AI Lab

I fw this heavy, extremely short even for a demo though

I found it a little annoying that I couldn't buy more GPUs after the 1st server one.
And it's a shame that the graph generation frequently freezes, I like seeing it be finished.
Speaking of the model creation, it feels counterintuitive for the "what direction will this be focused in" slider to only be available while already training a model. You'd think that decision would be made beforehand, right?
And when a model turns out to be low quality the discard option kind of feels like a trap. There is no real benefit to discarding (in this stage of the game) other than a negligible amount of computing power to be saved. It doesn't provide a discount on future training (that I've noticed) or anything else.

usually i dont support stuff that uses ai but i like this one because firstly the ai is used only for coding assistance which in my opinion is nowhere near as bad as anything else and the games actually neat

Agreed that this is a prototype. I spent more time searching for what to do next and finally discovering the "end demo" button than playing the game. Needs a bit more fleshing out to call it a demo

even though it was pretty confusing when i started playing, for something that was made with ai, i actually liked it! great game and continue developing it please!

ah yes, using ai to make a game about an ai collapse, how ironic. i'll just continue playing to see whether this is good :3

Compute rental is too op. It gives way more revenue than running ai service. So realistic lmao

I think you are confusing demo with playable prototype. There's not even 10 mins of content there.

I found it a little annoying that I couldn't buy more GPUs after the 1st server one.
And it's a shame that the graph generation frequently freezes, I like seeing it be finished.
Speaking of the model creation, it feels counterintuitive for the "what direction will this be focused in" slider to only be available while already training a model. You'd think that decision would be made beforehand, right?
And when a model turns out to be low quality the discard option kind of feels like a trap. There is no real benefit to discarding (in this stage of the game) other than a negligible amount of computing power to be saved. It doesn't provide a discount on future training (that I've noticed) or anything else.

Weird design choice that training ai models consumes data. If I use some data in training a model, what prevents me from reusing the same data in training future models?