When Building Software Became a Conversation

I'm building a complex software platform right now with just a couple of people.

We're using what Andrej Karpathy calls vibecoding. You describe what you want in natural language. AI writes it, refines it, debugs it. You verify, integrate, and ship.

When we started, I told the entrepreneur we'd likely hit one of three outcomes. First, we might create a mockup that comes close to what we need. That would be a godsend for a dev team because it beats technical specs. Second, we could handle the marketing site and application shell while devs focus on difficult components. Third, we might handle everything ourselves.

We slid into that third outcome easily.

The Moment I Realized the Bottleneck Had Moved

The realization hit when we got to the core solution of the platform. Over 300 individual accountancy checks. This is exactly the kind of complex, repetitive logic that traditionally requires a senior developer to architect carefully.

AI shines here because it handles repetitive tasks and gets them right.

I didn't need to know how to code it. I needed to understand the business and how software works. Not the coding, but the IT stack. The structure. The database setup. The integration points.

I've seen entrepreneurs create impressive software in Lovable. Really impressive. But they screwed up with bad database setup. They optimized for visuals and added features without understanding the structure underneath.

Another funny aspect: GitHub and similar tools were always the domain of developers. Even product owners had no clue. Now as a vibecoder you need to know how GitHub works. It helps a lot.

A New Kind of Literacy

There's a gap between "I can make the AI build what I describe" and "I know what structure will actually work."

Pure enthusiasm can't replace this understanding.

I've been both a philosopher and a business/tech person. When I'm structuring this application, I draw on decades of seeing how systems fit together. I understand the entire IT stack without needing to code it. I know where the integration points are. I can spot when a database schema will create problems six months from now.

The data backs this up. AI now generates 41% of all code globally in 2025. That's up from essentially zero two years ago. But here's the paradox: while developers expected AI to make them 24% faster, a rigorous MIT-backed study found experienced developers actually took 19% longer when using AI tools.

They still felt 20% faster.

This perception gap reveals everything. The bottleneck shifted from typing code to verifying it, integrating it, and knowing what to build in the first place.

The S-Curve We're Riding

Tony Seba taught me to see technology disruptions as S-curves. Gradual uptake, explosive growth, then tapering off. No successful technology in history has been adopted on a linear basis. Ever.

AI-driven development follows the same exponential pattern.

Seba also defines a market ripe for disruption as one where a new technology is 10x better or 1/10th of the price. Vibecoding enables small teams to achieve what previously required large development departments. That's a fundamental cost and capability shift.

We're not improving the old model. We're replacing it.

A Stanford University study found that employment among software developers aged 22 to 25 fell nearly 20% between 2022 and 2025. This coincides exactly with the rise of AI-powered coding tools.

The identity crisis is real.

Three Responses I've Observed

Developers respond to this shift in three ways.

Denial. Some insist AI is just autocomplete on steroids. They point to bugs in AI-generated code as proof it can't replace human developers. They're right about the bugs. They're wrong about what that means.

Resistance. Others see the threat clearly and fight it. They emphasize the craft of coding, the artistry, the deep knowledge required. All true. All irrelevant to the economic forces at play.

Evolution. The forward-thinking developers recognize their value lies in knowing what to build and addressing higher-order problems that emerge when building becomes fast.

I see a fourth path through synthesis thinking. The tension between "AI will replace us" and "AI is just a tool" dissolves when you understand that the human role is transforming, not disappearing.

We're moving from craftsman to curator and architect.

The New Scarcities

When speed stops being the constraint, judgment becomes the bottleneck.

Vision and taste. Knowing what's worth building matters more than knowing how to build it. 75% of engineers use AI tools, yet most organizations see no measurable performance gains at the company level. Individual developers get faster, but systemic bottlenecks remain unchanged.

The constraint moved upstream to vision.

Integration capabilities. A Forrester survey reveals developers spend only about 24% of their time writing code. The rest goes to design, testing, debugging, and stakeholder collaboration. AI accelerates the 24%. You still need humans for the 76%.

Verification skills. While 52% of developers report positive productivity effects, 46% say they don't fully trust AI results. When teams report considerable productivity gains with AI, 70% also report better code quality. With AI review integrated, quality improvements soar to 81%.

Verification became more valuable than execution.

Ethical judgment. As building becomes easier, the questions of what we should build and who benefits become more urgent. Abundance without wisdom creates dystopian outcomes.

From Scarcity to Abundance Thinking

I've spent years studying both disruptive technologies and human consciousness. The pattern I see connects to something deeper than software development.

We're living through a shift from scarcity to abundance. Not just in technology, but in how we think about human value.

The scarcity mindset says: if AI can code, developers become worthless. The abundance mindset says: if AI can code, developers become free to focus on wisdom, integration, and vision.

This mirrors the journey from logical thinking to intuitive wisdom. From the Magician archetype (mastering logic and materialization) to synthesis with the Monk archetype (consciousness and observation).

The developers who thrive will be those who can hold both. Technical understanding AND human judgment. Speed AND wisdom. Building AND knowing what's worth building.

What This Means Practically

For founders: Start building with existing AI tools. The barrier to entry dropped dramatically. You don't need a development department to validate your idea anymore. You need clarity about what you're building and why.

For developers: Evolve your skills toward judgment and architecture. Learn the entire IT stack, not just your specialty. Understand business context. Get good at verification. Your value lies in knowing what the AI can't know.

For companies: The bottleneck moved from finding people who can code to finding people who know what's worth building. Hire for vision and taste. Train for technical literacy. Measure success by shipped value, not lines of code.

The interface for building software is now conversation. Tools no longer require deep technical knowledge. They require clear articulation of what needs to exist.

This isn't the end of human relevance in software development.

It's the beginning of human flourishing in it.

When I'm vibecoding, I enter flow states that produce exceptional work in short bursts. The AI handles the repetitive logic. I handle the judgment calls. The integration decisions. The architectural choices that will matter six months from now.

The work became more human, not less.

We're not becoming the house pets of AI. We're becoming more fully ourselves. More intuitive. More focused on wisdom rather than execution. More able to create abundance rather than fight over scarcity.

The bottleneck moved. The question is whether you'll move with it.