Byline.
← Notes · Feb 17, 2026

The Last 10%

WritingAI

People are single-shotting entire books from podcast transcripts now. Saw someone post an intricate prompt to convert an Acquired episode into a manuscript.

I can't even single-shot a 2,000-word article.

I've tried. Engineered prompts. Tested workflows. Gone down Reddit rabbit holes looking for what I'm missing. Watched the YouTube tutorials.

No matter how much I tweak the prompt, the output is close. But never the texture I'm looking for.

Till a few weeks ago, I kept agonizing. What's wrong with my process? Where's my killer-app prompt?

Only recently did it hit me: I'll never write a prompt that gets me there. Because I know what I want only when I see it. I cannot describe it before I see it.

The single-shot gets me to a draft that's usable, maybe 80% or 90% there.

Then I spend five hours on the remaining 10%.

Cutting. Replacing. Moving. Contemplating. Despairing. Rejoicing.

That 10% is where the essay stops being content and becomes mine.

And here's what I keep coming back to: I can't prompt my way through that last 10% because I can't describe what I'm looking for.

That's not a limitation of the tools. That's the definition of craft—the stuff that lives in judgment, not language.

Which means anything that's truly crafted cannot be single-shotted. Not now. Maybe not ever.

Observations from my failed attempts at shipping content faster.