Usage cap kills custom GPT game experience

I’m working on building custom GPTs to run word games and narrative games, and the usage cap complestly kills the play experience. A game requires multiple prompts and responses, and can quickly time-out with current limits.

For example: I’m working on a word game where the GPT comes up with a secret pattern that words either match or do not match. The player guesses words and the GPT tells the player if their guess matches. The player wins when they can correctly guess the rule. This kind of gameplay obviously requires lots of exchanges with the GPT. Imagine narrowing things down until you are sure you have a great guess only to be told that you have to wait four hours before you can complete your game. So frustrating!

I also made a GPT to pay 20 questions lol. I’m going to have to narrow the scope and call it “ten questions.”

Are GPTs always going to be terrible at this kind of use case (i.e., one that requires multiple responses to complete an experience)?

IDK if anyone else out there is feeling this pain, but if so I will say that given current constraints, I’ve had more luck with designing games that have a fixed number of required inputs from the player. For example, I’ve got a GPT that does a madlibs style game and generates an image based on the result. That one works by limiting player inputs to seven (plus one more if they opt to generate the image).

So maybe that’s just a desing constraint that we’ll have to live with for now?

1 Like

to be honest I wouldn’t be surprised if they disappear again in a couple of weeks/months.

IDK if anyone else out there is feeling this pain, but if so I will say that given current constraints

I canceled my chatgpt subscription today because the limit got on my nerves. I liked that they had a nice ui and everything, but getting ripped out of your flow every hour to be told you need to wait two hours is not worth 20 bucks.

So maybe that’s just a desing constraint that we’ll have to live with for now?

I hope you don’t. the company needs to feel your dissatisfaction, otherwise this trajectory is not gonna change.

1 Like

:raised_hand: 1st post. I feel your pain, one thing I find useful is using the free version of gpt to craft a prompt that will minimize the prompting.

1 Like

Yeah it’s a pain but I think it’s temporary.

1 Like

It wasn’t like this before. Now everything counts to the point where some days I don’t even want to use it.

1 Like

I too have had that issue with quick play games. I wrote a 20 Questions game but abandoned it because it just uses up the cap too quickly. AND, default ChatGPT plays 20 Questions just fine on its own. Similarly with a Word Association Game I wrote - fun for the first 20 interactions, then I have to wait 3 hours - not so much fun. Abandoned that too.

1 Like


i loss time and message limit. for check file text.

I have the same problem. I created a challenge where players have to obtain a secret from GPTs. It consists of six increasingly difficult stages (aka six GPTs) and while the first stage might be solvable with a single prompt, the later stages often require users to write a lot of prompts and they hit the cap before solving it. For this kind of game this might be fine since you can start one stage after another whenever you like. But I actually developed the backed for a different kind of game and was planning to set a $ reward for the first person to finish a series of riddles. But as it stands, it would be the persons who manages to come back exactly every 4 hours who gets the reward and not necessarily the best riddle solver.

Same problem here (to a degree), for a narrative-based game.

At the moment, I try to get a single playthrough to finish below 40 exchanges via a sort-of timer mechanism which is part of the narrative (assuming the Custom GPT limit is 40, not 100% sure about that, I might have to adjust it).

Anyhow, my assumption is that limits will eventually increase. Hardware costs will go down, they’ll distill a cheaper, faster and slightly dumber model and call it GPT-4.5-Turbo (hopefully it won’t be much worse), etc., so building a more complex game now is not necessarily wasted time, as long as you are building something which does not hinge too much on the exact current version of ChatGPT.