Definitely, games that try this do need to abandon traditional narratives to some extent, which is a big ask for a typical publisher - and unfortunately, the complexity of such systems also puts them out of the reach of indie games which can otherwise afford to be more experimental (although Rain World is supposed to have some pretty amazing AI simulation too) - these kind of projects really have to be AA games, big enough to manage the complexity, but small enough to be allowed to actually do it instead of flattening everything in pursuit of more market share. And the AA side of the industry has unfortunately massively declined.

I tend to disagree. Dwarf Fortress was made by 2 people. The biggest problem is that the cap-ex of discovery of how these systems should work makes it very difficult to attach the rest of the "game design" portion of the industry to it. Companies make idiotic time tables, a company would not be able to release Dwarf Fortress even with a AAA budget and the game as it exists today simply because there'd be too many meetings for every time you realize a disastrous edge case in the simulation causes rework and scheduling delays for all the assets.

Also just as an aside I think a lot of this is hella worse during post-covid, not just from a cost of doing business POV, but from a explaining basic concepts to non-technical and technical people alike POV. I'm having a really rough go of it at my current company trying to get PMs, UX/UI designers, and even engineers wrap their heads around what it means to user interaction to have asynchronous jobs running in the background providing the functionality that was previously fully synchronous.

I'm having a hard time in the year of our lord 2024, explaining what multi-seat /collaborative data entry is, and how we need to start thinking about it instead of writing software for the 2000's.

In my world an example would be, last year I rebuilt our data modeling system so that we can introspect our models, tag how their properties work in standard data functionality throughout the system, automatically generate types, type guards, validators, and factories as well as create generic functionality that uses the model tags as well as all the developer guard rails of linting organization, etc. This all supports structural typing, and is fully declarative. I did all this in 3-4 months. Which has been one of the fastest projects of this size/complexity that I've done in my life.

Every 2 weeks on the project I'd be in a meeting with people asking me if it's done yet, and what's plan b? The goal of the project was to support a bulk export system for arbitrary data in our system. There was no plan b. The only realistic plan b was to write import export functions for each and every data component that was used which would be slower to implement, less consistent and ultimately not fix our underlying issues that the modeling was crap. But there's always the dumb guy question of "why is this not done yet?" and those can be asked basically for free by any available dumb guy.

That's why people don't take risks, interesting things don't get built. You get just get burn out as a reward.

Grand strategy games like Crusader Kings also utilize such dynamic systems to great effect while forsaking typical narrative, but they’re their own niche (and while Paradox does make cool stuff, they do also nickel-and-dime people for like 100 different DLCs per game…)

Honestly I don't really mind this... the DLCs go on sale a ton, and it's honestly the only realistic way to fund a modern game. Otherwise games just wouldn't get updates free or otherwise. We can argue about price points or the efficiency of the way these games are made and designed and what that means to the theoretical cost of upkeep (e.g. DLC) but I'm gonna have to tap the sign:

the reality of software is that software is made in spite of the management of the business. That’s always the #1 hurdle.