The Forge Was a Side Effect

I have written somewhere north of six million words for money, all long before AI writing was a thing. Computer books, mostly, from the mid-nineties to 2010, the kind with a version number in the title and a shelf life just a bit longer than the milk in your fridge. People who learn this sometimes ask if I’d do it again, now. I used to give a wistful answer. The honest one, the one this whole run of essays has been circling without quite saying, is no. Not a chance. If you handed me a machine today that could produce those six million competent, on-deadline words while I did something else, I’d take it before you finished the sentence, and I’d whistle while walking away.

That should bother you, especially if you’ve read your way here through my other essays. The earlier essays all pointed the same way: stand on your taste when the machine comes for your work. Stand proud, you own the judgment the model can’t make. Relocate there. I believe it. I’ve bet my creative existence on it.

But I’ve never been straight with you about where taste comes from. And it’s kinda important, because the honest answer opens three cracks in the floor I’ve been quietly walking us around, and the last one is mine.

Sawdust

Taste came from the forge. The reps: the millions of words written badly, then less badly, the only known process for converting effort into judgment. I made that case about taste and the forge already in The Gate Was the Product . The thing is, I never ran the forge to build taste. That’s the part I kept walking past. Nobody does. I ran it because rent was due and the books had deadlines. I was busy building furniture. The taste was the sawdust, seemingly worthless at the time, the stuff that piled up on the floor while I made the thing I was actually being paid for. I didn’t even know it was valuable. I found that out decades later, when the furniture got old and rickety and ended up in flea markets, and the sawdust turned out to be the one thing in the shop the machine couldn’t manufacture.

Nobody set out to make this. It's what the work left on the floor.

That’s a weird admission. The most valuable thing I own was a byproduct of work I’d now refuse to do. The thing I spent all those years sweeping toward a dustpan is the only part that held its value. Which means when I tell a twenty-two-year-old to do the reps, find your taste, I am prescribing a forge I would no longer light for myself. It’s the cheapest kind of advice there is: free to give, expensive only to the person who takes it. Nobody runs a forge for the sawdust. I didn’t. None of us should pretend the kid will.

The forge wore a disguise

So where does the next taste come from? Before I can get there I have to back up, because I’ve let you believe a couple of things that aren’t quite true: that something as tidy as a deadline ran the forge, and that the forge was the writing.

What ran it was actually fear. I left a salaried job building software to freelance as a tech author, which meant trading a paycheck for the realization, renewed at the completion of every single book, that there might not be another one. I was good at the work and never once trusted it to continue. So I worked roughly twice the hours a sane person would, partly because I’d gladly do that to own my own clock, but mostly because terror is an extraordinary motivator and I spent fifteen years quietly terrified between contracts.

I never turned down a project. Seriously. Saying no to one felt like betting against the whole thing, so I said yes to all of it — overlapping books, articles, side gigs, whatever came — and from the outside that read as arrival. Name on a spine in a real bookstore; he’s made it. Inside it was a semi-desperate guy rowing hard to keep the boat from going under, taking the paycheck work to keep alive the chance of the work he actually wanted. Not even joking, I still occasionally have a nightmare and wake up freaked out that I’m chapters behind on a book deadline. What complicates any clean story I could tell about it: I loved it anyway. Both stayed true for years — the craft was a genuine pleasure and the grind was survival. The terror and the joy ran on the same engine.

The point about taste is that the bigger half of those hours wasn’t writing. For every book I wrote, I read ten — competitors, predecessors, piles of documentation, the people who’d already done the thing better than I was about to — because you cannot write six million words on a target that moves every eighteen months without reading several times that just to keep up with it.

And that ten to one reading ratio, I’m lowballing. If taste is caught more than taught, acquired from whatever work you surround yourself with — the case I already made about my own tells — then the reading was the real forge. More precisely, the sorting inside it. A thousand small verdicts about what to borrow and what to leave, what was load-bearing and what was padding somebody added to hit a page count; none of them recorded, most settled months later when the thing I’d built on them either held or didn’t. The finest point on that sorting is knowing when to cut. I killed 550 words out of this essay a few days before publishing it, then put a different 220 back. The writing was the part with a deadline stapled to it. The reading was what gave it its shape. Fear was what made me do years of both.

Which would be the escape hatch, if you wanted one. Reading builds the taste, the machine doesn’t touch reading, so relax and read on. I wish. I didn’t read those ten books for love; I read them because the eleventh one I owed a publisher wouldn’t write itself, and because the rent was a recurring threat. Take away the fear and the job it rode in on, and almost nobody reads nine dense manuals for the pleasure of the tenth.

