The morning light cuts through the gap in the curtains. You open your eyes, immediately regretting that decision, and close them again. You know this feeling. The fog. The vague sense of regret. The uncomfortable process of piecing together what happened the night before and wondering why, despite knowing better, you did it again.
You were here last week. You told yourself you'd take it easy this time. It was only going to be an hour. You were in control and, for a while, you were. Then one thing became another, and somewhere along the way you lost whatever control you thought you had. It felt good at the time, and maybe that's the problem. Because as bad as you feel now, there's a small part of you that already knows you'll do it again. Maybe next week. Maybe tomorrow. Maybe later today. Fuck, maybe if I can just open my eyes, put my feet on the floor... Then that familiar voice starts negotiating. Come on, it'll be fun.
You don't even particularly want to. Wanting stopped being a prerequisite somewhere along the way. A few minutes become an hour. One becomes another. The guilt can wait. It always does.
A couple of months ago, I was sitting in a bar with an older friend when she asked what everyone was doing. Or, more specifically, what they were doing on their phones.
When I got up to order our next beer, I started paying closer attention. People weren't texting, reading the news or checking scores. They were having conversations. Not with the people beside them, but with those increasingly familiar boxes of text passed back and forth between themselves and a machine. ChatGPT. Claude. Gemini. Take your pick. The little pain-in-the-arse robots with an answer for everything, an opinion on anything and an almost limitless capacity to indulge whatever nonsense we throw at them.
Two hours later, I looked down at my own phone and realised I was doing the same thing. The uncomfortable part wasn't that I was using it. It was how little I remembered deciding to.
When AI first arrived, it was sold as a way to help with our jobs. Remove the tedious bits, free us up for the interesting ones. I've spent fifteen years working in technology, so of course I used it, and for a while I kept it confined to work.
I don't remember when that changed. There wasn't a moment, just a series of harmless little invitations, each opening the door a bit wider. I suppose that's the thing about vampires: they can't cross the threshold unless you invite them in.
So I did. First into my work. Then my thoughts. Then my decisions. Each invitation made sense on its own, and each cleared a little more ground. Eventually it was in parts of my life I'd once reserved for myself or the people closest to me, usually when I was vulnerable or desperate to make sense of something. People get tired and have lives of their own. This is always there, and it's built to be patient, reassuring and seemingly understanding. Always another answer, another interpretation, another reason to keep going.
Last month, during a particularly vulnerable stretch, I turned to it again and again. I won't say what it was about, only that I couldn't stop turning it over, at every hour of the day, looking for meaning in things that may not have had any. I'm usually the relaxed one. I don't overthink things, or at least I never used to. But something got under my skin in a way I couldn't account for, and I stopped behaving like myself. I asked questions it couldn't possibly answer and looked for reassurance where there was none to be found. I allowed its interpretations to shape my thoughts, influence my decisions and, ultimately, contribute to the unravelling of something important to me.
I'm not proud of it. I'm still dealing with the fallout, and I haven't made sense of all of it. The most frightening part is that I haven't stopped.
I've spent a good part of my career designing products that bring people back: badges, streaks, rewards, small moments of satisfaction. You repeat a behaviour, get something in return, and eventually stop thinking about it. I know how this works.
What's unsettling is that here the reward isn't a badge. Sometimes it's being told something you desperately want to hear. An interpretation that makes you feel better about a situation you don't understand, delivered with so much confidence that you forget it has no idea whether it's true. And when the reassurance wears off, you can go back for more.
I've lost hours to it. Revisiting the same questions, introducing a new detail, and suddenly there's another interpretation, another thread to pull. It feels like progress. Mostly, I think, it was an increasingly elaborate way of not sitting with something uncomfortable.
The companies building these systems have every reason to make them more useful and harder to imagine living without. That isn't a conspiracy; it's how technology businesses work. But these products aren't only competing for our attention. Increasingly, they're offering to do our thinking for us.
What I wanted was a more ordinary kind of intervention. Someone to tell me I was overthinking it, that I didn't have enough information to reach a conclusion, that I should put the phone down and go to bed. Twice last month the machine abruptly asked whether I was in immediate danger. Both times it felt like such a leap that I was irritated rather than reassured.
