I was asked recently whether I think I’m addicted to AI coding. My first answer was yes. It came fast, before I’d actually thought about it. That speed is its own small piece of evidence. But once I slowed down, I noticed the yes was built on two assumptions I hadn’t examined, not just one. The obvious one: that “addicted” meant something like compromised, when the real question was about what the experience feels like. The quieter one: that “coding” was even the right word for what I do. It isn’t, not anymore. I’m not writing the lines myself. I’m directing several AI agents at once, assigning the pieces, reviewing what comes back, redirecting the ones that wander. Orchestration is the more accurate term, and it’s not a small correction: directing several threads of work at once isn’t incidental to the activity, it’s the activity itself. Both corrections point the same direction, toward being precise about what’s actually happening before deciding what to call it. From here on I’ll use the accurate word.
What it feels like is specific enough to name precisely. Building, solving, and multitasking are happening at once, and each one is occupying a different layer of attention at the same time. That’s not incidental to the appeal. It is the appeal. I recognize the feeling from being a teenager in front of Missile Command: the pleasure wasn’t really the game, it was that every channel of attention was full simultaneously, with nothing left over to spend on anything else. Orchestration gives me the same total occupancy, except now there’s an artifact at the end of it instead of a score. That difference, stimulation that also produces something real, is exactly what makes the question harder to answer honestly than it would be for a video game.
There’s a second source, close behind the first one. I was a software
developer before I was anything else: computer science degree, then years working up from Junior Analyst Programmer through the ranks to CIO and CTO. In that first stretch, a day could contain the whole arc: see a problem and fix it, or have an idea and build it, and go home that evening having actually finished something. That particular satisfaction didn’t survive the promotions. The CIO and CTO roles gave me plenty worth having: scope, leverage, the ability to shape outcomes at a scale no individual developer could reach. But they didn’t give me that. I noticed the absence for the better part of four decades, and I want to be precise about what I did with it, because it isn’t the clean story I told a draft ago: I didn’t just let it go. I chased it, frequently, as new technology matured enough to plausibly justify it: the internet, the intranet, collaborative tools, social media, and others since. I’d get intrigued by what had just become possible, and then go find a problem in the organization that justified deploying it. That’s the actual pattern, not the tidied-up one: a legitimate business case, arriving right on cue, standing in as the reason instead of the pull underneath it. So I won’t say the feeling never drove me to go looking. It drove me to go looking, over and over, for forty years, and I was good enough at my job that I never had to call it what it was. What it means now is that AI orchestration isn’t only reactivating an adolescent stimulus loop. It’s also the least disguised version yet of something that’s been running for most of my working life. Two threads, so far, landing on the same activity, and one of them has a forty-year track record I’d been narrating as professional curiosity.
There’s a third source, and I left it out of both of the first two. I’m self-diagnosed, not clinically assessed, as both autistic and ADHD, and I want to be careful about what weight that claim can carry here: it isn’t a diagnosis, it’s a pattern I recognize in myself with enough confidence to build on, and no further than that. What it might explain isn’t a third craving alongside the other two. It’s a possible shape underneath both of them. Autistic attention and ADHD attention don’t usually pull in the same direction: one wants to go deep on a single thing and resents being dragged out of it, the other wants to be in several places at once and gets restless the moment it isn’t. Usually, having just one of the two already takes real energy to manage. What I seem to have is both, at once, which doesn’t add a second problem on top of the first. It changes what the depth and the scatter are working against. Instead of competing with each other, they’re being used together. If that’s right, the three-dimension itch I described at the start isn’t only a taste I happened to develop. It’s closer to a shape of work that matches how I’m actually built. The teenage feeling in front of Missile Command and the forty years of finding a business case for new technology may not have been two separate cravings that happened to look alike. They may have been the same pull in me, finding whatever version of itself was available at the time. AI orchestration would just be the first version that doesn’t ask me to translate it into something else first.