And there’s a worse problem waiting for the reading itself. A forge made of reading is only as hot as what you read, and I happened to steep in human work since there was no other kind at the time: uneven, argued-over, copyedited, occasionally great, all of it written by people with something to defend. Taste needs a corpus worth catching it from. We’re diluting the source with the machine’s house style now, trained on the pile of us and settling toward the average. The canon hasn’t moved; you can still go find the excellent human stuff. But you have to choose it, against a current running hard the other way, and growing by the hour. Choosing the hard input on purpose is the same rare act I keep returning to and keep doubting anyone will perform. So the machine never has to take the reading away. It just changes what’s easiest to reach, and absorption does the rest.

That’s the first crack, and it’s a generational one. My taste was subsidized by an economy that no longer exists, one where a person with judgment and no shortcut had to grind out six million words to build a temporary career, and got the judgment thrown in. The kid behind me gets the shortcut on day one. Good for the kid; I’d take it too. But the shortcut routes around the exact grind that, in my case, was the only thing that ever built the judgment. We are about to find out whether taste can be acquired any other way. I don’t know the answer. I notice that nobody selling the shortcut is even asking the question.

The harbor doesn’t keep you

Suppose my way wasn’t the only way, and taste survives anyway, in some stubborn few who do the reps for reasons of their own. Then they arrive where I’ve spent this whole run of essays telling them to go: up on the judgment the machine can’t reach, above the waterline, somewhere the surface can move without taking them with it. Call it the harbor.

There’s a second crack, and it’s the one I find harder to admit, because I made it myself.

I had the thing I’m telling you to build. Name on forty-seven spines, fifteen years of standing, readers who came back. An audience. Then the industry that supported it started folding, and I moved quietly into something adjacent and stopped writing in public for fifteen more years. I didn’t decide to let the audience go. I never once thought about it, which is worse. I just moved on to the work that was available, and with the dawn of the iPhone that work became mobile app development.

The taste carried forward intact. It’s the only reason any of this exists. But the standing evaporated completely, and the second act never rebuilt it: the apps I built that found real users were team projects behind a bigger brand, and the ones carrying my own name never found enough of anybody to remember. So I’m sitting here with exactly the thing I keep telling you to get, starting over anyway. Taste and standing are not the same asset. One of them survives being ignored for fifteen years. The other one doesn’t.

Which I’d like to call bad luck, except I’ve now done it twice.

At least it was mine to get wrong. Where I went next, I spent that whole stretch watching the same thing happen to people who never got the option.

By 2009 the App Store I shipped into had hundreds of fart apps, and over Christmas 2008 the reigning one — iFart — took the #1 paid spot: a ninety-nine-cent soundboard sitting above every real piece of software on the chart, clearing 38,000 downloads on Christmas Day alone .

They were the slop of their moment, every one built by a human who paid Apple ninety-nine dollars and sat through a review queue to ship a digital whoopee cushion. People made slop by hand, gladly, the instant the cost of making it dropped far enough.

What the machine changed is the waterline. A fart app still cost a guy a weekend of actually building the thing, and that was the friction — small, but real, and it quietly metered the supply. The distribution gate never did that work and still doesn’t. App Store Connect is smoother than it was in 2009 and still a hassle, and I say that as someone who fights it every release. What collapsed is everything upstream of the upload button. Pull the cost of making toward zero and the supply runs to infinity while the queue stays exactly as long as it ever was. The genuinely good thing and the ten thousand things that resemble it now land in the same feed on the same morning, indistinguishable not because no one can judge them but because no one will ever scroll far enough to try. I keep saying friction is the whole subject. But I’ve mostly meant the friction that forms a maker. This is a different kind I hadn’t accounted for: the friction that used to protect makers by thinning the field. It’s gone, and the field is bottomless.

So when I tell you the standing doesn’t keep itself, I’m not theorizing. I watched a portfolio of my own work go under exactly that way, and the books followed on a slower clock: a tech title’s royalties rarely outlived its index by more than a couple of years, and today the whole shelf of them pays me roughly nothing. The forty-seven spines built the taste. They do not buy groceries. Whatever standing they bought me on the supply side — the credentials, the decades, the earned judgment, the walk-by at Barnes & Noble — is no life raft on the demand side, where the only question is whether anyone finds you in the noise. I’m not above this. I’m in it on the same terms as someone who started last Tuesday, minus the part where I already know how it ends.

This isn’t where I planned to land it. I pointed at a harbor, and it might be neither safe nor a harbor.

Getting in was never the hard part.

The one move left

There is one move, and I’m at least mildly worried it won’t hold.

