I was thirteen, and the book I could not put down was an ammunition reloading manual. I actually still have it and as chance would have it, I found it while cleaning in the basement. Speer Reloading Manual, number 12. It has a bucolic, sunlit workbench scene on the cover; a workbench with all the kit you would need to load your own ammunition.
How a reloading manual came to be within reach of a thirteen year old is not a question anyone in my house would have understood. This was, afterall, early 1994. Firearms were as common as pencils in my household. I grew up around them the way other kids grow up around board games. My grandfather gave me a small bore shotgun for my fifth birthday, which sounds, to a certain kind of reader, like the opening line of a case study, and was in fact an ordinary act of love in the place I am from, the Iron Range equivalent of a first bicycle. The manual was shelf furniture. Nobody assigned it to me and nobody monitored my reading of it, because there was nothing to monitor. It was a book about a chore.
The front half of a reloading manual is instructions. The back half is tables: velocities, ballistic coefficients, drop, smokeless gunpowder measured in grains. The tables were the part that got me. Somewhere inside those columns was a curve, the arc a bullet actually travels, sagging under gravity, bleeding speed into the air, and I wanted to compute it. The mathematics that computes it is calculus. I was thirteen. I did not have calculus.
What I had was a copy of Visual Basic 3, obtained by an older neighbor friend for me from his university’s bookstore, the computer in the house, and the conviction that the area under a graph could be filled with rectangles, and that if you made the rectangles small enough the error would stop mattering. And that would allow me to plot a bullet’s trajectory. I did not know that Bernhard Riemann had gotten there first, roughly a hundred and forty years earlier, or that the trick had a name, or that it was the front door of the exact discipline I was missing. I reinvented his sums badly and implemented them worse, and the program drew an approximate trajectory on the screen, wrong in every decimal and right in its shape.
Nobody thought anything was unusual, including me. In that house, the guns were the normal part. The strange object under that roof was the boy computing where the shot would land, and there is no checklist anywhere with a box for that.
This essay is not about marksmanship and it is not about software.
Even medicated, even stable, even with the band as narrow as the pharmacy can make it, I get fixated. Not interested. Fixated. A topic arrives, usually sideways, and the world reorganizes itself around the topic with a completeness no hobby has ever asked of anyone. People call these things shiny objects, which captures the arrival and misses everything that comes after it. Something shiny is a thing you glance at. This takes up residence. It sets its own hours and holds no opinion whatsoever about how many evenings a topic is worth, proportion being the faculty it suspends first.
I run a monitored brain, and I have written about the apparatus: the sleep tracked like a canary, the spring moods interrogated for velocity, the small black insects counted at the border of vision. The apparatus works. What it does not do is audit interest. There is no instrument pointed at what I care about, because sustained, methodical, productive interest in a technical subject is not a symptom on anyone’s sheet. It is my job description, and the same personality that built the career that pays for the pharmacy. I have written before that productivity is the best camouflage mania has, and I would revise that now, because camouflage implies something underneath is being hidden, and at baseline there may be nothing underneath at all. A fixation gets through the monitoring the way a man in a safety vest gets through a checkpoint, and the vest is not a disguise. He really does work here.
The trait is older than the treatment and it has run at every amplitude the treatment permits. At full amplitude, before the diagnosis, the fixations did not confine themselves to objects; the algorithm could acquire people, and the costs of those runs were not measured in evenings. The conduct of that season appears on this blog in exactly three words, risky and uncharacteristic, and the three words are still carrying the weight. The bargain I signed capped that amplitude. Last winter’s hypomanic burst got past the cap and produced a working programming language and most of a processor in a few compressed weeks. But that was an episode. It had a clinical name, a beginning and an end, and the people around me saw it eventually, because episodes surface.
What I am describing here runs underneath all of that, and it has no clinical name. It is the fixation that operates when nothing is happening. No episode, no warning from the people who know what to watch for, and no reckoning afterward, because there is no event to have an afterward. Over the years that setting has renovated a farmhouse, among other things. Lately it does its work in a text editor.
External ballistics was the original shiny object, the one from the tables in the manual, and late in 2024, thirty years after the rectangles, I picked it up again and could not put it down.
The return started as a passing thought, which is how they all start. The thought was: with AI handling the boilerplate, I could concentrate on the math. Anyone who writes software for a living will recognize the truth in this. Every side project of my life had eventually drowned in its own scaffolding, the build scripts and configuration and plumbing that stand between a person and the part he actually wanted, and by late 2024 there was a machine that would write the plumbing.
The thought was accurate. It was also the sound of a gate opening. Tedium had been the natural predator of every fixation I ever ran, the thing that ends them when the grind outlasts the glow, and I had just removed the predator from the ecosystem. A year later an actual episode would come through that same gate. In late 2024 there was no episode coming. There was a man at baseline with the friction gone.
I write, on my other blog, about Jevons’ Paradox, the nineteenth century observation that making a resource cheaper to use does not reduce consumption but increases it, the efficiency gains eaten by the appetite the gains create. The idea comes from the observation that increased efficiencies in coal-fired boilers did not result in less coal being used, in fact, quite the opposite. I have admitted on this blog that the paradox fascinates me more than it probably should. I can now report why. The paradox runs on fixation. The AI made every unit of progress cheaper. I did not spend less. The appetite ate the gain, and then it ate the evenings, and then it went shopping.
