I mean, is HN really an unbiased sample source? Ask my mom or grandma if she prefers dark mode over light mode and she’ll say whatever she has right now on her phone.
Is that because your grandmother doesn't understand the question/choice, or did she evaluate the two options and decide which she prefers?
I think if I asked my grandmother a question like that she'd probably get concerned that I was about to change something she was already accustomed to, without necessarily understanding what change was being proposed, and would default to invoking the 'don't fix it if it ain't broke' principle.
My mother on the other hand is younger and still has an adventurous willingness to try new things, and last I saw was using the dark mode in Windows 10.
It's not surprising that power users are more acquainted with the options available with a developed preference from trial and error.
Polling someone who likes light themes because they know no better is an answer to a completely different question.
Also, while trendy, "bias" isn't a word you should throw around unless you also have established the question being asked and what the bias is. Informed answers to the poll aren't biased, especially compared to someone who doesn't know what a dark mode is.