Embrace a future where code is disposable and human creativity endures.

Ben Linders argues that AI turns code into a write-only medium.

3 min readInfoQ
Embrace a future where code is disposable and human creativity endures.

Ben Linders' piece lands on a truth that many in the software world are only beginning to whisper: if AI writes code that humans can't read, then the code's only real audience is the machine. That's not a failure of discipline. It's a fundamental shift in what code is for. When tests define behavior and become the documentation, the source file stops being a letter to your future self and starts being a build artifact. The implications are uncomfortable, but they're also clarifying. We're not moving toward a world where we read less code because we're lazy. We're moving toward a world where reading code at scale is simply no longer a viable quality gate.

The practical consequence for teams is stark: if you can't review AI-generated code line by line, and rewriting is cheaper than debugging, then your entire engineering strategy has to change. The old instincts don't just become outdated; they become liabilities. You don't invest in meticulous code review rituals that assume a human can hold the full system in their head. You invest in faster feedback loops, sharper test suites, and a willingness to throw away a module that's become too tangled to reason about. That's a hard pill for organizations that have spent decades building institutional knowledge through code ownership. But intent has already decoupled from implementation. The "why" lives in the prompt, the test, and the conversation, not in the syntax. So the real skill isn't writing or reading code anymore. It's being able to articulate intent clearly enough that the AI produces something worth keeping.

What we would tell a reader who asks, "So what do I actually do on Monday?" is this: stop trying to understand every line your AI assistant generates, and start treating it like a junior developer who writes too fast. You can't review all of it, so you build guardrails. You demand tests that are as expressive as the code they protect. You get comfortable with the idea that a 2,000-line function might be cheaper to regenerate from a well-crafted prompt than to refactor by hand. And you accept that your primary value is no longer in the mechanics of typing code, but in the creative judgment of deciding what the system should do next. That's not a diminished role. It's a more interesting one.

The open question we're watching is whether the industry's tooling will catch up. If code becomes write-only and disposable, then the debugger, the profiler, and the test runner become the last line of defense. But who maintains the tooling that validates behavior when the code itself is too dense to inspect? That's the detail worth tracking. Because if the tests become the documentation, then the quality of that documentation depends entirely on the discipline of the person writing the prompts. And that's a skill we're going to have to teach, measure, and value differently than we ever have before. The takeaway to quote: "Your value is no longer in writing code; it's in defining intent clearly enough that the code becomes a byproduct."

From InfoQ

Artificial intelligence (AI) makes all code write-only,. It’s too dense to read, and tests define the behaviour and become the documentation. Code is also disposable; it becomes easier to rewrite than to debug. Humans can't review AI-generated code at scale. Intent decouples from implementation; developers should focus on creativity.

Read the original at InfoQ