There were late nights that winter pinned to a single error, the kind every engineer knows, and this is where I have to report what the monitoring said, because I did check. Sleep is the canary in my coal mine and it is the first thing the checklist asks about. The nights read as ordinary engineering, which is not a failure of the audit but the finding itself. When a fixation lives inside the same behaviors as your profession, the instrument returns the reading it was built to return, which is normal, for an engineer. The canary is always fine. That is what makes this the one symptom the apparatus cannot see.
Somewhere in those months I also registered more than twenty domain names for a ballistics calculator that runs on my laptop. The project needed one. Arguably it needed zero. I have written before about the caching squirrel, single minded in autumn, burying nut after nut with no theory of why. The squirrel has a credit card, and the registrar stays open all night.
Then there was the rewrite. The first version was Python, a language entirely capable of computing where a bullet goes. I rewrote the whole thing in Rust anyway, which is what happens when the answers stop being the point and the instrument becomes the point. The arc of the bullet did not get more correct. It got faster than any question I was asking it.
Two years on, the thing is real. It is robust in the way I was trained to make things robust and thorough in the way I am thorough, which is the same thoroughness that once itemized an emergency room ceiling for a social worker. It takes ballistic coefficients and drag curves and produces plausible trajectories, or statistics about a shot that has not been fired. It models three degrees of freedom with extra knobs, which is most of the way to four, which is more physics than any deer has ever required. The repository is public, on GitHub, under my name, dual licensed under MIT and BSD, documented. Anyone can audit it, and every audit will return the same finding mine did: a healthy engineer doing healthy engineer things.
Here is where the thirteen year old comes back, because he had the method all along.
He wanted a curve he could not have. The mathematics that produces the exact answer was closed to him, so he did what an engineer does with a locked door, which is to approximate. Rectangles under the curve, error shrinking as the rectangles shrank, never zero, eventually negligible. Numerical methods are what you do when the closed form is out of reach, and most of them amount to the same confession. I cannot know this perfectly, so I will sample it finely and manage the error.
I have never had the closed form solution to my own mind. There is not one. What there is, is sampling. The sleep log is one rectangle. The mood check is another. So is the session every other week, and the psychiatrist’s questions, and the counts of small black insects, and the sentence said out loud to Melissa on the days that need it, which is half the protocol and always has been. None of them is the curve. Each is a small honest box laid against it, and the practice of laying them down, finer and finer, for the rest of my life, is the entire treatment plan underneath all the clinical names it has worn. The pharmacologists measure the medication itself this way: the exposure that matters is called the area under the curve, a coincidence so tidy I distrusted it for a week before deciding to keep it.
The fixation gave me the method that manages the fixation. At thirteen it taught me what to do about an unreachable function. You do not solve it. You box it in.
So which is it? Two years of steady work, public code, a documented repository, a project that looks from every angle including mine like the healthiest thing a person can do with an evening. Is that the narrow band holding, a lifelong interest finally running at a sustainable amplitude? Or is it the longest and best hidden run this fixation has ever managed, sitting inside the one behavior nobody will ever flag?
I do not know. The checklist reads green, the way it read green through the late nights and at the registrar and, for that matter, at thirteen.
Then I checked the mail.
Because the repository is public, people find it. Technical people, from around the world, write to me with questions about the software. Some have found bugs and were kind enough to tell me, which is its own small astonishment if you have spent any time on the internet. Others have folded the engine into projects of their own. Two people in Georgia, the country, are building devices you set downrange to read wind and conditions where the shot will actually travel, and they feed that data into a running instance of my engine.
None of that settles the clinical question. The quality of the work never settles it, and I have written a whole essay about a man whose work was extraordinary while his life came apart. But it settles something I can actually measure. The fixations I fear are the ones that cost people. Those run up debts and empty rooms and leave somebody worse off than they were before I arrived. This one fills an inbox with strangers being generous.
I am proud of it. I want that stated plainly, because the last time I wrote about pride it came braided with shame, the way it does after an episode, and there is no braid on this one. I have also spent years refusing, in writing, the idea that any of this is a gift, and I am not starting now. The tendencies were never optional. The object sometimes is, and that is not a gift. It is a steering input.
I keep thinking about the two Georgians and their sensors, because the image is doing more work than they know. External ballistics is the study of a flight already underway, and the one thing the solver cannot know from the bench is the air the flight will pass through. For that you need instruments downrange, positioned where you are not, reporting conditions you cannot read from where you stand. I have been running my life on that architecture for six years. Melissa is downrange. The therapist is downrange, every other week. The friends who once said dude, that’s crazy, and meant it as a warning, are downrange. The engine and the treatment plan are not metaphors for each other. They are the same design, arrived at twice by the same person, the way I arrived at the rectangles before I knew whose they were.
So I sample it. I lay the boxes against the curve, I take the wind calls from the people downrange, and the rectangles get smaller.
The error never reaches zero. It just stops mattering.
Quiet Mind, Loud Brain 
Leave a comment