A lot has been written recently about the development of artificial intelligence tools hitting a wall: that is, we’re reaching a point where the pace of improvement in models has slowed considerably. Some have made the point that this may not matter as we’re not yet close to exploiting even a fraction of what the existing models can do.

I think this is a reasonable take. One of my main uses of ChatGPT has been to help with the finer points of coding. I’m not really a coder, but I occasionally throw together a bit of PHP or Javascript to solve a specific problem: to tweak the output of an ical feed, for example, or to tweak the layout of a webpage, or to use an api to very quickly check public transport departures for a specific stop. This has been made much easier by being able to paste the code into ChatGPT and ask: “Why isn’t this working?!”

But just recently, I’ve been playing with the ChatGPT api and plugging in into some of those small scripts—with great results. When my alarm goes off in the morning, ChatGPT gives me a quick, sensible verbal briefing on my calendar events, tasks and so forth before I’ve even opened my eyes. I plug it into scripts where I’d like the wording to be a bit varied rather than identical every time, with pretty good results each time.

None of these things are lifechanging, but they are the sorts of small quality of life improvements that haven’t yet become commonplace—but will no doubt spread over coming years.