For a while last year, as a hobby, I spent many evenings browsing old issues of The New Yorker online. Some articles I read, some I skimmed, most I skipped, and many I kept a note of to maybe read later. As I did this, I also kept a note of the very long articles, just out of curiosity, and made a reasonably comprehensive list of them – see this earlier blog post.

The other day, I was reading an article in the London Review of Books. I noticed that the article had a wordcount at the top – as do all LRB articles. Indeed, their whole archive back to 1979 is properly digitised with helpful wordcounts. So it should be possible to make a “longest articles in the LRB” list not by spending months browsing back issues but automatically, by writing a little computer program to go through all the issues of the magazine, then all the articles in those issues, and text-scraping all the wordcounts.

Unfortunately, while this is not really difficult, it’s not something I can do. (Most of our MSc Data Science students learn how to do this, but, slightly embarrassingly, I’ve just never learned myself.) However, there’s a tool now that could write the program for me – ChatGPT. I told ChatGPT what I wanted doing, what the URL structure for issues of the paper is, and where in the HTML for each article the wordcount can be found. Then ChatGPT wrote me a little Python script that goes through them all and writes me a textfile with all the wordcounts of every LRB article. Once I’d installed some Python things on my computer, the script worked perfectly first time. Thanks, ChatGPT!

This is the first time I’ve used one of these AI things to actually do a useful task for me that I genuinely wanted doing (rather than learning how to use it, just playing around, or getting it to make a picture that I wanted to look obviously AI-generated). I’m very sure it’s not the last time, but rather it was my first little glimpse of what the future might be like.

Anyway, if I’ve done it right – and if ChatGPT’s done it right – this should be a list of every LRB article longer than 25,000 words (this century) or longer than 15,000 words (last century):

I didn’t count Alan Bennett’s series of annual diaries (36 articles, 1984–present, 200k words, books: Writing Home, Untold Stories, Keep On Keeping On, House Arrest)

I didn’t count Jenny Diski’s memoir series about cancer, Doris Lessing, and other things (17 articles, 2014-16, 80k words, book: In Gratitude).

The longest fictional piece I found was The Uncommon Reader by Alan Bennett (2007, 23k words, book of the same title).

Art Spiegelman’s comic In the Shadow of No Towers was serialised in the paper (10 parts, 2003).