I want to be careful here, because this is the most flattering explanation I’ve offered so far, and the most flattering explanation is exactly the one I should trust least. Knowing why something pulls this hard doesn’t settle whether the pull is fine. An explanation like that, if it holds up, accounts for the shape of the itch. It doesn’t get to also excuse how much of my week it takes, and I don’t want to use it that way even by accident.
The instinct, once “addiction” is on the table, is to go looking for external damage: am I sleeping enough, am I present enough. I did this myself, within seconds of being asked. And I have a rebuttal ready for each one. Not sleeping enough? I nap whenever I want. Neglecting the relationship? My partner is genuinely proud of the tools this produces. Both rebuttals are true. Both also function as an exit from the question rather than an answer to it, and that’s worth noticing on its own terms. Not because the rebuttals are dishonest, but because having a clean counter for every symptom is one shape a functional dependency can take. That’s not a rule I can prove, and I know it’s a rigged one if I leave it standing as written: a weak rebuttal would read as avoidance, and now a strong one reads as suspicious too, which means nothing I say could ever count as evidence in my own favor. So the fair version has to leave room for the plain reading as well: that the naps genuinely cover the sleep debt, and the relationship genuinely has enough in it to take this. If those things are true, they have to count as real evidence, not get relabeled as a cleaner disguise because the more interesting story needs them to be one. “The people who struggle to see their own patterns clearly are rarely the ones failing outwardly” isn’t a law I can generalize. It’s just true often enough of me, in this instance, that I don’t get to dismiss it either.
So the checklist is the wrong test, not because the concerns behind it are wrong, but because it’s checking the wrong location. Sleep and relational presence are downstream proxies for something upstream: whether I’m choosing this, moment to moment, or whether the loop is choosing for me. A nap solves a sleep debt. It doesn’t tell me anything about whether I can be pulled out of a build mid-thread by something that actually deserves my attention, and leave cleanly. Not grudgingly, not still half-running the thread in the background, not irritable until I get back to it. That’s a different question than “can I stop.” I’ve never actually had to test that one hard enough to know for certain it’s true, which is itself worth sitting with. What I do know is that stopping was never the part I doubted. Leaving without residue is the part I don’t have an honest answer for yet.
The relationship rebuttal deserves more honesty than I gave it in the first pass. “She’s proud of the tools” and “she feels neglected while I build them” are not in tension. They’re both just true, at the same time, about the same behavior. Using the first to answer the second is the move I want to flag in myself, because it’s a rhetorical trick more than an argument: it changes the subject from how I was present to what I produced, and those are not interchangeable currencies no matter how good the output is. Pride in the artifact doesn’t retroactively correct the hours it took to make it. I don’t think the answer here is to stop building. I think the answer is to stop offering the output as a rebuttal to a complaint about presence, because it isn’t one.
I don’t think I land on a clean verdict, and I’m suspicious of any version of this essay that would. Calling it flow lets me keep the whole thing without examination. Calling it addiction offers a different exit: the word carries an assumption that things aren’t working, and I can point to the shipped tools, the caught-up sleep, the relationship’s real strength, and use them to argue my way back out from under the label. Flow lets me skip the inquiry. Addiction lets me argue my way out of it once it’s raised. Both get me to the same place, not having to sit with the question, by different doors. What I actually know is narrower than either label: the thing that lit up in me at fifteen in front of a screen full of incoming missiles, and the thing I’ve been feeding under cover of legitimate technology projects for forty years, are both lit up at the same time now, in the same activity, running as one thing instead of two, though neither one needs a business case to justify it anymore, and nothing outside it draws the edges the way a deployment memo used to. The career version had a deadline doing that job whether I wanted it to or not. This version doesn’t have one, which means the job of stopping belongs entirely to me now, and I don’t have a long enough track record with that yet to say how well I do it.
Whether that’s a problem isn’t something I get to settle by finding the right label for it. The test I’d actually trust is smaller than a label: the next time something outside this needs my attention, do I notice on my own, in time, or do I notice late. I don’t know that answer yet.