This is an old revision of the document!
Using BotW as a benchmark to compare
Game Mechanics
Replay
Generally does things well, slightly outstay it's welcome per area, base game v1 was perhaps a tad too big compared to botw which seems to have gotten the size just right to make players tapper out and run out of interest near the end while still ending at a still positive note. GI v1 felt like it outstayed that by a bit.
v2 and further, would probably feel better in that regard if I came at it from the start, finished the previous v in 3-4 months, and left starving for more. As I came in at v2.4 I already had a backlog, that grew with v3 coming out before I was done with v2, and same with v4 etc. But by that point other issues also appeared.
For the actual experience itself. The towers are ok, I'm kind of mixed on the whole system from different games (including botw), the teleporters are pretty common, chests and the various oculi are generally well paced and spread out. The actual environment/map itself tends to be well setup, I like the more natural feel and style to the layout. That you can often walk around to find easier ways and so on. Roads and paths makes generally sense.
In that regard I both liked and disliked v2's islands, as they make for good bit sized contents to engage with. But other parts of the game drags you around so much you're not allowed much time for that. And by v3 I was just barely getting done with v2 and just put entire v3 on indefinitive hold because I was sick of it.
Mechanics. some ups and downs, you get tired of the seelies in the long run, as well as many of the different types of puzzles like the torches and various timed elemental altars etc. And the world in general doesn't scale difficulty enough to really feel that engaging after a certain point in PP. Still over all manages to do it well, especially if you can keep it to more bit sized experiences instead of trying to dive into everything all at once. Dragonspine might still be my favourite, despite how annoying that cold effect can be.