You can ask GPT to give you a plan for world peace and it will.
Will it work. Who knows. Will it be any good. *** no.
Knock it off with rubbing fake AI on everything.
Addon: JSE
|
Addon: JSE
wasn't talking to you brother. you have a copy.
You are right, AI isnt going to spit out something perfect, but at the end of the day its a tool to make writing it easier. have it lay the foundation for the addon, have it verify that the syntax you are writing works properly. people *** miss the point of these things, its not about having it spit the answer out to you its about having it give you something to work from.
[+]
from my experience today it definitely has potential, but you need to be very specific about what you want and be willing to make it try several times and correct it when it goes off course. it can make helpful suggestions for things you might not have thought about, which is entertaining.
Ive used chatgpt to make an addon to open up rema stone boxes then stagger trades to a holder. I eventually got it to work, but its really not that good at it, and there was a lot of “theres an error on line X, are you sure this is a windower function”, when it was not a windower function. The code is bloated as ***now and could probably be cut down by half.
I could probably also make it significantly better by not leveraging two separate addons (pounches and tradeplayer) to run actions ¯\_(ツ)_/¯ Its ok at building a skeleton and checks, but you have to have an idea of what youre doing to make it work.
[+]
It can't go past the training data and it can't comprehend complex interactions. For something incredibly simple like this addon, it should work fine. You might need to curate the input datasets some, but the logic is well within what it can handle.
Like Nynja said, it also bloats like crazy. I'm thinking of a GPT addon someone posted in the Ashita channel that was over 1000 lines when it should've been 150. That sort of code is not only much more tedious to read/fix for people who know what they're doing, it likely incorporates performance issues as well. [The addon broke the imgui state for other addons too, so definitely wasn't without visible flaws.] If you wanted to make something larger in scope like one of the crossbars, gearswap, etc.. you're dreaming. Learn to code and maybe GPT can save you a little time on it (more likely, it pigeonholes you into a design structure you aren't familiar with and doesn't actually save time, though). Asura.Darthsploder said: » have it verify that the syntax you are writing works properly.
[+]
It's like the heavy frameworks from the early 2010s that promised everything and came with everything included in them, DBs, Rest APIs, interfaces, UIs, you name it.
As soon as you want something a tiny bit more than just a bog standard basic UI that maybe kinda works out the box, it's a ***show to fix. You end up spending more time "fixing" broken ***than you would have spent writing the whole thing from scratch. You do you though.
[+]
my major hurdle with coding is knowing the right syntax for everything. the time saved throwing it at gpt and then getting help with the logic from actual people, for me, is very high.
probably a lot of personal learning bias, my first real programming class used emacs on some version of linux when I had never seen anything but windows/ms dos before and the shock just about killed me :p I mean I can't stand Lua. I think it's quite possibly the worst syntax used in a programming language since.. idk, Pascal maybe.
Rather write 100% TCL than write another line of lua. Lua's fine once you get used to it. Tables are ultra flexible and reasonably performant; going back to any strictly typed language always feels tedious.
Sure, Python is more or less a better Lua in terms of writing code. But, it's not that big a difference, and Lua(especially LuaJIT) is far more performant as an embedded language. My only serious complaint is that it's 1-indexed, but c'est la vie.
[+]
It's just so convoluted, can't stand it.
If you're going to have function if end end everywhere might as well write C and use { and }; I've heard the LuaJit performance argument before. It's nowhere near a good Python library with C bindings. Which gives you a syntax that doesn't suck, with C-like speed. Anyway, here's a good Python tutorial that assumes no prior knowledge, with an in-browser interpreter. I'd take Lua over objective-c. Though I also hate Lua with a passion. C# would have been nice even if it comes with a lot of overhead
Not a big fan of lua, but I probably prefer it to tcl, although I've done much more in tcl, although tcl is slow...like really slow.
And holy thread derail, but I guess it beats treadmills.
[+]
I mean I wasn't totally serious about TCL.
Maybe. I think. I do prefer it to Lua. And TCL is not slow, it's just not suited to interactive stuff. Dodik said: » If you're going to have function if end end everywhere might as well write C and use { and }; Due to a number of factors, especially that EVERYTHING IS A TABLE, lua lends itself pretty well to ifless flows. Most recent stuff I write for myself has very little IFing going on! |
All FFXIV and FFXI content and images © 2002-2025 SQUARE ENIX CO., LTD.
FINAL FANTASY is a registered trademark of Square Enix Co., Ltd.