This post was really easy to get into, was fluently written and generally was engaging to read.
It actually taught me a potential new mental motion, or at least heavily clarified one, in the way it illustrated how bottlenecks in mazes influence the optimal strategy. I’ve done an undergraduate in CS, and so done lots of path-searching problems, and was familiar with the basic concept, but this connected my abstract idea of identifying bottlenecks with my intuitive planning sense.
Great use of pictures and visual media. I have a guess that we are lacking that on LessWrong, and that a lot of our posts could be made a lot better with more diagrams. And for that I also liked how the diagrams were kinda janky photos of pieces of papers with crayon on them, because it set an example of how you can add diagrams to a post without needing to put in a ton of time and effort in illustrator or something.
Cons:
I think I would have loved a bit more generalization of the concept of mazes. There is the difficult question of what problems are actually shaped like the maze problem analyzed in this post, and I think the biggest thing preventing me from using the ideas in this post, is a lot of uncertainty about when it is actually a good idea to apply them.
This is good as a stand-alone post, and I think it’s valuable to have stand-alone posts, but I am also always in favor of linking to existing discussion of a concept, that we’ve had on LessWrong. In this case links to the section of the sequences were Eliezer talks about a bunch of graph-search algorithms, and some of the existing writing on chunking would have been great, but I understand that finding those posts can be quite a bit of a pain.
Promoted to curated, here are my thoughts:
Pro:
This post was really easy to get into, was fluently written and generally was engaging to read.
It actually taught me a potential new mental motion, or at least heavily clarified one, in the way it illustrated how bottlenecks in mazes influence the optimal strategy. I’ve done an undergraduate in CS, and so done lots of path-searching problems, and was familiar with the basic concept, but this connected my abstract idea of identifying bottlenecks with my intuitive planning sense.
Great use of pictures and visual media. I have a guess that we are lacking that on LessWrong, and that a lot of our posts could be made a lot better with more diagrams. And for that I also liked how the diagrams were kinda janky photos of pieces of papers with crayon on them, because it set an example of how you can add diagrams to a post without needing to put in a ton of time and effort in illustrator or something.
Cons:
I think I would have loved a bit more generalization of the concept of mazes. There is the difficult question of what problems are actually shaped like the maze problem analyzed in this post, and I think the biggest thing preventing me from using the ideas in this post, is a lot of uncertainty about when it is actually a good idea to apply them.
This is good as a stand-alone post, and I think it’s valuable to have stand-alone posts, but I am also always in favor of linking to existing discussion of a concept, that we’ve had on LessWrong. In this case links to the section of the sequences were Eliezer talks about a bunch of graph-search algorithms, and some of the existing writing on chunking would have been great, but I understand that finding those posts can be quite a bit of a pain.
Overall, very happy about this post.