The Next Abstraction Layer: Value and Beauty in the Age of AI

Over the past year, I’ve been trying to understand what AI coding agents actually mean for software engineering.

Like many others, I’ve listened to discussions from leaders across the AI world and experimented with these systems myself. One evening, while deep diving into AI coding agents, something about this shift became clearer to me.

This essay is my attempt to explain that realization: AI is not replacing software engineering — it is becoming the next abstraction layer above it.

The Question Everyone Is Asking

Over the past year, conversations about artificial intelligence have been dominated by a single question:

Will AI replace software engineers?

The question surfaces everywhere — on podcasts, in conference halls, and in interviews with the people shaping the direction of this technology. Leaders such as Demis Hassabis, Dario Amodei, Andrej Karpathy, and Alex Karp regularly discuss the rapid progress of AI systems and what it might mean for the future of work.

For many people, these discussions create a sense of uncertainty. If machines can increasingly write code, debug systems, and even design software architectures, what happens to the profession built around doing exactly those things?

I found myself thinking about this question often.

Over the past few months, I had been listening closely to these conversations, trying to understand where the technology was actually heading. But the moment that clarified things for me did not come from a conference stage or a podcast interview.

It came late one evening while I was deep diving into AI coding agents.

As I experimented with these systems — guiding them, observing how they generated code, and watching how they reasoned through problems — something important began to crystallise.

What we are witnessing is not the disappearance of software engineering.

We are witnessing the emergence of a new abstraction layer on top of it.

A Pattern in the History of Technology

To understand what might happen next, it helps to look at history.

Technological progress rarely arrives as a single isolated event. Instead, it unfolds through waves of transformation that reshape how humans create value and interact with the world.

The Industrial Revolution mechanised physical labour. Machines amplified human strength and productivity, fundamentally changing manufacturing, transportation, and economic systems.

The computing revolution mechanised calculation and information processing. Tasks that once required enormous human effort — complex calculations, record keeping, and data processing — became dramatically faster and more reliable through computers.

The internet then connected the world’s knowledge and communication infrastructure. Information could move across the planet instantly, enabling entirely new industries and transforming how people collaborate, learn, and build businesses.

Today, we appear to be entering another wave.

Artificial intelligence is beginning to automate parts of reasoning, problem-solving, and creation. Systems that once required direct human effort are increasingly becoming collaborative processes between humans and machines.

Each of these technological shifts initially generates uncertainty. People worry about disruption, about jobs disappearing, and about whether their skills will remain relevant.

Yet history shows something interesting.

Societies adapt.

New technologies rarely eliminate the need for human expertise. Instead, they raise the level of abstraction at which humans operate. What once required deep technical interaction gradually becomes something humans can control at a higher level.

Another pattern appears repeatedly as well.

When a transformative technology first emerges, access to it is limited. Only a small group of institutions or individuals can use it. Over time, the technology spreads, infrastructure improves, costs fall, and access becomes democratised.

Eventually, what once felt revolutionary becomes ordinary.

Electricity is a perfect example. When it was first introduced, access was limited to a small number of facilities and wealthy households. Over time, the electrical infrastructure expanded across cities and countries. Today, electricity is so ubiquitous that we rarely think about it — yet it powers nearly every aspect of modern life.

Many emerging technologies follow this trajectory:

privileged access → democratisation → commoditisation.

Artificial intelligence appears to be moving along a similar path.

From Vibe Coding to Agentic Engineering

Within software development, this shift has become visible very quickly.

In 2025, a phrase began circulating widely in the developer community: “vibe coding.” The term, popularised by Andrej Karpathy, described a new way of building software — one where developers could describe what they wanted and AI systems would generate large portions of the implementation.

For many engineers, the experience felt almost magical.

For the first time, it felt less like writing software line by line and more like collaborating with a system that could reason alongside you.

But something changed recently.

Toward the end of 2025 and into early 2026, improvements in AI models created what felt like a step-function increase in capability.

Coding tools began evolving into something more powerful.

They began acting like agents.

Instead of merely generating code, these systems could plan tasks, explore codebases, run tests, iterate on solutions, and refine their own output.

This shift is increasingly being described as agentic engineering.

And it has led many people to believe that software engineers may soon become obsolete.

But this interpretation misses something fundamental.

AI coding systems are not replacing the foundations of software engineering.

They are building on top of them.

The Foundations Beneath the Abstraction

To understand why this shift does not eliminate software engineering, it is important to look at what actually makes modern software possible.

Beneath every application lies a vast technological foundation.

Programming languages define how instructions are expressed. Operating systems coordinate how programs interact with hardware. Infrastructure — whether on-premise or in the cloud — provides the compute, networking, and storage that allow systems to run at scale.

