--- title: "The receipt was zero, the habit was not" description: "I wrote that a year of AI subscriptions bought me nothing. That was true about the invoice and wrong about everything else: the addiction, the ideas, the reading, and what happened when I pointed it at something else." date: 2026-10-10 --- Two weeks ago I published a post arguing that a year of AI subscriptions bought me nothing. No shipped product, no users, no output, just a line on a card that kept coming back. I meant every word of it and I still do, about the money. The receipt really is zero. But a friend read it and asked the question this post is about: is that really all you got? Because from where he stands, I am a different person than I was a year ago, and the difference is not nothing. He is right. The invoice column is empty. The person column is not. ## The addiction is real and I am not apologizing for it Let me start with the part that sounds worst. I code with AI every free moment I have. The moment there is nothing to do, my hands open the terminal. My head is full of new ideas all the time, one after another, most of them bad, some of them interesting, none of them leaving me alone. A tech Discord server I am in has a whole vocabulary for this: people call themselves AI addicts without irony. One person described waiting for the quota to reset as a form of rest. Another called himself a storytelling addict: the habit does not wait for the reset. Someone got a 20x plan and the first message was that it was dangerously addictive, might actually pay for it. When the free taste runs out, the verdict from the room is always the same: they gave you the taste so you would get hooked. I recognize all of it because it is me. Here is the thing I want to say plainly: the addiction is the asset. Not the subscription. Not the models. The habit of reaching for the tool, the reflex of opening the repo instead of the feed, the itch that shows up in an empty evening. A year ago my empty evenings went nowhere. Now they go somewhere, even if the somewhere is usually a dead end. A dead end you walked is still walking. The muscle that walks it does not care which destination you picked. ## What the year actually deposited I know more than I did. That sounds empty until I list it, so let me list it. I read GitHub repos the way some people read novels. I used to open a repo, see the file tree, and close the tab. Now I read source the way the channel reads source, following how someone structured an orchestrator, how they handled the failure I am having this week. I read papers, or at least the parts of papers that matter to what I am building. I read changelogs. I read issue threads the way other people read comment sections, because the fix for my problem is usually three replies deep in someone else's argument. None of that shows up on the invoice. All of it shows up in the person. The ideas are the other deposit. My head generates them constantly now, and I know most are worthless, and that is fine, because the generation itself is the skill. A year ago I waited for a good idea to arrive. Now I produce ten bad ones a day and the eleventh is occasionally real. The habit of having ideas is upstream of having a good one. You cannot filter a stream that does not flow. I want to be careful about levels here, because I wrote the rules myself in the last round of edits. These are observations from one person's year, judged by that person, with no control group and no measurement. I cannot prove the subscriptions caused any of it. Maybe I would have read the repos anyway. But the timing is suggestive and the mechanism is visible: the tool lowered the cost of starting something to near zero, so I started things constantly, and starting things constantly is a practice with or without a shipped product at the end. ## The transfer Here is the part that changed my mind about the year. A few weeks ago I started trying to move the addiction. Not quit it, move it. English. I have been putting off serious English study for years, the way you put off anything with no deadline. Then one evening I caught myself reaching for the terminal out of pure reflex, nothing to build, just the itch, and I thought: what if the itch does not care what it scratches? So I pointed it at English. Same evenings, same restlessness, different object. And it worked, or at least it started working, in a way that years of intending to study never did. The habit was already built. It just needed a new address. That is when I understood what the year had actually bought. Not a product. Not a skill with a name. A trained capacity to sit down every evening and go at something. The AI was the training equipment. The equipment was rented and overpriced and I do not miss the invoice. But the muscle is mine, and it transfers. The channel figured out the economics of this before I did: one evening someone noted that data plus English gets a fresh grad a twenty-million-dong offer. The tool changes. The learning habit is the thing that compounds. ## Both posts are true So let me reconcile the two, because I do not want to retract the first post. It stays. The money bought no output. True. Every finished thing I can point to from that year fits in a folder of dead ends. If you measure a subscription by shipped product per dollar, mine rounds to zero, and the post stands as written. The year built a person who starts things. Also true. The addiction, the reading, the ideas, the reflex to build instead of scroll, and now the discovery that the whole apparatus moves to a new target when I tell it to. None of that was on the invoice. All of it is in the room with me as I write this. The mistake in the first post was not the accounting. It was the unit. I measured the year in products, and in products it is zero. Measured in habits, it is the most productive year I have had. The subscription was overpriced training equipment for a capacity I did not know I was training. I would not pay that invoice again. I am glad, in a way I did not expect, that I paid it once.