I was very much anti-AI. I refused to believe it could produce anything good and I was sure its results would be utter crap. I was wrong: some AI tools can help in some tasks. The challenge is figuring out how and when to use them.

Hominidae

All my writing before 2026 is 100% organic. No AI involved. But I started using LLMs to review my writing in early 2026. I still write all my posts and other documents myself and now I’m using LLMs only as a reviewer: to check for grammar mistakes (I’m not a native English speaker), too many repeated words, awkward and long sentences, sequence of thoughts, completeness.

I like to ask it for completeness of ideas. More than once it found a paragraph, or a section, that was missing an ending and guided me to improve it. This is not something I would have noticed and improved without external help.

And then I ignore all suggestions to use em-dashes, I don’t know how to use them anyway. Every time I asked an LLM to review something I wrote, it suggested that I use these goddamn em-dashes. It is annoying.

But this is a slippery slope. It is tempting to just accept the edited version by the LLM and call it a day. That is something I struggle with. I don’t want to vibe-write but some suggestions are too good to ignore. The boundary between human-written and LLM-rewritten is very fuzzy.

I feel like I learn something from every review by an LLM.

My writing is now ~95% organic. The remaining ~5% is English mistakes that an LLM fixed or a few small suggestions here and there. But a year from now, will this 5% grow? Will my personal satisfaction of doing it myself vanish in favour of supervised writing? What skill will I develop further: writing or prompting? If writing is thinking and I outsource it, am I still thinking?