved on. His new preferred term: "agentic engineering."
"'Agentic' because the default is that you are not writing the code 99% of the time, you are orchestrating agents. 'Engineering' to emphasize that there is an art and science and expertise to it."
The casual phase was a phase. What replaced it is a discipline. And the distinction matters.
What's Actually Happening
The Non-Technical Builder
The tools are real. Lovable, Replit Agent, Claude Code, Cursor, v0. Non-technical founders ARE building and shipping.
25% of YC Winter 2025 startups had codebases 95% AI-generated. Garry Tan: "You don't need a team of 50 or 100 engineers. You don't have to raise as much. The capital goes much longer."
88% of global organizations reported using AI in at least one part of their business in 2025, up from 78% the previous year.
What Works
Quick MVPs, internal tools, prototypes, landing pages, basic CRUD apps. A non-technical founder can absolutely get a working v1 shipped. The "Minimum Viable Product" has become the "Minimum Vibable Product."
What Breaks
Production security. Scale. Maintenance. Edge cases. Compliance.
The gap between "it works on my machine" and "it works for 10,000 users without leaking data" is exactly where non-technical builders hit walls.
The Pieter Levels Model
Pieter Levels is the poster child of solo AI-powered building. $3M/year, zero employees. His fly.pieter.com (an MMO flight simulator) was built in 30 minutes, scaled to 320,000 players in 17 days, generates $87K/month. Photo AI does $138K/month.
But Levels is a self-taught programmer who's shipped 40+ projects over a decade. He uses PHP because he mastered it years ago. His AI advantage sits on top of deep product intuition and shipping muscle.
He's not a non-technical person who learned to code with AI. He's an expert who uses AI to 10x his existing expertise. The 8-to-80 pattern.
The Cal Newport Counter
Cal Newport, writing in The New Yorker (December 2025), titled his piece "Why A.I. Didn't Transform Our Lives in 2025."
The agents Sam Altman promised? They didn't arrive in the way predicted. Newport's conclusion: "We actually don't know how to build the digital employees that we were told would start arriving in 2025."
His ask for 2026: stop reacting to what AI might do. Start reacting to what it actually does today.
This is the sober counterweight to the hype. The tools are dramatically better. The transformation is real. But it's slower and messier than the narrative suggests. Most people who vibe-coded something in 2025 stopped maintaining it by 2026.
The Three Outcomes of Barrier Lowering
Outcome 1: More Software, Same Quality Distribution
When the printing press lowered the barrier to publishing, more books were printed. Not better books. More books. The distribution of quality remained the same: a small fraction of excellent work, a large middle of mediocre work, a long tail of garbage.
AI coding tools will produce the same distribution. More software. Not better software. The excellent software still requires excellent judgment, excellent architecture, excellent taste. The tools accelerate production at every quality level simultaneously.
Outcome 2: Expertise Becomes More Valuable
When everyone can produce a v1, the v1 isn't differentiating. What differentiates:
- Knowing which v1 to build (product sense)
- Making it work at scale (engineering judgment)
- Making it secure (security expertise)
- Making it beautiful (design taste)
- Making it succeed (distribution knowledge)
These are all forms of expertise. None of them are automated by the tools. All of them are multiplied by the tools.
Outcome 3: The "Quality Gap" Replaces the "Skill Gap"
The old world had a skill gap: can you code or can't you? The new world has a quality gap: can you produce production-grade software or just demos?
The skill gap was binary. You either knew Python or you didn't. The quality gap is a spectrum. Everyone can build something. The question is whether it works at scale, handles edge cases, stays secure, and remains maintainable.
The market is flooding with AI-shipped software that looks finished on the surface but has no architecture, no security, no error handling, no monitoring. It works in the demo. It fails in production.
The person who can tell the difference—and fix the difference—becomes more valuable, not less.
The Commoditization of Implementation
Sidu Ponnappa's line concept applied to consulting:
Below the line (commodity):
- Building chatbots
- Setting up automations
- Connecting APIs
- Creating landing pages
- Basic RAG implementations
Anyone can do these. The barrier is gone. The pricing follows: what cost $5K now costs $500.
Above the line (premium):
- Architecture for scale
- Security for production
- Domain-specific AI systems
- Legacy integration
- Compliance in regulated industries
The barrier is still high. The expertise required is deep. The pricing holds.
What This Means
For Non-Technical Builders
Build the v1. Ship it. Validate the idea. The tools let you do this. This is genuinely transformative for entrepreneurship.
But know the limit. When the v1 needs to become a production product, you need expertise you don't have. Not coding expertise. Judgment expertise. Architecture, security, scale, maintenance.
For Engineers
Your value shifted. From "I can build this" (everyone can now) to "I know what to build and how to build it right" (few can).
For the Market
The flood of AI-built software creates its own demand for quality assurance, security auditing, architecture review, and production hardening. The person who ships with AI and the person who makes AI-shipped code production-grade are two different roles. Both are needed.
We're in the "flood" phase. The "discipline" phase is beginning. The people who navigate the transition from vibes to engineering will define the next era.
Compiled from Andrej Karpathy's vibe coding origin and evolution, Pieter Levels' solo builder model, Cal Newport's counter-narrative, Garry Tan's YC data, and market analysis across 15+ sources.