The Age of Coding Is Not Ending. It Has Already Ended.

The biggest mistake in conversations about AI today is that people talk about the future as if it has not arrived yet.

They say: "Soon, cars will drive themselves." But in some cities this is no longer "soon." It is already part of the urban layer.

In Phoenix, San Francisco, Los Angeles, Miami, Orlando, Austin, autonomous vehicles are no longer science fiction. Waymo cars pick up passengers without a human driver. Zoox is testing purpose-built robotaxis. In China, Baidu Apollo Go and Pony.ai are operating at massive scale. In Abu Dhabi, WeRide and Uber have launched fully driverless commercial robotaxi operations.

For some people, autonomous vehicles are still a future trend. For others, they are already transportation.

Some people are still debating whether AI will write production code. At the frontier, that debate is already over.

Boris Cherny, the creator and head of Claude Code at Anthropic, said in an interview that he has not edited a line of code by hand since November. His work is now authored by Claude Code.

This is not a junior developer experimenting with a new tool. This is the person building one of the most important coding agents in the world — and he has already moved from writing code to directing a system that writes code.

For decades, a software engineer was the person who translated intent into machine instructions by hand. Now, at the frontier, the human increasingly defines the intent, architecture, constraints, taste, direction, and verification — while the operational layer of writing code moves to the machine.

Google has already said that 75% of its new code is AI-generated and then approved by engineers. That phrase matters: AI-generated and engineer-approved. It does not mean engineers disappeared. It means their role moved up the stack.

The human is still in the loop, but the work has changed: approve, orchestrate, verify, direct.

OpenAI leadership has spoken about AI tools moving from writing 20% of code to 80% of code in only a few months.

At the startup layer, the shift is even more visible. Founders no longer need to start by hiring a full engineering team to build an MVP. They describe the system, guide the AI, review the output, fix what breaks, deploy, and iterate.

Lovable says its users have created millions of projects and are building tens of thousands every day. Replit has said that Zillow employees created thousands of internal apps in a single year.

Not all of these apps are good. Not all of them will survive. That is not the point. The point is that the barrier between intent and software has collapsed. A person who once would have been only a user of software can now become the originator of software.

The wrong conclusion is: software engineers are dead. That is too simple. The better conclusion is: the manual operating layer of software engineering is dying first.

The bottleneck is no longer: who can type the code? The bottleneck is: who knows what to build? Who can define the right constraints? Who can judge whether the output is correct? Who can see the architecture? Who understands the product? Who can verify security? Who takes responsibility for what ships?

First, the human stops writing every line of code. Then the human stops manually moving every task through a CRM. Then the human stops manually answering every inbound lead. Then the human stops manually assembling reports, reconciling transactions, checking documents, launching campaigns, and coordinating workflows.

Eventually, a business stops looking like a group of humans operating software. It begins to look like a network of autonomous systems directed by human intent.

The age of human-operated software will not end all at once. It will end layer by layer.

First, hand-written code. Then manual operational management. Then entire business functions. Then the company itself as a human-operated machine.

Human Beyond is building for the world where the human no longer has to be the operator. The human remains the source of intent, taste, judgment, meaning, and responsibility. Machines take over execution.

So the thesis is not: "AI will change coding." Too late.

It is not: "Software engineering is evolving." Too soft.

It is not: "Agents are coming." They are already here.