A useful analogy is natural selection. The two features, mutation and selection pressure, are present in software. From the inside, the mutations seem rational, controlled, planned, but from the outside they are random. The selection pressure is itself complex, but success can be measured by "number of copies in the world", which is ultimately driven by social success and luck as much as anything. Software also has a (unique?) "win more" mechanic where a popular tool tends to get more popular. The pressure happens recursively at different scales, and sometimes very large branches tend to die off because of big topics like "lack memory safety" or "supply chain attacks".

Interestingly, language diversity seems driven by school curricula, which becomes comfort which becomes hiring practice, the change driven by academic boredom with a given language.

> Dev Tools are the hardest business on earth

Don't most people say that about their niche of the tech industry? Healthtech. Online dating. Edtech. Game dev. You name it.

That part rings true, if slightly hyperbolic.

It’s like opening a restaurant that only serves other chefs. Not easy!

There is some truth here but I find this take to be overly cynical.

I built Cronitor in 2014 not because I liked making dev tools or was even particularly good at it, but because in my day job at Zillow my team had a lot of problems with silently-failing cron jobs. I had friends who were developers at other companies so I had the ability to validate this problem with other people in my network.

With any business it’s very important to solve a real problem. Steve Jobs called this approach starting with the customer and working backwards to the technology. So, you build what you know you need, with the belief that you are not as different from your peers as you might think. In contrast we have a lot of people starting with LLM technology and trying to work backwards to customers. I think that’s the real honeytrap.