Brendan Leonard’s “Do It 14,000 Times Slower With This One Trick” uses a deliberately overlong writing process to compare human authorship with text generated by large language models. The illustrated Semi-Rad essay, listed as updated Aug. 5, 2026, begins with Leonard’s account of seeing a social-media post that said, approximately, that ChatGPT could produce an essay in 15 seconds while a person might spend three hours writing one.
Leonard’s response is that his own first draft takes eight hours, not three. From there, the piece turns the gap into a series of charts and captions about procrastinating, typing, editing and worrying over whether the work is any good.
The target is not a measured test of ChatGPT’s output speed or writing quality. Leonard describes his own process, in comic terms, as including making food, taking walks or runs, sleeping, reconsidering sentences and seeking feedback from his editor, whom he identifies as his spouse.
What does Brendan Leonard mean by 14,000 times slower?
The number is best read as comic exaggeration rather than a literal performance claim. Leonard says an LLM could create an essay in 15 seconds and says his first-draft typing takes eight hours. Eight hours is 28,800 seconds, which is 1,920 times 15 seconds. A 14,000-fold comparison at 15 seconds would equal roughly 58 hours.
The figures therefore do not support a precise 14,000-times calculation, and the page offers no benchmark or controlled comparison. They do make the title’s joke clear: Leonard is describing creative work that occupies much more than the time required to place words on a screen.
After the first draft, Leonard says he may spend another 30 minutes to four hours second-guessing what he wrote. He then sends the piece to his editor for suggestions, makes changes and publishes it. The essay presents that loop as part of the work, not administrative debris to be optimized away.
What does the essay say separates human writing from LLM output?
Leonard’s contrast is qualitative. He writes that, by the time he finishes an essay, he has lived through the days, months or sometimes years that inform it. He also says he remains uncertain about the result’s quality, then closes by asking whether an LLM can have the conflicted desire to write while also complaining about the act of writing.
That leaves the piece’s answer properly unresolved. Both a person and a model can produce text, Leonard suggests, while the writer’s accumulated experience, revisions and doubt remain part of what happened before the reader got the finished page.
This story draws on original reporting from Daring Fireball.