There is a price to pay when you use LLM generated articles. They may be functional, but they do not capture the human experience. I'm rewriting them all.

Ibrahim Diallo

A developer reflects on using DeepSeek to write blog articles last year and the dissatisfaction of reading them back. The post walks through the full process: the original prompt, the LLM's first draft, subsequent edits, and the published version. The author concludes that while LLM-assisted articles are functional, they lack personal voice and authentic human experience, and is now rewriting all such articles to recapture their original intent.

Editing my LLM assisted articles

I think the more one writes the better they become at it.
I also regret using AI too much in the wrong ways to write my early articles. Today i am still using AI, a lot, but i am using it mostly for feedback and grammar correction.
“Tell me why is this blogpost shit: [blogpost]”
“Please fix the grammar of that blogpost while keeping it verbatim”
I also use it as a crutch for areas where i am still learning, for example when coming up with a good progression, i tell it what i want to do and then ask it to suggest improvements.
What i completely stopped doing is asking prompts like “Please write me a blogpost about x from a y perspective”
I want my blog to represent my ideas and i honestly learn by bringing them to paper and i dont want to hand that off to AI.

When I was heavily writing (3+ articles per week), I was using CrewAI to do web searches, parse search results, and create article frameworks. Unfortunately, after the growing and egregious use of AI replacing authentic internet sources, I have had to step away from public articles and have focused on internal documentation and writing instead.