Claude Code is changing the developer profession: code is now written by agents, not people

What does it mean to be a developer in 2026? If three years ago the answer was obvious — a human writes code — then today the line has blurred. More and more people are creating working programs without ever typing a single line by hand.

And experienced engineers spend their workday not writing code, but managing agents. This is the subject of a recent episode of The Vergecast podcast. The hosts invited Paul Ford — a writer, entrepreneur, and one of the most thoughtful technology commentators of the past two decades.

Ford has long been watching how programming changes: back in 2015 he wrote the famous essay "What is Code," which took up an entire issue of Bloomberg Businessweek. The starting point of the conversation was Claude Code — an Anthropic tool that allows you to delegate programming tasks to AI agents directly from the terminal. With its appearance, more and more people have started calling themselves "coders" — even those who have never written programs professionally.

Claude Code accepts instructions in natural language and figures out how to implement them: it creates files, fixes bugs, runs tests. But it's not just about beginners. The main shift is in how the work of experienced developers is changing.

Professionals now spend less time on the actual writing of code and more time on task formulation, checking results, and coordinating multiple agents simultaneously. This is like a transition from a craftsman to a foreman: you are still responsible for the result, but your hands are occupied with something else. A logical question arises: what is left for humans?

If agents handle the routine — writing functions, debugging, refactoring — what is their value? Ford believes: in understanding context, in the ability to ask the right questions, and in responsibility for consequences. An agent doesn't know why this function is needed or who will use it.

A human does. A separate concern is what this means for new developers. Previously, the path into the profession lay through writing code: lots of code, with errors, with debugging.

This very experience shaped intuition. If today a beginner starts right away with managing agents — they won't have that foundation that allows them to understand why the agent made a mistake. This is a potential systemic problem for the entire industry that no one is yet addressing.

Claude Code and similar tools make programming more accessible — and that's good. More people can create tools for themselves and others without spending years learning. But at the same time it creates an illusion of competence: a person sees a working program and thinks they understand technology.

The gap between "wrote a prompt" and "understand what's happening" is enormous. Ford frames this as a question of trust: who do we trust — code that we understand, or code that works? When responsibility is blurred between human and agent, figuring out the reasons for a failure becomes fundamentally more difficult.

The future of development is not the end of the profession or its continuation in the old form. It is something new, which doesn't yet have a precise name. And as always with technology — how good this new thing turns out to be depends largely on how consciously we manage the transition.