The basic premise is to solve a medium-difficulty-puzzle, while spending a lot of attention on noticing exactly what your brain is doing at every step. You can then notice which pieces of your process are doing most of the work, and which are wasted motion.
This is not only valuable for improving your puzzle-solving abilities, but for generally improving the feedback loops that improve your cognitive abilities. I've found this to be one of the most essential rationalist skills that I've learned.
We'll be meeting in the Lecture Hall in the Walled Garden, at noon PT. (Follow this invite link to enter the Garden. Follow the path to the Tardis, and then enter the "Bacon Lecture Hall")
I'll give a short talk on the theory-and-practice of tuning your cognitive algorithms. We'll spend two 20 minute periods doing individual exercises, and then talking about what we learned.