One of my dream games is a truly open world text adventure. I got a glimpse of it by having ChatGPT run this game, but it started hallucinating and misremembering after a few rounds. It has to be perfect to avoid breaking the immersion, but I'd pay $100 for such a game even without graphics.

▲

alisonatwork 6 hours ago | parent | next [-]

Isn't this what MUDs are? I tried a few in the early days of the internet and even back then they were like much bigger and more dynamic versions of text adventures of the 80s. For me I bounced off the idea that I had to role-play with other humans - I thought it was far more interesting to chat with other humans about real-world topics - but if you are looking for a large, text-based role-play experience then it's probably worth trying out a few. There might even be some that can be soloed these days, there are so many.

I think the challenge of trying to make an "endless" game using an LLM is the same challenge that all procgen games face - they are boring for people who are seeking a well-paced narrative. There are players who enjoy the mechanics of looting/crafting/trading/etc who will gladly play games where the story is incidental or emergent, but if you're specifically looking for something with a bit more narrative depth, I'm not sure procgen will ever work. Even if there is a system that tries to project coherent storylines onto the generated world, you still need the player to do things that fit into a storyline (and not break the world in such a way that it undermines the storyline!), otherwise the pacing will be off. But if the system forces the player into a storyline, then it breaks the illusion that the world was ever truly open. So you can't have it both ways - either there is a narrative arc that the player submits to, or the player is building their own narrative inside a sandbox.

AAA games try to have it both ways, of course, but it's always pretty clear when you are walking through procgen locations and leafing through stacks of irrelevant lore vs when you are playing a bespoke storyline mission that meaningfully progresses the state of the world.

▲

JKCalhoun 4 hours ago | parent | next [-]

What I wanted in MUDs was a simple editor to allow people with little technical skill a means to create a world—or extend an existing one. And then I wanted a way to join MUDs together—like if you leave a forest by a certain path you are, unbeknownst to you, rerouted to a different MUD that picks up where the forest left off.

In this way I imagined in time a world larger and richer than any that had come before it—where you could really just keep going, keep playing, never see all of it.

▲

breve 12 minutes ago | parent | next [-]

> What I wanted in MUDs was a simple editor to allow people with little technical skill a means to create a world—or extend an existing one.

Those are MOOs. They're fully programmable in MOO code. Here's the original MOO: https://lambda.moo.mud.org/

There's no point to a MOO other than to be itself, although LambdaMOO does have an RPG system in it you can play: https://en.wikipedia.org/wiki/LambdaMOO

Server resources: https://www.wrog.net/moo/

Programmer's manual: https://www.wrog.net/moo/progman.html

yduJ's venerable duck tutorial: https://jkira.github.io/moo-cows/docs/tutorials/wind-up-duck...

▲

alisonatwork 4 hours ago | parent | prev [-]

I never got deep into it, but I remember reading magazine articles back in the 90s that that's exactly what the new generation of MUDs were. Wiki has pages on MOO, TinyMUCK, MUSH etc - these are basically platforms where the players themselves can expand out new objects and locations, presumably in a similar way to Second Life or other MMO sandboxes do today.

So the tools already exist, but it seems to me that they primarily appeal to a very specific type of gamer, one that doesn't have much overlap with the type of gamer who would like an "endless" open world or the type of gamer who would like a tightly-plotted narrative experience. I think it's more something that appeals to fans of table-top RPGs, people who are looking for a collaborative storytelling environment.

I think many gamers have the imagination of an epic infinite metaverse style game, but then when they actually get the opportunity to participate in one, it turns out that that's not really what they wanted after all, because it requires a level of creative labor that they weren't expecting. This is why I think the market has naturally segmented into sandbox builders, survival/roguelikes, traditional narrative adventures etc.

	▲	egypturnash 3 hours ago \| parent [-]
		My experience was that in practice all that mapped-out world of most social mu*s was largely ignored by players; they'd all end up in a few gathering spots, or in private spaces disconnected from the main map, open only to their owners and people they teleported in.

▲

shevy-java 5 hours ago | parent | prev [-]

> chat with other humans about real-world topics

You can do this with regard to a MUD too, but typically out of character and not every MUD would allow OOC chatting within the game world, as that is disruptive to those players who seek immersion.

It seems to me as if you may not have found a good roleplaying MUD back when you played MUDs. You may be missing out on that experience. I retired from playing MUDs about 11 years ago permanently, but the in-world roleplay was the only thing that was interesting to me since it was the creation of a unique storyline potentially involving many other playercharacters.

	▲	alisonatwork 3 hours ago \| parent [-]
		I think I just don't really vibe with roleplaying in realtime with other humans, to be honest. I grew up trying to play tabletop RPGs (my dad was a DM and used D&D mechanics as a way to make storytime more engaging), but while I really enjoyed making up characters, I never had much fun actually doing a campaign. The thing I love about computer games is that I can go through them at my own pace, pause whenever I like, hang around looking at a cool visual, go back to an old save and try something different, whatever. Multiplayer takes all that freedom away because everything has to progress on somebody else's timetable, which isn't as fun for me. Nowadays being expected to perform on a time limit just reminds me of work, which is the last thing I want when I'm playing a game.

▲

TylerLives 7 hours ago | parent | prev | next [-]

Instead of relying on the model's memory alone, you could have it read/write to a file.

▲

hackshack 8 hours ago | parent | prev | next [-]

You're on to something. I tried this too, a few months ago, with offline Ollama/Magistral on Mac. "You're a dungeonmaster for a single player adventure game, with me as the player..."

It lost track of things almost immediately. But the foundation was there.

Maybe if we had a MUD-tuned model...

If it has an approximate way to track state, and a "pre-caching" method where it can internally generate an entire town all at once, room by room, so hallucinations are rarer... actually starts to sound like a traditional DM's method of world building for a campaign.

Maybe something like an LLM-assisted Inform (interactive fiction engine). https://ganelson.github.io/inform-website/

Side note: been playing Aesir, then the Aesir 2 MUD since 1994. It's still up!

	▲	c22 3 hours ago \| parent [-]
		Maybe you could ask an LLM to build a campaign then ask another one to run it.

▲

mynjin 5 hours ago | parent | prev [-]

Would Avalon count?

https://www.avalon-rpg.com/intro/mud