Advent of Code is a yearly series of 25 programming puzzles that get harder as you go. The presenter used the first puzzle to test an AI coding assistant called windsurf, learning both what the problem required and what the tool could do. Using a standard challenge like this gives you a systematic way to evaluate AI capabilities.
Check out the Advent of Code. There is an exercise that happens like every year for the last decade, which is called the Advent of Code, which is going to be 25 puzzles. They're put together, they're just brilliant and challenging. The first few are easy, and then they get really, really hard. And I think people who are interested in building, even people who wouldn't describe themselves as coders, would get a lot from playing with this, if not to learn these data structures and algorithms type principles behind the scenes to learn how to get the AI to build some of these things because it actually makes for some pretty bite-sized projects that AI can do a really good job with. I experimented with that on day 1 of Advent of Code in this particular year. And I found that I learned one, a lot about the question that was being asked. But I also learned a lot about like what windsurf can do. By using and evaluating tools through this kind of, you know, standard mechanic, you get a repeatable process, you can gain a whole lot of understanding what these tools can do that you couldn't do before.
