If the scope is solidly defined, this isn't a bad idea; cramming everything into RAM in one go is a perfectly acceptable approach. I do have mild worries about how brutal the intialisation phase is, but I'm assuming it's streaming from the HD rather than from a disc, so that should alleviate things considerably
That said, 'if the scope is solidly defined' is important; it doesn't leave him with much in the way of leeway. I've worked quite close to RAM limits on a few games, and it can get to the point where you get terrified to add anything new because it might push you over the limit. Hopefully he's got a comprehensive and accurate memory budget laid out.
It's a pain to write good, solid streaming code; I worked on the PSP version of Gun, and we had an awful lot of niggling bugs in the streaming code which required a lot of work to smooth out. That said, if you *do* nail the streaming code, suddenly you're able to produce environments considerably larger than the memory limitations allow. A The Witness produced for a system with a smaller memory pool - so they were forced into having to make streaming code - could in turn be significantly larger than this one...
...but there's no point in doing that if you don't have to. Strictly speaking, it's not a particularly economical approach to memory usage - but it's an extremely sensible, extremely economical approach to time usage, and that's something that it's sometimes hard for the end viewer to comprehend and appreciate.