Code is a Liability
There’s a lot of noise right now about the bottom falling out of the software market. Panic in the Slack channels, grim-faced posts on LinkedIn about sinking salaries. The truth is a lot less dramatic and a lot more damning. The market isn't collapsing. It’s just sobering up after a decade-long bender where it wildly overvalued the one thing that was never that valuable to begin with: the code itself.
Let’s be clear. Code is text. Text is cheap. Anyone with a functioning laptop and a Wi-Fi connection can churn out thousands of lines of it. The magic was never in typing:
if (x) { doThing(); }The real work—the valuable work—was in understanding the system, defining the right problem, and, most importantly, having the discipline to know when not to build something.
Somehow, we confused the receipt for the meal. The code is just the evidence that thinking happened. Or, increasingly, that it didn't.
The Cult of the Tool
This got worse somewhere in the last decade when engineers started rebranding. You weren’t a problem solver anymore; you were a “TypeScript developer” or a “Senior Python Engineer.” It’s absurd. That’s like a carpenter calling themselves a “hammer specialist.”
When you define yourself by the syntax you’ve memorised, you’re telling the world the only thing you bring to the table is a set of keystrokes. Tools change. The syntax gets updated. The craft of building sound, functional systems is timeless. By chaining your identity to a tool, you made yourself spectacularly easy to replace.
The Inevitable Correction
So why is the reckoning happening now? It’s not complicated.
Oversupply: We trained legions of people to write code, but not enough to solve problems. Now we have too many typists and not enough engineers.
Globalisation: Why would a company pay a £90k salary in London when they can get the same lines of code for £15k from Bangalore? If the only thing you’re selling is the text, you will lose to the lowest bidder.
AI: And now, here comes autocomplete on steroids, ready to spit out boilerplate in any language you ask for, instantly.
The glut of cheap code isn’t a breakthrough. It’s a liability.
And Then We Poured AI on It
This is where the farce really accelerates. The current obsession with AI coding agents is focused on the worst possible metric: output. “Look! My new AI buddy generated 30,000 lines of code in 45 seconds!”
Congratulations. You now have 30,000 new lines of liability. Each one needs to be tested, secured, maintained, and understood by a human who is already stretched thin. We are generating digital rubble faster than we can ever hope to clear it. Does the AI understand the non-functional requirements? The security context? The virtue of systemic simplicity? Of course not. It just confidently hallucinates a solution based on a prompt that was probably flawed to begin with. AI isn’t solving the problem; it’s just making the mess bigger, faster.
What We Actually Need
With the cost of producing code dropping to zero, the only thing left that has any value is the one thing the industry seems determined to ignore: engineering.
We don’t need more code. We need:
- Clearer definitions of the actual problem.
- The courage to ask if the problem is even worth solving.
- A deep understanding of constraints—reliability, usability, security, budget.
- The discipline to deliver the smallest possible system that works.
The industry doesn’t celebrate restraint. It celebrates feature velocity and lines of code. And so we continue to drown in receipts, while the actual meal—the careful, considered thinking—gets scarcer every day.
Code was never valuable. The ability to write it was just overvalued for a while. AI is making cheap code cheaper, and the desperate need for real, disciplined engineering has never been greater.
And, of course, what we’re getting is just more code.
