[hnbad]

Anecdotally, "unique custom experiences" like this sound great on paper but fall apart in practice. This is not just true for games but also for other digital media:

I remember reading about various techniques to adapt online trainings to each learner's performance by offering additional exercises and content when certain performance thresholds weren't met and from a technological standpoint this seemed exciting but in practice it meant learners could not be reliably expected to have received the same knowledge (which is bad for any form of compliance) and their results might be difficult if not impossible to compare (which is bad for any form of analytics, not to mention actual conclusions about individual performance).

It's cool and flashy and you can make great demos but when it comes to actual hands-on experience, the drawbacks usually outweigh the benefits, plus you end up creating a lot of content most people don't actually ever get to see.

For games the only counter-example I can think of is cosmetic changes like Zelda TOTK's dialog lines changing to take weather, time of day and sequences of events (e.g. getting a fetch quest for an item you already have in your inventory) into account. But those are fairly cheap to implement, add minimal extra code paths (which is important for QA) and don't change the overall experience.