Page 4 of 4
Re: CapLab & KI
Posted: Fri Nov 13, 2026 6:19 am
by deepCapital
arxbombus wrote: Tue Nov 10, 2026 9:26 am
I'm trying to figure out if you're completely mad or if it's the ramblings of an insane genius. Are you training a ML model or are you putting data into a LLM chatbot and asking it to analyze data for you? I'm also struggling to understand the hexagrid planets - very, very confused.
I also have a hard time following, but I always click the thread when a new post is made because I find the whole thing intriguing. I think the board game is separate ordeal. But the suggestions for Capitalism Lab is straight forward (CSV files for AI training data) and separate from the board game.
Initially, I thought CMcF was doing an interesting CapLab-to-board-game, digital-to-physical, "regional" gameplay (similar to SimCity 4 regional play) pipeline, wherein CapLab gameplay feeds the board game which then feeds back to CapLab which also then feeds back to the board game type of gameplay loop.
Which actually got me thinking: maybe Capitalism Lab should have a more detailed "regional" gameplay, beyond just having cities. Like maybe there should be "cross savefiles" that influence a new game somehow. I don't know if that's possible, but it's pretty interesting, because I do find myself reaching the endgame with so much cash and little competitors that I just start a new game.
Re: CapLab & KI
Posted: Fri Nov 13, 2026 8:35 am
by CMcF
Mea culpa
I am certainly to blame for the general confusion.
While I did mention my plan, I also overwhelmed everyone with nonsensical lists.
On top of that, I threw in mentions of board games and hexagons—a really poorly handled approach on my part.
So, let me start from the very beginning.
Once upon a time...
I’ve known about *Capitalism* for over 30 years. I’m currently pushing 70 myself.
Recently, I picked up *CapLab* along with all its DLCs, and at the same time, I began experimenting with emerging AI tools.
I noticed a button I wasn’t using—and I really don’t like that sort of thing!
It turned out to be for *.CSV files. I tried feeding some into the AI, and the results were fantastic. Naturally, I went on to input all sorts of *CapLab* CSV files. That’s how the idea was born: to get the AI to help me. I provided the AI with the game manual and instructed it to gather all available information on *CapLab*, including content from the forums and elsewhere. This created a data package that serves as the foundation for our collaboration. I’m also providing the AI with additional information through a large collection of screenshots and videos. Here are my game settings, which were, of course, listed by the AI:
Basics:
Difficulty: Custom 287% – Extremely difficult
50 competitors, all starting with low capital (~$20 million)
1 city: Mineral, fixed; no new cities (though the AI is permitted to found new ones)
1 seaport (consumption + 1 industry), import quality: tiny, constant import supply: no
Product customization: enabled, range: high
Industry-specific knowledge only: enabled, knowledge point system: disabled
Farm expansion: enabled, agricultural equipment: enabled (cannot be imported)
Banking DLC:
Bank interest rate: moderate (5.00%)
Insurance: disabled
Bonds: disabled
Global stock market: disabled
AI banks: tiny
Service Industry DLC – all enabled:
General business, shopping mall, import firm
Export duration: 5 years
Logistics: moderate
Freight costs: 100%
Digital Age DLC:
Software/Internet index: 100
E-commerce: enabled
Maximum share: moderate
Talent system: full
Experimental DLC:
City size: fixed
Integrated advertising unit: enabled
In short, I’m using all the DLCs, though not every single option—plus the "Real World Mod" 15.23.
The idea is for the AI to analyze the entire economy.
It can answer questions like, "Who sells the best computer, and at what price?" or
"Find the 24 most profitable goods," or "Which company is on the verge of bankruptcy?"
Essentially, it acts as my personal secretary—analyzing, evaluating... everything a computer is capable of doing.
And it all works perfectly.
The only snag is that I don’t have a CSV file covering all the goods—specifically data on purchases, sales, revenue, and pricing. There are 84 goods currently, but that number will grow, making the market increasingly complex to track. And if you add cities to the mix... things could easily descend into madness.
That’s why the "CapLab & AI" project has been put on hold for now—at least until a product list (in .CSV format, perhaps) becomes available.
So, what do I do instead? I took a board game and set the AI loose on it.
Since the rulebook for *Gaia* is quite difficult to grasp, the AI was initially intended to serve merely as an interactive, explanatory rulebook. But, naturally, that wasn't enough for me; I wanted the AI to actually participate in the game. To do that, it needs to be kept informed about any changes occurring on the game board. I’ve now managed to get it to recognize each of the 133 hexagons individually and correctly identify their contents whenever asked.
The AI is capable of interpreting an.d explaining the game’s symbolic language.
I’m curious to see if I might even be able to train the AI to act as an opponent.
And "train" really is the right word here—you can’t simply "program" a free version of an AI.
You have to go through the rather laborious process of coaxing it to do exactly what you want it to do.
If you’re interested, I can go into more detail about this process—it’s equal parts frustrating and amusing.
The next level would be how AI deals with luck and chance.
I hope this helps clear up some of the confusion I’ve caused!
And don’t forget: it’s all just a game, all just for fun.
I look forward to seeing any further contributions from you on this topic.
Re: CapLab & KI
Posted: Sat Nov 14, 2026 2:45 am
by CMcF
Good board gamers don’t have a "board in front of their heads"
But some board games have an incredibly steep learning curve.
Or perhaps a rulebook that is simply too difficult to grasp.
A single, small sentence—seemingly insignificant—can have a massive impact on the game, yet it just refuses to sink in. Both issues apply to *Gaia Project*.
At the very least, I’m still struggling significantly with a few of its mechanics. And I usually love reading rulebooks!
Let’s take a deep breath—let’s compose ourselves.
*Gaia Project* has a predecessor—one I didn’t care for back in the day: *Terra Mystica*. It was designed for at least two players and featured a sort of "deity track" that served no real purpose other than generating victory points. Dreadful. *Gaia Project* is a different story. While the core gameplay feels similar, it’s no longer themed around mystical deities; instead, it’s set in a sci-fi universe. That "ritual track"—or whatever it was supposed to be—has been replaced by a technology tree that offers actual rewards, progression, and victory points. That’s a major improvement.
And it gets even better: there’s an "Automa" (solo) version.
It was developed by some clever folks from Italy and is generally highly regarded.
I can’t personally vouch for it yet, though—I’ve never actually played *Gaia Project* myself.
So, this is a first for both me and the AI.
And since the game is about “power,” I suppose I should say: May the power be with us.
To be continued...
Edit:
(By the way, the game is available in many languages. It was even released in Japan at one point. I can’t say for sure if it’s still available in all those languages today. This is not a purchase recommendation or an advertisement!)