Distributed systems move data across machines. Protocols define how systems communicate. Security mechanisms protect information and ensure trust across networks.

This entire computational stack continues to exist regardless of how code is written.

Even if AI systems eventually generate a large portion of the world’s code, that code must still run somewhere. It must interact with infrastructure. It must integrate with existing systems. It must be architected, monitored, secured, and maintained.

None of these layers disappears.

In fact, as software becomes easier to create, the complexity of the systems that support it often increases.

This means that deep technical understanding does not become less important — it becomes more important.

Engineers who understand how these layers work — from programming languages to distributed infrastructure — play a crucial role in ensuring that complex systems function reliably and efficiently.

AI may change how code is produced.

But it does not remove the need to understand the systems that make software possible.

Value

At its core, the importance of technical expertise comes down to one idea: value creation.

Throughout the history of technology, the individuals who have shaped industries and economies were not simply those who used tools. They were the ones who understood the underlying systems deeply enough to create new possibilities.

Deep knowledge allows people to build systems that scale.

An engineer who understands distributed systems can design infrastructure that serves millions of users. Someone who understands databases, networking, and system architecture can build platforms that power entire industries.

These outcomes are not merely technical achievements.

These outcomes create economic value.

When a system improves efficiency across an organisation, enables a new product, or allows a company to operate at a global scale, it directly influences economic activity. The impact may extend from a single company to an entire industry, or even to the economy of a nation.

This has always been true.

And it remains true in the age of AI.

Agentic coding systems can dramatically accelerate the process of writing and iterating on software. They can reduce friction in development workflows and allow engineers to move faster than ever before.

But the ability to produce valuable outcomes at scale still depends on understanding what to build, how systems should be structured, and how different technological layers interact.

Tools can amplify execution.

But judgment, architecture, and deep understanding remain human responsibilities.

In this sense, the engineers who adapt to these tools may become dramatically more productive. Instead of replacing expertise, AI systems may allow experienced practitioners to multiply their capacity to create value.

Beauty

While the idea of value explains why deep technical expertise remains essential, there is another side of this technological shift that is equally remarkable.

It is the beauty of accessibility.

For most of the history of computing, building software required specialised knowledge. Programming languages, system architectures, and development tools created a barrier that separated technical professionals from everyone else.

Artificial intelligence is beginning to soften that boundary.

With the help of AI systems, people who come from entirely different professions — accountants, lawyers, designers, entrepreneurs, researchers — can now interact with software in a fundamentally different way. Instead of needing to understand programming languages or system internals, they can describe what they want to build and collaborate with intelligent tools to bring those ideas to life.

For many people, this represents a profound shift.

Tasks that once required hiring developers or learning complex technical skills can now be approached directly. Individuals can automate parts of their work, build small applications, create internal tools, and experiment with ideas that were previously out of reach.

This is where the beauty of the AI era becomes visible.

Technology is no longer confined to those who specialise in it. It becomes a creative medium that a much larger group of people can explore.

However, the scale of impact often differs.

A professional using AI to build a tool that improves their workflow is creating something useful and meaningful. But the systems that shape industries, power global platforms, or influence national economies still require deeper technical expertise and architectural thinking.

Both forms of creation are valuable.

One expands personal leverage.

The other creates large-scale economic impact.

The emergence of AI allows both to exist simultaneously.

The Next Abstraction Layer

Seen through this lens, the rise of AI coding systems feels less like the disappearance of software engineering and more like the arrival of a new layer in the evolution of computing.

Each technological wave has raised the level of abstraction at which humans work. Machines mechanised labour. Computers mechanised calculation. The internet connected knowledge across the world.

Artificial intelligence appears to be raising the abstraction once again.

In many ways, interacting with machines is becoming more natural. Andrej Karpathy once described this shift succinctly when he wrote that “English is the new programming language.” Increasingly, humans describe intent while machines translate that intent into executable systems.

Yet beneath this new interface, the foundations of computing remain.

Programming languages still exist. Infrastructure still powers the systems we rely on. Networks continue to move information across the world. Complex architectures still determine whether systems are reliable, scalable, and secure.

What changes is not the existence of these layers, but how humans interact with them.

Experts who understand the foundations can now operate with greater leverage than ever before, guiding powerful systems toward meaningful outcomes. At the same time, millions of people who were previously excluded from the world of software can now participate in creating tools for their own work and ideas.

In that sense, the AI era reveals two complementary forces at once.

Value, created by those who understand the depths of the system.

And beauty, created by making those systems accessible to many more people.

Seen together, these forces suggest that artificial intelligence is not simply replacing parts of the technological world we know.

It is building another abstraction layer above it.

And like every abstraction layer before it, it will not eliminate the foundations beneath it.

It simply changes how humans build the next layer.