Grieving the Death of Coding

Coding is dead. It may not have hit your organization yet, but it will. The act of typing in programming language syntax letter by letter is over. No more autocomplete. No more memorizing IDE shortcuts. No more vim motions you spent years wiring into your fingers. No more extract method, inline variable, rename—the physical rhythm of reshaping code by feel. That's over.

In April 2025, I built my first RAG pipeline with a coding agent. A knowledge base chatbot—from scratch, with an AI and human collaborator. Learning something new, watching it actually work, building alongside a machine that could keep up. It was thrilling.

By July, the thrill had curdled.

Denial

I started going back and forth with the agent 20 times to get code to a standard I was satisfied with. One week I was excited about a new model. The next I wanted to throw the whole thing away and go back to my IDE.

"Maybe I'll just take a break from agents," I told myself. "Go back to traditional coding for a while."

That break never lasted.

Anger

By August, I was watching the industry lose its mind. Developers spinning up 10 agents at once. Working on 10 user stories simultaneously. Bragging about speed.

WIP (work-in-progress) should be 1. We can't even ship one thing well most of the time—why would I want 10 tracks of work that I then have to code review at the end?

The problem wasn't the tool. The problem was that nobody was asking what problem are we solving?

Bargaining

In September, I rediscovered TDD.

"Fine," I thought. "If I'm going to use coding agents in production, we're doing it right." TDD. Prefactor before adding changes. Clear, readable tests that read like specifications. Pair with the agent AND a human to reach an optimal design.

My bargain wasn't retreat, it was: I'll accept this, but on my terms.

Depression

November. I attended the Samman Coaching Open Space looking for someone who would talk me out of this.

Someone who would tell me I was wrong, that I shouldn't be going down this path.

Nobody did. Everyone was in their spring—happy, excited, discovering possibilities. The same stage I'd been in six months earlier.

I found Llewellyn Falco and told him one-on-one: "I'm disappointed. I enjoy coding. I miss it."

He said his grandma enjoys knitting sweaters. But that's not what we do. We've found a better tool and we need to accept it.

It snapped me out of it. The next day I was back on the coding agent bandwagon.

But February brought it back. At Artium, the consultancy I work for, we read AI Vampire by Steve Yegge. On that call, I admitted something out loud: I have an AI addiction. If AI is supposed to make our lives easier, why am I working more than ever? Taking fewer breaks? Prompting an agent every chance I get?

My colleagues related. I wasn't alone on this one.

Acceptance

Acceptance isn't "okayness." It's accepting the new, permanent reality.

I can still have fun refactoring code—through agents. Now I'm focused on optimizing for quality and optionality, not speed. I still ship faster than before, but I always want to see how far I can push the maintainability of the software I deliver.

I don't want to rush. I'm a crafter. I've been one for years. Whether I write code in an IDE or guide a coding agent, I will deliver high-quality software to users in small batches and delight customers through agility and fast feedback.

For years I've focused on good design, TDD, refactoring—principles of software that haven't changed much in 50 years. I never chased frameworks, languages, or tools. That attention to software design has paid off.

Coding is dying, but engineering is alive and well.

If you're somewhere in this journey—the excitement, the frustration, the loneliness of being ahead of the curve, hang in there. There's pain around the corner. There will be ups and downs. But the craft survives the tool.

Know a developer who's struggling with this shift? Share this with them. Sometimes it helps just knowing you're not alone.