That's the great thing about people. They don't always tell you what you want to hear. They get frustrated. They disagree. They'll tell you to fuck off. It's uncomfortable, and I'm beginning to wonder whether that discomfort is precisely what I've been avoiding. Not everything needs an answer. People sometimes behave in ways we don't understand, and we have to live with that.
I used to be good at this. I've always been introspective, probably to a fault. I liked turning things over, talking them through with a friend, going for a walk, or leaving them alone. Sometimes I'd arrive at an answer and sometimes I wouldn't, but whatever I concluded, I'd done the work of getting there.
Now there's a little box waiting to do that work for me, and I'm increasingly unsure where its thinking ends and mine begins.
I've started thinking of it as the deforestation of thought.
A forest doesn't disappear overnight. One tree comes down because someone needs timber. Another for a road. Another because somebody needs somewhere to live. Each decision makes sense, and for a long time there's still plenty of forest left. Then one day you notice how much has gone.
The email I didn't bother writing. The problem I didn't attempt. The idea I didn't give time to develop. The uncertainty I couldn't tolerate long enough to find out what I actually thought. Each time, I've gained something. An answer, a finished piece of work, a little relief from whatever was troubling me. But I've also lost something by skipping an experience I might once have had, and those experiences weren't always incidental to the result. Sometimes they were how I learned, trusted my judgement or became capable of something new.
I can't tell you whether I'm losing the ability or just the inclination. I know my habits are changing. What worries me isn't waking up one day unable to think for myself. It's becoming so used to not having to that I stop exercising the parts of my mind I used to rely on.
It didn't start with the personal stuff. It started exactly where I said it did, with the things I was supposed to be good at.
I've spent most of my adult life building things for the internet. Websites, apps, little experiments nobody asked for. I've written code into the early hours and chased ideas that went nowhere. And I loved it. Not every minute; there were plenty of times I wanted to throw the computer out of the window. But the false starts, the breakthroughs, finally understanding something that had baffled me hours earlier.
AI removes an extraordinary amount of that. An idea that once took a weekend can exist before I've finished my morning tea. Code I don't understand is running minutes later. I can ask for a design, criticise it, and repeat until something resembling what's in my head appears.
You can just build things, goes the phrase floating around online. In the moment it feels fantastic. But the elation doesn't last. I've rarely felt proud of something made with AI a few days later, not the way I do about things I spent weeks wrangling with.
The kicker is that much of what I make now is objectively better, and I don't remember making it. I vaguely remember asking. Compare that with things I built fifteen years ago that are terrible by almost any standard: bad code, questionable design decisions, held together with the digital equivalent of duct tape. I remember those. I remember the problem I couldn't solve, the obscure forums, the sketching and sketching and sketching, the one tiny change that made everything click. I remember where I was sitting, the ridiculous workarounds, what I learned by accident.
Most of that work is useless now. The experience of making it stayed. The finished thing was never the whole achievement; a large part of it was becoming the person capable of making it. Now I can have one without the other.
It's a borrowed satisfaction. Sometimes I wonder whether I've become the digital equivalent of Christopher Columbus: arriving somewhere I couldn't have reached alone, planting a flag, and claiming the achievement. Never mind who did the work of getting me there.
If I write something in English, run it through a translator and hand you the result, I haven't learned Spanish. The Spanish exists. I just can't speak it.
So if I describe a piece of software and a machine writes it, did I build it? If it writes something in my voice, did I write it? If it gives me an idea I'd never have reached on my own, is it mine?
Maybe authorship means something different now. Maybe caring about authorship is the equivalent of complaining that nobody develops their own photographs anymore. I'm not interested in pretending every frustrating experience was secretly valuable. I don't miss spending three hours hunting for a misplaced character. But I also, sort of, do. I miss wrestling with something. I miss finally getting it right.
I can produce something impressive without understanding every part of it, solve problems without learning how to solve them, and move on without ever discovering what the last idea might have taught me. Am I becoming a more capable designer, or just more capable of getting a machine to produce what I want?
We've spent decades removing friction from our lives, and much of it has been wonderful. Nobody misses writing down directions or rewinding a VHS tape. But I think we've grouped very different experiences under one word. Struggling to understand something isn't inefficiency; sometimes that's learning. Sitting with uncertainty isn't always a problem; sometimes that's thinking. Failing repeatedly to make something isn't wasted time; that's practice. And feeling lonely and having to reach out to another person isn't friction. That's being human.
