Why My AI Wrote Badly in Three Languages (It Wasn't the Model)
I asked my AI assistant a fairly rude question this week: why is your writing bad in English, Chinese and French?
The answer wasn't the model.
Buried in my own setup were instructions written months ago, when the models were weaker. An email template that opened with "Hope you're doing well!" A rule that said always end with a call-to-action. Each upgrade gave me a better model, and those old instructions kept dragging its drafts back to 2025.
We deleted eight of these this week. The part that surprised me: the downloaded plugins were all fine. The rot was in the files we wrote ourselves, back when we didn't trust the model to write a plain sentence on its own.
So if your AI still sounds like a template, maybe check what templates you fed it before blaming the model.
Jury's still out on whether mine writes like me now. This post is the first test.
