The solution might be cancelling my AI subscription
AI coding agents are turning us into digital hoarders, and we're paying for the privilege.
Analysis
AI coding agents are turning us into digital hoarders, and we're paying for the privilege.
David Wilson's honest confession resonated with me because I've lived it. You ask Claude for a quick script. Forty-five minutes later, you've accidentally built a full-stack application with CI/CD pipelines, comprehensive test coverage, and a README that reads like a software company's pitch deck. The original problem? Still unsolved. But boy, does that repo look professional sitting there, gathering dust alongside the forty-seven other repos you spun up last month.
This is the dirty secret of AI-assisted development that nobody in the hype cycle wants to talk about: the tool works almost too well. The friction between "I have an idea" and "I have a functioning prototype" has collapsed to nearly zero. And humans, it turns out, needed that friction. Not because suffering is noble or because building things should be punishing, but because the struggle was doing crucial sorting work. It was filtering signal from noise, passion from whim, necessity from novelty.
Think of it like the difference between growing a vegetable garden and Amazon Fresh. When you spend weekends tilling soil, battling aphids, and waiting months for tomatoes, you develop a deeply personal relationship with those tomatoes. You eat every single one. You preserve the extras. You remember that garden. When a box of organic heirlooms shows up at your door twenty minutes after you thought "I should eat healthier," the psychological weight is entirely different. Half of them rot in the fridge.
AI coding agents have given us infinite tomatoes. And our fridges are overflowing with rotting ones.
But here's where the story gets genuinely interesting, and where I think the lazy "AI bad" takes miss the plot entirely. The ADHD responses in that Hacker News thread aren't outliers or anomalies. They're pointing at something real and important that I think has implications far beyond neurodivergence.
For people whose brains are wired to chase novelty compulsively, AI agents are functioning as a strange kind of prosthetic executive function. The agent provides the scaffolding. The ADHD brain provides the creative horsepower and the ability to hyperfocus when sufficiently stimulated. Together, they're producing output that neither could manage alone. That person maintaining inbox zero for the first time in their life? That's not trivial. That's someone whose relationship with technology just fundamentally changed in a way that improves their daily experience of being alive.
So we have two seemingly contradictory realities coexisting. For neurotypical users like Wilson, AI agents are accelerating a kind of productive-looking but ultimately hollow activity pattern. For many ADHD users, those same agents are unlocking genuine capability that was previously gated behind neurological barriers. Same tool, opposite outcomes.
This forces us to confront something uncomfortable: the "discipline" framing might be entirely wrong. Wilson's solution is curtailing use, developing self-control, treating the tool like alcohol at an open bar. Fair enough. But what if the problem isn't the user's relationship with the tool, but rather that we've built something optimized for a specific cognitive profile and are surprised when it interacts differently with other profiles?
The deeper issue is that the tech industry has spent years building productivity tools while implicitly assuming a universal human default. ADHD, autism, dyslexia, anxiety disorders — these aren't edge cases. They're millions of people with different relationships to focus, motivation, and completion. AI coding agents are accidentally revealing just how much of "best practice" software development was optimized for neurotypical brains all along.
What I find genuinely fascinating is the attention problem Wilson identifies. Calling it a "thermonuclear ADHD amplifier" is vivid and probably accurate for many people. But attention has always been a finite resource under assault. Television didn't destroy our ability to focus. Social media didn't either, though it certainly eroded it. AI agents represent the latest and perhaps most potent version of a technology that offers immediate gratification with minimal commitment — the techno-optimist's dream and the psychologist's nightmare simultaneously.
The sustainable path forward probably isn't willpower or subscription cancellation. It's building better interfaces between these tools and human intention. What if your AI agent asked, before spinning up a new project, whether you've finished the last three? What if it tracked your abandoned repositories and surfaced them as gentle shame notifications? What if the tool itself had opinions about your attention budget?
That sounds dystopian to some, but we already let apps track our screen time and send wellness reminders. The difference would be that the intervention happens at the moment of temptation, not after the damage is done.
Wilson's realization that a tool producing cheap reward with minimal friction is inherently a liability is genuinely profound, even if he arrived at it with some despair. But I'd push back on his conclusion that curtailing use is the only viable response. The ADHD users in that thread aren't using AI as a toy. They're using it as an assistive technology. For them, curtailing use would be like telling someone with clinical depression to just think positive thoughts.
The real contribution of AI might not be the code it writes. It might be the mirror it holds up, showing us just how different our cognitive architectures actually are, and forcing us to build a world that accommodates that diversity instead of pretending it doesn't exist.
Disclaimer: The above content is generated by AI and is for reference only.