How can this ever be sustainable without a huge monthly subscription?
post
From their blog on Steam:
For the paid release, we plan to account for expected AI usage in the gameβs price. Players will not be charged separately for their individual AI token usage after purchasing the game. This isnβt because each player is particularly expensive to support. The total cost has increased because so many people are playing, and we have taken out a bank loan to help keep the demo running.
I do not understand. They have to factor in token usage to the game price, but also the longer people play the more they will lose money over time? How is that even an operating model.
If they don't let players hook up their own API key, it's a guarantee they shut it down after a year.
My guess is that they are just morons. It is really easy to be confident in your projections if you are a stupid person.
It seems that it's one of those games with zero replayability
They probably bank on streamers make fun videos in 4-5 hours of gameplay total, then the people that buy the stuff and never play it
Btw it uses Gemini and gpt, it's expensive to run that.
What if someone pirates the game and starts wasting tokens?
Loans are normal and you take losses to capture the market then you enshittify
But loans donβt scale with the amount of players and played hours. They typically purely factor into the cost of development and support.
Server costs.
Typically scales fractional to the user count and often in steps. These costs stay relatively stable no matter if you have 1 or 10000 players (for example), then rise a bit to support 18000 players, etc. And they should not start out high enough to require a loan just for them.
If you build an MMO and get a sudden influx of millions on players⦠ok. Then maybe. But they typically also grow somewhat organically.
Each user causing you direct costs that even increase with their usage isβ¦ not optimal. But hey, itβs worth a shot, I guess.
Why are there so many opportunities to post this meme?

The only model I program with is a Model M.
Well maybe its a Mini M, but close enough XP
Why can't they just fine tune a small locally hosted model? Are they dumb?
Most gamers have a GPU anyways
Exactly my thought. Surely you could and achieve at least a similar result.
I would say yes, yes they are dumb. Taking out a loan for a game that isn't even out yet according to the article to keep an ai going is pretty stupid. Even if the game does well, the overhead is massive for what could be done with a smaller specialized model. Hell, I bet payig someone to train an existing model would be significantly less than a $1000 a day when you aren't even making money yet.
Because this is vibe-coded. You think an LLM is going to give up how to make your own?
I literally let deepseek + opencode modify the nanogpt code for my own custom model and suggest training datasets lol
Ran out of patience after a while because crappy GPU, but it was training and words were being worded when I prompted it after the first day of training
"Markets can remain irrational longer than you can remain solvent."
So they didnβt even do the math for a basic cost forecast before launching the demo. Also this shit makes them vulnerable for bad actors. Just put this game on several PCs and run it 24/7 to bankrupt them.
What makes you think that the demo just got massively more popular than expected? This is probably an occam's razor type of thing.
I'm more curious about how they were planning on structuring the full game to make a profit. Hopefully they're just planning on using a client side ai, but that would likely make the demo somewhat deceitful.
Minimum Specs : 96GB VRAM
AI running a video game, somehow still less power efficient than a playstation 1. Good job guys. Just make an actual video game.
me explaining to future generations why all the water dried up
How long before the internet is flooded with clips of the little sister saying all kinds of nasty stuff after people prompt inject the AI?
The AI disclosure on Steam claims the sister's voice is generated by a local model. If they're being charged for tokens I somehow doubt that.
Another engineer I work with had a really well put description of proper AI implementation that really applies to this:
"AI should help you build the automation, AI should not be the automation."
It's every bit as horrible as you're probably imagining. She has one of those talks-too-fast fake slop voices like on the YouTube spam shorts.
Hilarious article to put out tbh. If the game wasn't broken before, it's gonna be now.
AI powered games donβt work.
We did a super cool project for a game jam last year. Cost like 50c per game for us.
Cool for the 100 or so players who tried it, but thereβs no way to make a profit from it.
Local models maybe.
For just simple and guided actions and dialogue, local would be fine for this. Or better, have AI design the large database of what to say or do when a player does things. No need to reinvent what players are going to duplicate. The token cost was worse when you consider all the same stuff being prompted over and over.
If they're actually hitting a cloud LLM, there's no good reason not to wrap that and memoize the results... it would probably be hard to generalize the inputs (it could just be the whole game state for all we know π€·ββοΈ) but youd build up that database of all the most common input/outputs in real-time
Exactly, that's just smart programming (maybe that's one of the issues... but I can't say that without evidence).
I saw someone years ago develop that concept for a text to speech idea for NPC responses, and for the same reasons. Why repay for something when you can just capture the sound file for each new phrase? The cost is in the creation and addition of things, not in the actual running of every person's game (or app in this case).
In case it might be interesting info, maybe want to check out WebGPU, WebLLM, Wllama, or Transformers.js. You can write web apps that run an llm on the client's browser.
It requires permission and compatibility and isn't... it might be tough to really make accessible to the average user.
But it's an option and kinda cool.
Given that people playing this have GPUs, why not run the AI model locally? I can't imagine it needs to be a very large model.
Actually sounds like the perfect use case for a model like Jev. I think Jevos should be able to handle that locally.
Because even very small models will take up a large chunk of your cards VRAM, if not all of it. And then you still have to render the game on top of it.
There are 1-2gb vram models that should fit in anything. Game doesn't appear to be texture heavy, and targetting a 6gb gpu, would cover anything less than 10-12 years old.
It looks like a very simple game graphically, so having say 2GB VRAM left over for a small model would probably be viable?
I canβt say for certain but based on how the game is described, I donβt think itβs likely a 2GB model would be viable at all.
That sounds like the most boring game imaginable and I hope they go bankrupt from it
Guess you've never heard of Desert Bus.
top 73 comments