discussion AI is horrible... just a rant.

[LewisCarroll]

I want to throw every AI tool I have, save one, out the window. There's a point where it's just not very helpful and that point is at the beginner level. For factoids, meta-analysis and general data gathering it's king. For actual work, it's pretty useless.

I'm learning Ren'Py. I'm past the basics, so AI has become almost useless. (If anyone has links to professional classes with tools link them below).

My issue is I have a persistent variable that I'm tracking over several menus, but every time i scroll back and select a new option it just adds to the total over and over again. I'm sure the solution is very simple, but ChapGPT, CoPilot, GIT, DeepSeek, have all fallen short. None of their solutions have worked and some of them have been extremely complex. It's like working with Stable Diffusion where you have to be an expert in prompt writing, have deep understanding of the models to even come close and even then something comes out looking like a horror show.

I tried using a bit for work to see what it would spit out, it wasn't even in the ballpark of accuracy. At this point, if AI is taking your job, your job isn't very challenging creatively.

Anyway... if anyone knows how to control a persistent variable in Ren'Py so it isn't modified by menu changes, that would be great My only other option seems to be to backwards engineer popular games and see how they manage their programming.

 

[goons]

I think your luck might be with trying game dev forms or the like. ai tools, at least the general ones we use seem to struggle after reaching more advanced knowledge that has less free available resources for it to use

 

[rocketsheep]

I want to throw every AI tool I have, save one, out the window. There's a point where it's just not very helpful and that point is at the beginner level. For factoids, meta-analysis and general data gathering it's king. For actual work, it's pretty useless.

I'm learning Ren'Py. I'm past the basics, so AI has become almost useless. (If anyone has links to professional classes with tools link them below).

My issue is I have a persistent variable that I'm tracking over several menus, but every time i scroll back and select a new option it just adds to the total over and over again. I'm sure the solution is very simple, but ChapGPT, CoPilot, GIT, DeepSeek, have all fallen short. None of their solutions have worked and some of them have been extremely complex. It's like working with Stable Diffusion where you have to be an expert in prompt writing, have deep understanding of the models to even come close and even then something comes out looking like a horror show.

I tried using a bit for work to see what it would spit out, it wasn't even in the ballpark of accuracy. At this point, if AI is taking your job, your job isn't very challenging creatively.

Anyway... if anyone knows how to control a persistent variable in Ren'Py so it isn't modified by menu changes, that would be great My only other option seems to be to backwards engineer popular games and see how they manage their programming.

Maybe just take the obvious pedestrian solution. Instead of a single persistent counter variable, use many persistent boolean variables and compute the counter from them when (if) you need it. I am assuming you have something like this:

Situation 1a: counter += 1
Situation 1b: (no change)
Situation 2: counter += 1
Situation 3: counter += 1
Situation 4: if counter > 2 then ...

Replace it by something like this:

Situation 1a: counter_1 = true
Situation 1b: counter_1 = false (or no change, depending on what you are trying to achieve)
Situation 2: counter_2 = true
Situation 3: counter_3 = true
Situation 4: counter = number of true values in (counter_1, counter_2, counter_3); if counter > 2 then ...

If you need a lot of such boolean values, maybe use an array.

 

[Sarya]

It's like working with Stable Diffusion where you have to be an expert in prompt writing, have deep understanding of the models to even come close and even then something comes out looking like a horror show.

It was like that years ago, it improves widely, even with a really simple prompt (and some model are trained to natural language too), you can do some good stuff, on perchance or on SD: your vision of AI is a caricature of the reality.

 

[zyxion0]

i don't really like AI, it gets almost the same result

 

[LewisCarroll]

Maybe just take the obvious pedestrian solution. Instead of a single persistent counter variable, use many persistent boolean variables and compute the counter from them when (if) you need it. I am assuming you have something like this:

Situation 1a: counter += 1
Situation 1b: (no change)
Situation 2: counter += 1
Situation 3: counter += 1
Situation 4: if counter > 2 then ...

