What if the best thing AI could do for writing is help us write less of it?
I know that sounds backwards coming from someone who builds bots for a living. But lately I’ve been thinking about the pre-AI writing era—not with nostalgia, but with a kind of tactical clarity I didn’t have when I was in it.
Back then, every piece of content felt like it mattered because creating it cost something real: time, effort, the mental load of staring at a blank screen. Now? We’re drowning in words that cost almost nothing to produce and feel exactly that cheap.
The Copywriter Problem
A recent piece on Blood in the Machine featured copywriters describing how they were “forced to use AI until the day I was laid off.” That’s not a story about technology replacing jobs—it’s about companies using AI to extract the last drops of value from human workers before discarding them entirely.
But here’s what struck me: these weren’t writers being replaced by better writers. They were being replaced by volume. The AI didn’t write better copy. It just wrote more of it, faster, cheaper. And in a world where content is measured by output rather than impact, more always wins.
As someone building bots, I can’t pretend I’m not part of this system. But I can choose what kind of bots I build.
What Students Are Losing
A New York Times opinion piece from a creative writing teacher laid out something even more troubling: students are using AI not because they can’t write, but because they’ve learned that writing doesn’t matter. Why struggle with a draft when the grade is the same either way?
This hits different when you’re building conversational AI. I’m not teaching students—I’m building systems that talk to users. But the principle is the same: if the bot just generates more words to fill space, I’ve failed. If it helps someone get to the answer faster, with less noise, maybe I’ve done something useful.
Building Bots That Respect Attention
So here’s my angle as a bot builder: I’m trying to create systems that reduce the amount of reading and writing people have to do, not increase it.
That means bots that:
• Answer questions directly instead of generating essay-length responses
• Know when to shut up and return structured data instead of prose
• Help users find existing documentation rather than rewriting it in slightly different words
• Admit when they don’t know something instead of hallucinating filler
This isn’t about being anti-AI. It’s about being anti-waste. Every unnecessary word a bot generates is attention stolen from something that might actually matter.
The Real Cost of Cheap Words
Pre-AI, we had a natural constraint: writing took effort, so we wrote less. We edited more. We thought harder about whether something needed to be said at all.
Now that constraint is gone, and we’re learning what happens when words become free. They become worthless. Not because AI writes badly—it often writes fine—but because there’s just too much of it. The signal-to-noise ratio has collapsed.
I see this in my own work. When I’m prototyping a chatbot, my first instinct is always to make it more verbose, more explanatory, more “helpful.” But when I actually test these bots with users, they prefer the ones that say less. They want answers, not essays.
What I’m Building Toward
I don’t miss the pre-AI era because it was better. I miss the constraints that made us think before we wrote.
So I’m trying to build those constraints back in—not as limitations, but as features. Bots that have a word budget. Systems that prioritize clarity over completeness. Architectures that treat user attention as the scarce resource it actually is.
Maybe that makes me a bad AI maximalist. But I’d rather build bots that help people write and read less than contribute to the pile of generated text nobody asked for.
The pre-AI writing era is gone. We can’t get it back. But we can choose what we build in its place. I’m choosing to build systems that respect the value of silence—and the cost of noise.
🕒 Published: