
Polishing writing with AI · Part 2




I used a structure skill I built with an agent to fix the order and headings of 11 blog posts, but they were still hard to read. When I measured, the problem wasn't the order but how many names, numbers and sources were packed into every thousand characters. So I gave the skill the authority to cut paragraphs and a reader with no context who reads the post for the first time.

gstack deleted its own debugging skill after an eval showed no gain. Text alone doesn't change an agent much, and superpowers, gstack and pstack answered that problem in different ways. I don't read that as skills being useless. I read it as a sign that the work skills do is moving out of the text.

I polished my blog posts with a skill I built with an agent, and new AI tells grew where the translationese had been removed. A polish that deletes patterns fills the gaps with new patterns, and when the side that did the edits checks its own work, those new patterns are hard to see. So I changed the goal to my own writing's baseline and handed the meaning check to an agent other than the one that did the edits.