this post was submitted on 07 Oct 2026
99 points (99.0% liked)
Programming
28777 readers
254 users here now
Welcome to the main community in programming.dev! Feel free to post anything relating to programming here!
Cross posting is strongly encouraged in the instance. If you feel your post or another person's post makes sense in another community cross post into it.
Hope you enjoy the instance!
Rules
Rules
- Follow the programming.dev instance rules
- Keep content related to programming in some way
- If you're posting long videos try to add in some form of tldr for those who don't want to watch videos
Wormhole
Follow the wormhole through a path of communities !webdev@programming.dev
founded 3 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
If you ask an LLM for Pong, it will give you Pong, or rather a Pong, a simulacrum average of all the Pongs in its training corpus. But it won't be your Pong. You will not have decided the color, shape, and size of the paddles, or the background color, or speed of the ball, or the language it's written in, or the way to architect the code, or any of the other aspects, so many that they can't be enumerated. It's not possible to describe exactly what you want to the LLM because that would be the code itself. Anything short of that and you will have omitted something, forcing the LLM to make a choice all on its own. Write your own damned Pong.
That's the thing people don't want their thing as long as it works
the more detail you give, the more detail you’ll get.
The only difference is that unspecified things will be filled in by random selection of ideas around the average.
Devs now have ability to go high level when safe, and go detailed when needed. So much about pong is easy and obvious so the llm can just do that for you and speed things up.
This is a tool for going faster, especially over the trivial stuff. The code didn’t go away, you can still edit it.
Yes, yes it will. I can change it to do whatever. My tool wrote the code based on my guidance.
If some stuff is trivial, then it is fast to do. If it's not trivial, time needs to be spent on it.
The amount of delusions that vibecoders need to ingest constantly to keep their beliefs is crazy.
Also, why the fuck do you need your pong project to be done so fast? If you're doing it as part of your job, just do your fucking job, and if it's as a hobby, then do it instead of playing with a garbage generator to try to make it do it for you.
Also, no, it won't be your pong because it will be built on stolen code. If you copy paste the code of someone else and say you made it because you changed the name, you didn't make a pong, you're just a piece of shit that lies about it.
Unfortunately this is not true. That is why the word "boilerplate" exists.
That's what deterministic code generators are there for. Or macros/metaprogramming.
Which is... not trivial?
The code generator in my LSP at least is just a few keystrokes away. That's pretty trivial I'd say. The other things maybe less so, yes.
Are you seriously saying the code is stolen because it's copied? Are you stating you do not copy code from stack overflow?
Please bare in mind that AI from a month ago is shit compared to AI today so opinions on AI coding need to be constantly updated.
And lastly, who the fuck cares if Pong was AI coded for work? If it works, it works. Do you want a medal for coding it your self? Do you predict an edge case that would break AI's version of pong? There are absolutely somethings where perfect code and good enough don't have any real world impact.
Code under a license that you appropriate to yourself is stolen, yes, don't play stupid.
The typical slopbro defense of "no but it was shit ten minutes ago but now it's amazing and perfect" is bullshit, everyone knows it, and it's just a hypocritical loophole used to block every critic. Most of the critics on LLMs from months and years ago are still very valid, and while I try to be charitable I cannot find a scenario where this argument would be used in good faith.
As I said, if your pong is for work, you should do your fucking work without contaminating it with slop, and if it's not for work there's no speed requirement. "If it works, it works" is such an absurdly moronic argument, I don't even know where to start. Experimenting on humans is horrible, but brings advances to medicine, and "if it works, it works". Planting evidence on someone you believe is guilty is illegal, but they end up in jail so " if it work, it works".
I'm done, there's no way I can delude myself any longer into thinking that slop users do it innocently, when you have such obviously shitty and worthless arguments, there's no way you could be dumb enough to believe them. This is obviously a bunch of bad faith arguments and nothing more.
Tell me specifically why you think a game of pong written for work with AI and vetted by a human is inferior to a game of pong written by a person if they perform indistinguishably for the person playing pong.
Because it is made through stolen data, while destroying the environment and basically the world.
"What is the difference between a synthetic diamond and a blood diamond, if they are chemically and structurally identical in the end?"
As I said, these arguments are so unbelievably stupid that they cannot be said in good faith, and each reply has been proving it.
So, your issue isn't the AI being probably the most dangerous thing since the invention of the nuke or the quality of the response. It's the fact they don't have closed loop cooling, a better sources of energy, and they used data on the Internet?
When shit goes wrong you use ai again to band-aid and repeat and repeat until you're at the point when you cross your fingers.
People use ai because they don't know how to get there which means they usually don't really know what's going on under the hood even if they think they understand what's going on over the surface which means you just have slop. And then BOOM!
Or you just don't fire and forget. Anyone who doesn't review what AI is doing as if it was a jr developer has more problems than their AI usage.
I use AI because it's incredibly fast for things that would take hours if not days or would even require making programs to program for my program.
For example, restructuring where files live in a program. That would be a pain in the ass to do but trivial to review. It's exceptional at finding bugs I overlooked (though it loves overbuilding). Hell I just had a literally last minute demo of a feature I made. While the feature was tested and fully functional, I didn't have anything connected to it to show for an unplanned demo. In a few minutes it built me a rudimentary example for the feature so I could show people how the feature worked. No human on this planet could have done that in the time I had.
What is the actual percentage of people make their own changes after generating a project with an LLM? Most people give the wheel to the LLM and prompt their way out of the resulting mess.