Replace it by something like this:

Situation 1a: counter_1 = true
Situation 1b: counter_1 = false (or no change, depending on what you are trying to achieve)
Situation 2: counter_2 = true
Situation 3: counter_3 = true
Situation 4: counter = number of true values in (counter_1, counter_2, counter_3); if counter > 2 then ...

If you need a lot of such boolean values, maybe use an array.

Yeah, that was one of the five different methods. AI didn't even think of yours. AI lacks creativity. Sadly, I did think of it eventually, but I wanted to re-use menus and not have to calculate at every choice (menu). Wanted a universal variable where I could control from a single point. I've seen other games do it. Many games, in fact. With multiple variables. LC's dev area is pretty lacking. Will probably have to look into other forms. Ren'Py as a VN is super easy. As a real gaming platform it's a bastardized mashup making it too messy to code in.

 

[rocketsheep]

Yeah, that was one of the five different methods. AI didn't even think of yours. AI lacks creativity. Sadly, I did think of it eventually, but I wanted to re-use menus and not have to calculate at every choice (menu). Wanted a universal variable where I could control from a single point. I've seen other games do it. Many games, in fact. With multiple variables. LC's dev area is pretty lacking. Will probably have to look into other forms. Ren'Py as a VN is super easy. As a real gaming platform it's a bastardized mashup making it too messy to code in.

Why don't you just look how the other games do it? Some don't even try to hide their code, and AFAIK for the others there are tools. Unrpa for unpacking the rpa archives, and some other tool for decompiling rpyc.

 

[Redyscom]

AI is a tool and should be treated as such, if it's not useful for the job it's left aside and something more useful is used. It's also not good when people let AI do all the work, that sucks.

Anyway, for your problem, I think it's a good idea to reverse engineer games to learn how they work, although it's a bit of a clumsy and slow way depending on your skills.

The other alternatives would be to use YouTube guides from someone you know who has the skills you want to learn, maybe the author of a game who shares their knowledge.

Another way is to experiment a bit, small projects to polish the basics while learning new things, it's much better if you combine them.

At least those are the things I did while making a Pokemon fangame.

 

[LewisCarroll]

Why don't you just look how the other games do it? Some don't even try to hide their code, and AFAIK for the others there are tools. Unrpa for unpacking the rpa archives, and some other tool for decompiling rpyc.

That's what I said I'll have to do in my last sentence vs just finding a quick solution with AI instead of going line by line in code I'm not familiar with yet.

 

[LewisCarroll]

AI is a tool and should be treated as such, if it's not useful for the job it's left aside and something more useful is used. It's also not good when people let AI do all the work, that sucks.

Anyway, for your problem, I think it's a good idea to reverse engineer games to learn how they work, although it's a bit of a clumsy and slow way depending on your skills.

The other alternatives would be to use YouTube guides from someone you know who has the skills you want to learn, maybe the author of a game who shares their knowledge.

Another way is to experiment a bit, small projects to polish the basics while learning new things, it's much better if you combine them.

At least those are the things I did while making a Pokemon fangame.

Well, I did say this was a rant and I get what you're saying. Learning coding by yourself is like using a dictionary to learn Shakespeare. Yeah, you'll understand what he's saying but you won't understand the meaning behind it. I'd just like an option where I don't have to learn Iambic pentameter to understand that suicide is bad.

 

[panda]

AI for the purposes of research, incredible. AI for the purposes of entertainment, no thank you.

 

[xatfx]

AI, like any tool, requires training to use it correctly.

 

[Trialbaloon]

Its possible in the future AI will be great, but right now it feels like it's not really improving on existing tools, just cannibalizing creative people's work and then itself. . .

 

[ChaosLoli]

Yeah...the tools aren't great right now. Need some more time in the oven

 

[Gaminggxz]

I dont really prefer Ai but if it does actually have some effort like little animation or actually drawn images I'd give it a shot