• #2

If you don’t know about LudoForge, you should probably read the previous posts. In summary: I’m developing an app to evolve tabletop game prototypes according to a “fun” factor made up of plenty of fun proxies like player agency, strategic depth, etc. The code is now mature enough to run X evolutionary generations on demand, produce shortlists of the best games, and finish. That’s great because it means the code works well, although there’s always room for improvement. But the shortlist of four winners is absolutely terrible in a way that has had me giggling for a while. I’ll let ChatGPT explain these “games.”

---

Game 1: “The Shared Counter Chicken”

Two players take turns. There’s a shared number on the table that starts at 0 and can be increased up to 15. The “win condition” is simple: whoever makes it hit 15 wins instantly.

The game also includes a couple of other buttons that change other numbers… but those numbers can never actually produce a win. They’re basically decorative dials. One of them always gets set to 12, but the victory threshold for it is 20, so it’s like having a “Launch Rocket” button that always fuels you to 60% and then stops forever.

So what do players really do? They increment the shared counter… and then start playing a tiny psychological game of chicken: “I could push the counter closer to 15… but then you might get the final move.” So the game hands them some useless actions that function as “stalling.” If both players are stubborn, they can just keep stalling until the match times out.

It’s not strategy so much as who blinks first.

Game 2: “The Game Where You Can’t Move”

This one is my favorite kind of failure because it’s so clean.

The game defines two counters. The only action in the entire game says: “Decrease both counters.”

But the action is only allowed if both counters are greater than zero.

And the game starts with both counters at zero.

So on turn one, a player looks at the rules and… there is literally nothing they’re allowed to do. No move exists. The game doesn’t even fail dramatically. It just sits there like a vending machine with the power unplugged.

In other words: it’s a tabletop game prototype that begins in a stalemate.

Game 3: “First Player Loses: The Speedrun”

This one actually runs, which makes it even funnier.

There are two counters. One counter is the only one that can possibly lead to a win — the other one only goes downward forever, so its “win condition” is a permanent lie.

The “real” counter starts at 0, and there’s an action that increases it by 2. The victory threshold is 4.

Here’s what happens:

• Player 1 goes first. Their only sensible move is: 0 → 2.
• Player 2 goes next and does the same move: 2 → 4.
• Player 2 instantly wins.

So the entire “game” is basically:

“Player 1 sets up the win for Player 2.”

It’s like a sport where the first player is forced to place the ball on the penalty spot and then politely step aside.

Game 4: “The Unwinnable Ritual”

Three players this time. There’s one shared counter. Winning requires the counter to reach 4.

But the rules only let you do two things:

• Set it to 2.
• Decrease it by 2 (if it’s above zero).

Notice what’s missing: any way to make it bigger than 2.

So the win condition is a castle in the sky. The game is a ritual where players take turns setting the number to 2, or knocking it back down toward 0. It can never, ever reach 4. No amount of cleverness changes that.

It’s essentially a machine that cycles between “2” and “less than 2” until you get bored or the turn limit ends it.

---

ChatGPT finished its report with this note: “The first evolutionary run didn’t produce brilliant board games. It produced life, in the same way early evolution produced algae and sea foam. These were rule-sets that technically existed… but didn’t yet deserve to.”

These idiotic games already provided fun by existing, so as far as I care, this app has already been a success. The good thing is that I have code to handle degeneracy, fitness, etc., so I simply have to tighten that code so some of these nonsensical points would either kill a genome or penalize its fitness.

By the way, earlier tonight I was playing tennis in VR with people across Europe while I waited for Claude Code to work through the tickets of one of the app’s new features. And yet the world is terrible. We live in the lamest cyberpunk dystopia imaginable.