Davos Couldn’t Write

I spent nearly forty years in corporate IT before I seriously started writing The Revenant Strain, and somewhere in the last few of those years I watched something happen that I haven't been able to stop thinking about since.

I had a project manager once — I'll call him Davos — who was one of the best I ever had. He could motivate a team that had nothing left to give. He could sit across from a stakeholder who wanted someone's head and walk out with the relationship intact. He held people accountable without making them hate him for it, which is a skill I've seen maybe a dozen people in my career actually have.

He could not write to save his life.

Status updates, meeting notes, executive briefings, escalations — all of it took him hours, and most of those hours happened after 5pm, because the writing was the part of the job he could never get ahead of. I spent a fair amount of my own time coaching him through it, editing his drafts, trying to get the instinct to transfer. It never quite did. The talent was real. The writing wasn't going to catch up to it.

Then he started using AI to do the first pass — meeting notes generated straight from the discussion, executive briefings pitched at the right altitude for the room they were going into, team updates that shifted tone and jargon depending on whether he was talking to engineers or to his own VP. The judgment about what needed to be said was still his. The hours it used to cost him to say it collapsed to minutes.

That was the last piece of his puzzle. He made Program Manager not long after.

A friend of mine — I'll call him Jacob — has spent his career as a senior graphic designer for a major American city, and his experience with AI has been almost the mirror image of Davos's.

The requests changed before his own work did. Over the last couple of years, senior administrators and city officials stopped sending him briefs and started sending him mockups — AI-generated brochures, campaign concepts, layout ideas, sent over with some version of "something like this." He says most of them are bad. They ignore the district's brand standards, the accessibility requirements, the visual language his department spent years building. They look, in his word, amateurish.

But he's not dismissive of what's happening, because the quantity and the shift in direction told him something the individual mockups didn't: everyone above him had suddenly decided they had opinions about design, and a tool that would draw those opinions for them whether or not the opinions were any good. It was a preview of where the requests were heading, and a department that ignored it was going to spend the next five years fighting its own stakeholders over amateur mockups instead of doing the work.

So his department went the other direction. They built AI into their own process — bound by their standards, their brand rules, their accessibility requirements — so that a city official's rough idea could be turned into something on-brand and usable in a fraction of the time it used to take. Same force multiplier as Davos, aimed at a completely different bottleneck.

Jacob's close enough to retirement that he could coast out the door without saying any of this out loud. Instead he's writing something close to a manifesto for his department heads, telling them, more or less: use this for what it actually is, or spend the next several years being run over by people who did.

I've watched some version of Davos and Jacob's story play out across an entire career's worth of meetings. Some people started showing up with AI transcribing the call, summarizing the action items before the meeting had even ended, drafting the follow-up email while everyone else was still typing their own notes by hand. Some people started using it to audit spreadsheets in minutes that used to take an afternoon, to tighten a presentation from bloated to sharp in one pass, to turn a rambling paragraph of feedback into three clean sentences a client would actually read.

And some people didn't.

The gap between those two groups didn't stay small for long. I watched colleagues who'd adopted these tools early pull dramatically ahead. They'd stopped spending their hours on work a machine could do faster, and started spending those hours on work only a person could do. The ones who sat it out didn't get worse at their jobs. They just got slower, relative to everyone else, in a way that compounded every week.

I'd seen this movie before. I just hadn't expected to be standing in the theater watching it happen in real time.

A Very Short History of Writing Getting Easier

Someone, a very long time ago, scratched a mark into a stone because they needed to remember something and their own memory wasn't good enough. That was the first writing tool, and I promise you someone complained about it — that carving your thoughts into rock was cheating, that real memory lived in the mind, not etched into a surface that would outlast you.

Then came papyrus, which meant you didn't need a chisel and a wall. Then scrolls, which meant you could take the writing with you. Then, for a very long stretch of history, armies of monks in cold rooms copying manuscripts by hand, one at a time, because that was the only way to make a second copy exist.

Then Gutenberg, and suddenly a single press could do in a day what used to take a monastery a year — and there were people, real people, who insisted the printing press would cheapen the written word, that something sacred was lost when a book stopped being a singular, hand-copied object.

Then the typewriter, which took handwriting out of publishing almost entirely. Then the word processor, which took the typewriter's permanence away — you could delete now, revise without starting over, move a paragraph without literally cutting and pasting it with scissors and glue. Then WordPerfect, then Word, then the rise of software built specifically for long-form fiction — outlining tools, character trackers, timeline managers, the whole apparatus of modern noveling.

At every single one of these transitions, the tool made writing faster, and at every single one of these transitions, someone insisted the old way was more legitimate.

Most of the time, they were wrong. The old way often carries real value, even beauty.

What If Tolkien Had This

I think about Tolkien a lot, for obvious reasons if you've read anything else on this site. The man built entire languages from first principles — grammar, phonology, the whole architecture of a tongue that had never existed — and then built a world on top of that world, genealogies stretching back thousands of years, appendices cross-referencing appendices, all of it held together by one man's memory and a truly heroic amount of handwritten note-taking, done over decades, without so much as a word processor to search his own notes.

What he did stands on its own as one of the great achievements of twentieth-century literature, tools or no tools.

But I do wonder, sometimes, what he could have done with something that could instantly tell him whether a name he used in an appendix contradicted a name he used in a chapter written four years earlier. Something that could hold the entire genealogy of Númenor in memory at once and flag the one place where two dates didn't line up. The book itself was always going to be his alone to write. What it could have freed up were the hours he spent on bookkeeping, so he could spend them on the parts only he could do.

Middle-earth was already whole. There might simply have been more of it.

Where I Actually Use This

I write grimdark horror in a found-document format, which means I am, on a very small scale, doing exactly the kind of continuity-tracking Tolkien did by hand — dozens of characters, precise timestamps, a Curator's voice that has to stay consistent across a hundred different documents pretending to be recovered from different sources at different times. LivingWriter has AI built into it that I mostly don't use for generating prose — I use it, and I use tools like Claude, as a tireless second reader: checking whether a date I wrote in chapter nine contradicts a date I wrote in chapter two, whether a character's weapon has quietly changed names between drafts, whether a phrase I just wrote is doing the thing every AI does if you let it — reaching for the same tired similes, the same "heart hammering" and "eyes like" constructions that could have been assembled from a thousand other books instead of written for this one.

That second part is the part people get wrong, and it's worth saying plainly: used carelessly, AI produces the laziest, most homogenized prose imaginable — competent, forgettable — interchangeable with the sentence next to it. I've had to build an entire set of rules into my own project just to keep it from sanding the voice off my own manuscript. Used carelessly in a corporate meeting, it produces a summary nobody double-checks, a presentation that's technically fine and says nothing, a communication that's been smoothed until it sounds like it came from no one in particular.

Used well, in either setting, it's the opposite. It's hours back. It's the tedious, mechanical parts of the work handled so the actual thinking — the parts that are supposed to be yours — get more of your attention, not less.

So — a genuine question, and I actually don't know the answer: how many of you write with spell-check and autocorrect turned off?

— Q.B.