I asked an LLM to name my variables. It was worse than me.

I had a theory: if large language models are trained on all the code on the internet, surely they've seen enough good variable names to fix my worst habit.

The experiment

I fed the model a gnarly function and asked it to rename things for clarity. It returned handleStuff, tempResult, and — my personal favourite — dataData2Final.

Reader, I have written dataData2Final. The machine had simply learned from the best (worst).

What I actually learned

• Models are mirrors. Garbage naming in, garbage naming out.
• The fix for bad naming isn't a smarter autocomplete; it's caring two seconds longer.
• Also, final never means final.

Anyway, I renamed it to normalizedOrders. The model agreed it was better. We're friends again.