I watched this video and was annoyed by what he said here:
Why would the base Nintendo 64 not be able to keep a minimal number of true/false flags in its memory, as is what Majora's Mask's time mechanics must amount to? Why would that require extra RAM or virtually any resources at all?
A SNES could do it. A NES could do it. A freakin' Atari 2600 probably could easily hold all of Majora's Mask's possible time-related game flags in its very limited memory!
Whenever Link enters a new "scene", the Nintendo 64 just looks up the current day and time in the game (probably just a simple integer ticking away) and minimal boolean flags such as "has Anju given Link the letter?" (0/1), "has the Goron postman fetched the Anju letter?" (0/1), "has Link rescued the bomb lady?", etc., and then simply place the 3D models wherever they need to be at that moment.
He makes it sound as if each of those NPC polygon models have an entire simulated "brain" constantly "thinking" (eating system resources) and act on their own and walk around and perform an infinite number of actions based on a highly sophisticated AI, or something like that, when in reality, it's a very simple set of simple flags which can be kept by any ancient computer's memory.
Where did even such a myth origin? I never heard any such claims from Nintendo back then. They talked about the extra RAM being used to display more sprites and effects on the screen at once as well as powering the "movement strategy AI" for boss battles (always in confined areas). They didn't use it for higher resolution and, obviously, not for the NPCs' alleged "intelligence".
And if there actually is something to this after all, then I'm more than eager to be blown away by the revelation that my favourite game actually has "thinking" NPCs which I've never noticed! It would be remarkable if that were the case, but I do not see how it could be.