29 Apr 2025
I hadn't really intended this blog to become a "week notes", but I can't say I'm against it. This blog is a little late, on account of a hellish Saturday evening that had me working until 4:30am. Sunday is a blur.
I had thought this blog would have more technical content in it, but I've found my thoughts drawn to other things. A little while ago in Brighton a had a "mulberry crumble". An unbelievably delicious dessert, that started off a minor obsession myself and my partner with mulberries. The first thing to know is that, despite the nursery rhyme, mulberries grow on trees rather than bushes. They aren't native to our shores, but were brought here to try to stiumulate a silk trade as silkworms enjoy mulberry trees.
That effort did not pay off, and unfortunarely a lot of mulberry trees died with their efforts. But in a way mulberry trees have been following me and my partner for a while. Nearby Charlton became the centre of a lot of mulberry tree growing, and there are great mulberry trees to see at Charlton House, Greenwich University's Dreadnought Building at Lesnes Abbey.
There is a huge amount of information on London's mulberry trees on the fantastic Morus Londinium website. They even have a map to find your nearest mulberry tree! One of the most impressive features I've seen of black mulberry trees is there ability to completely grow back, even if cut right back into a stump. You can read more about these "phoenix trees" here. I do wish I had the resiliency of a mulberry tree.
One sort-of tech related thing I can say about last week is how I fear that AI might mean a sharp downturn in the quality of documentation. Let's face it: not all open source projects have great documentation as it is. But I had a situation over the weekend where what appeared to be a very simple setup that should have been just a straightforward change to configuration became a drawn-out battle between human and software.
The documentation included no examples, and the tutorials in the documentation created a configuration that the software complained was deprecated. The only way I found to get it to work was to bash it through a bunch of GitHub Copilot prompts until it worked. That's a terrible experience. I don't really want to support AI, I want to learn for myself, but and if the documentation is so bad I'm forced to turn to AI then I think that is an indictment of the documentation.
Unfortunately, I fear developers will neglect documentation even more if the expectation is that an AI can figure it out for you. Why bother to create a useful set of documentation with tutorials, references and how-tos if nobody will bother reading it and instead will get an AI to explain it to them instead? Are we just writing documentation for the consumption of robots now?