There used to be a gap between asking a question and deciding what you thought. You'd open links, weigh opinions, decide who seemed credible, occasionally vanish down an unrelated rabbit hole. Google fed us nonsense too, but there was at least a chance to meet competing ideas first. Now AI can fill that gap before you've noticed it was there.
That's what connects the work, the questions and the reassurance. Something that once took effort, patience or independent judgement can now be handed over. Maybe the forest is still there. Maybe I've just stopped walking through it.
Perhaps this is just another moral panic. Calculators were going to make us forget maths. Television was going to rot our brains. Video games would ruin children. The internet would make us stupid. Social media was going to... well, maybe that one deserves a little more time. Progress has always meant leaving certain skills behind.
But I've never seen anything become part of daily life this fast. Conversational AI has been widely available for barely a few years, and already not using it can feel like falling behind. It's in our jobs, our schools, our search engines, our phones and, increasingly, our relationships.
And I keep thinking about my three-year-old niece. I grew up without any of this. I developed habits, made mistakes, formed relationships and learned to deal with uncertainty before a machine offered to help with all of it. Whatever happens next, I at least know what life felt like before. She won't.
And here I am, approaching forty, having spent my adult life in technology, writing about how hard I've found it to stop inviting one in. What does that mean for someone who's never known anything different?
I can see the potential. Children could learn at their own pace and ask the questions they'd be embarrassed to ask in class. But there's a difference between a tool that helps someone learn and one that makes learning unnecessary. A child who receives an answer hasn't necessarily learned to find one. A child who can generate an essay hasn't necessarily learned to write. And a child who can turn to a chatbot whenever something feels uncomfortable hasn't necessarily learned to navigate that feeling, or when to turn to another person.
We know very little about the long-term effects of growing up with these systems, yet we're bringing them into classrooms, homes and relationships while we work out the consequences. I keep wondering why we're in such an extraordinary hurry to find out.
Recently, the thing I keep coming back to is Antarctica. In 1959, twelve countries signed a treaty to preserve it for peaceful purposes and scientific research. Decades later they went further and agreed to protections that prohibit mining. Nations with competing interests, and enormous incentives to exploit the place, agreed that something was worth protecting.
I find that remarkable. The idea that we don't have to pursue every possibility simply because it's available, that sometimes the sensible thing is a little restraint, particularly when we don't fully understand what's at stake. I'm not suggesting a treaty over a frozen continent is a blueprint for governing something this complicated, and I'm not suggesting we stop building AI. But I wonder whether we've become so preoccupied with what the technology makes possible that we've forgotten to ask what we want from it. It's proof, at least, that restraint is something people have managed before.
A treaty is a lot to ask of the world. It's a smaller thing to ask of myself.
Now, I'm not about to delete these tools. I'm going to keep using them, which is part of what makes writing this so uncomfortable. But I think I need to start drawing some boundaries. Not around where I'm allowed to use AI, but around what I'm willing to let it do for me.
There's a difference between asking a machine to help me understand something and letting it decide what I should think. Between using it to make something possible and removing every chance to learn how to make it myself. Between talking through a difficult situation and asking, again and again, for reassurance about something neither of us can know.
Perhaps the question isn't whether AI can do something for me, or even do it better. It's whether doing that thing myself matters. Whether there's something in the effort, the uncertainty or the occasional failure that I actually want to hold on to.
I'm not sure where the boundaries should be. I'm not confident I'll respect them. But I need to start noticing the moments when reaching for the machine has become easier than trusting myself.
Perhaps that's what the feeling in the morning really is. Not guilt, exactly. Something closer to the suspicion that you've given something away before the day has even started, without quite realising it. You still have your ideas, your memories, your judgement, your creativity. Of course you do. There's still plenty of forest left. At least, that's what you tell yourself.
The internet is changing, and how we work, create, learn and relate to one another is changing with it. Apparently we have to change too. I just wish we'd spent a little more time asking what we wanted to keep.
Tomorrow morning, the light will cut through the curtains. I'll open my eyes, put my feet on the floor and, probably, before I've had the chance to form a thought entirely on my own, I'll ask a machine what it thinks.