If the forge only ever ran when something forced it, then the move is to run it when nothing forces it. To choose the difficulty on purpose, after the machine has politely removed every reason you had to suffer it.

I can tell you it’s possible, because I’m doing it right now. The terror is gone — between the studio and the tools, no single project is the last one anymore — and I’m working the same doubled hours I worked when I was scared, except the engine flipped. A force multiplier pointed at thirty years of backlogged ideas is the most intoxicating thing I’ve ever felt at a desk, and nobody is making me feel it. So the forge can run on want instead of fear. That’s the hopeful sentence, and I mean it.

Here’s the one underneath it, and I’ve changed my mind about it twice while writing this. The easy version is that I came to this desk with the taste already built and paid for, funded by fifteen years of fear-driven writing followed by fifteen more of software dev before any of this existed, and that what I’m doing at double speed today is spending that account rather than adding to it.

Except even that isn’t entirely true. I read more in a week now than I used to read in a month. Call it twenty to one, and it’s bespoke, aimed exactly at whatever I’m chasing that day, which the old forge could never do. Researching Nobody Told the Scribes I came out the other side knowing real things about a fifteenth-century abbot and the arrival of print, checkable things that stayed. That’s acquisition. Whatever else is happening at this desk, the forge is busy.

And the shape of the day inverted, which matters more than the volume did. Writing those books, the overwhelming majority of the hours went to rote: buying and reading a stack of competitors, searching, sifting API documentation, writing sample code to find out whether the documentation was lying to me. The thinking happened in whatever gaps were left. Now the rote is the smallest part of the job, compressed into minutes, and nearly all of the time goes to weighing what comes back: pressing on a claim, asking for the version that disagrees with it, following a thread three levels past where I’d have stopped back when it would’ve cost me a week. The research runs deeper and stranger than anything I could sustain alone. It is, functionally, what a writer with a research team has always had, handed to somebody who never had one.

That should be unambiguously good news. Most days I’m sure it is.

The rote was doing two unpaid jobs, though, and I’ve only ever written about one of them. In The Duck That Argues Back I described the first: the busy work was also recovery, the micro-rests where your hands stayed occupied and the back of your mind quietly untangled the hard thing. Strip it out and the day becomes unbroken judgment and you end it slumped, staring at a wall. Everyone I know working this way has landed on the same complaint.

The rote’s second job is the worse one, because exhaustion announces itself and a forge going cold does not.

Which is the third crack, and the one I’m still standing in. What I can’t establish is whether it’s the same forge. The old one made me sort, and the sorting included the junk: the chapter that turned out to be padding, the manual that was flatly wrong, the afternoon lost to documentation describing a function that had been quietly deprecated. I didn’t value a minute of that at the time. I’m not certain it was worthless. The new one hands me the sorted version. I still choose, and I’m choosing harder and at a higher altitude than I ever did with a stack of manuals. But I’m choosing inside a narrowing something else performed, and I can’t see what fell out on the way. Maybe I’m being flattered at speed by a machine with no stake in whether I’m right.

The multiplier works by closing the gap between what I can imagine and what I can produce, and that gap — what used to be the slow humiliating distance between taste and ability — is the exact place the scared years put me until judgment formed. So I choose difficulty and take the shortcut in the same motion. That looks like a contradiction from outside. From inside it’s a change of unit. I used to be the only pair of hands; the difficulty was doing every piece of the work myself, every word, every line of code. Now the pieces come back finished, and the difficulty I kept is the verdict on each one: hold it, send it back for another pass, or kill it outright. I gave up the gap and got back the thousand small verdicts that used to hide inside the doing, at a volume the old forge never ran. And nothing dilutes them anymore: every one is a real decision, the busy work that used to space them out is gone, and they do not let up. Would the version of me who started this way — handed the multiplier at twenty-two, never once afraid — have built any taste to spend? I might be one of the people I’m worried about. The only reason I can’t tell is that I got mine before the getting got easy.

I’m not taking the advice back — the alternative, betting on craft the machine matches in seconds, is plainly worse, and taste is still the best thing you’ve got. I’m taking the certainty out of it. “Find your taste or be swept away” was always half a sentence. The other half is that finding it may not be enough, and the path that built mine is being paved over while I recommend it.

Whether anything is piling up on the floor of this shop, I can’t see from in here. I couldn’t see it the first time either. The sawdust sat there for years looking like exactly what it was, waste, and I only learned different when the furniture wore out.

So I don’t get to close this clean. What’s left sits in two other essays: The Long Thank-You, about a debt that took four years to come back around, and The Tax of Mastery, about a guitar I have no business picking up at my age. Neither answers the worry. Both are reasons the fire’s still lit.