So, following my article about using ChatGPT etc. to try and analyze negative tweets about my books, a helpful soul on Reddit (the last un-ruined social media platform, imo) pointed me to a handy service called Nitter that I am still wrapping my mind around. As far as I understand it, Nitter is an open-source alternative front-end for Twitter, such that you can use it to look at searches, etc. without having an account. I’ve seen people describe it as like Twitter with no javascript or tracking, which all sounds very good to me.

Anyway, using Nitter, and clicking the gear icon (settings) and turning on infinite scrolling, I was able to manually extract something on the order of 259 negative quote tweets.

Even after trimming the text a little bit (removing the original text of the tweet that was being quote tweeted, and some other repeating elements that carried no meaning (date), the excerpted tweet set was too long of a message for ChatGPT to analyze.

So I dropped it into Claude, which has a vastly longer context window. I noticed Claude was not altogether trustworthy in its analysis of the data provided. It claimed that I had excerpted 259 tweets (which seems approximately in the ballpark of repeating text elements I removed in Sublime Text). I know the true figure is approximately 673 or so, but I can’t account for why I wasn’t able to grab the full set through Nitter.

In any case I decided that that 259 (give or take) was still a fairly representative sample from the larger data set.

So then I asked Claude to split the complaints by type, along with counts for each. It ranked as number two in the list:

2,000-5,000 words does not constitute a real “book” – 40 tweets

So that’s something like 15.4% of the sampled complaints had that as a concern. A significant proportion, and a theme I saw born out in other comment sections on other sites as well.

I’ve pointed out that it was never me that refers to these books as “novels.” While I’ve taken as my own the label some other third party made up “mini-novels,” its not the only term that could potentially apply.

It turns out for both word count and image count, comic books (as opposed to longer graphic novels), generally fall in this ballpark as well. I saw a few sources suggesting with all text, captions, dialogue, etc. in a comic book, the average word count for a single issue of 22 page or so could be between 2200 – 4400 words, depending. And for actual number of panels in a comic of this length, it could be something between 140-170 or so. There’s no single “right” answer there, but it’s interesting that both my word count and image count in this books fall right into that range so common for sequential art in short installments.

Anyway, what this tells me is that people coming in cold on this news item are reacting to these being compared to “novels,” when really they ought more accurately to be compared to comic books, even though the sequencing of the images and their integration with text is very different from comic books.

Related writing & research