you are right, and that has been bugging me from the start and never got to fixing it. I tweaked it a bit to make it hopefully make much more sense now. Also I didn't know enclose.horse, and loved playing it, so thanks for pointing it out and for the nice feedback.
You are definitely correct, and that is how I am solving all the puzzles as well. However, I've encountered many people that couldn't wrap their head around even the mirror across the fold line logic. For such people, the techniques I described help to come up with puzzles that feel "hard" for them. Thanks for giving it a try.
Author here, happy to answer questions about any part of this. The scoring weights are hand-tuned from playtesting, not learned from data. Would love to hear from anyone with psychometrics or spatial cognition experience.
Microsoft’s real goal is selling Copilot seats and pushing Azure, not building a neutral playground for third-party agents. There is just no money for them in being the backend for someone else's AI.
As for the AI spam, GitHub's internal metrics have always been tied to engagement and PR volume. Blocking all that AI slop would instantly drop their growth numbers, so it is easier for them to just pass the cleanup cost onto open-source maintainers.
Really cool idea, however as a product developer myself, I know how hard it is to hit the ground running especially for a product that requires an already existing audience to function. I tend to lose myself in the coding and design part, while not realizing that the 90% of the real deal is actually marketing. You have to delve much more into ways to market the idea, rather than develop it.