Programmer Humor

When I want to be lazy and make some simple excel macros is about the most iv trusted it with that it manages to do with out fucking up and taking more time then just doing it my self.

I don't think it has much to do with how "complex or not" it is, but rather how common it is.

It can completely fail on very simple things that are just a bit obscure, so it has too little training data.

And it can do very complex things if there's enough training data on those things.

I've also found it useful for simple Python scripts when I need to analyze data. I don't use pandas/scipy/numpy/matplotlib enough to remember the syntax and library functions. With vibe coding it, I can have a script in minutes for reading a csv with weird timestamps, scale some of the channels, filter out noise or detrending, perform a Fourier transform, and do a curve fit towards a model.

But then obviously I know every intermediate step I want to do.

I don't see how it would save time as someone whose job is to currently undo what "time" it "saves". You can give Claude Code the most fantastic and accurate prompt in the world but you're still going to have to explain to it how something actually works when it gets to the point, and it will, that it starts contradicting itself and over complicating things.

You said yourself he has to reiterate through the code with the LLM to get something that works. If he already knows it, he could just write it. Having to explain to something HOW to write what you ALREADY know can't possibly be saving time. it's Coding with extra steps.

I don't think it saves time. You spend more time trying to explain why it's wrong and how the llm should take the next approach, at which point it actually would've been faster to read documentation and do it yourself. At least then you'll understand what the code is even further.

I do the same, I am not sure if it saves time. Some times not. Other times if it is a task I really don’t want to work on this helps me to get started and break through procreation

Agree, my spouse and I do the same. You need to know how to code and understand the basic principles otherwise it's a bit like the Chinese room thing where you may or may not operate currently not have no actual clue of what you're doing. You need to be about to see when llms follow their hobby and blow three lines of code unnecessarily out of proportion by adding 60 lines of unneeded shit that opens the door to more bugs.

Oh man, I love ChatGPT for one thing in particular: "Hey chatbot, is there some library or standard library function for that very specific, yet still kinda generic thing I'm trying to do, so that I don't have to write it myself?"

It does frequently give a helpful answer. That is, it doesn't give me working code, but a helpful pointer to some manual where I can find good instructions for how to use the thing to solve my problem.

I don't want to dismiss your point overall, but I see that example so often and it irks me so much.

Unit tests are your specification. So, 1) ideally you should write the specification before you implement the functionality. But also, 2) this is the one part where you really should be putting in your critical thinking to work out what the code needs to be doing.

An AI chatbot or autocomplete can aid you in putting down some of the boilerplate to have the specification automatically checked against the implementation. Or you could try to formulate the specification in plaintext and have an AI translate it into code. But an AI without knowledge of the context nor critical thinking cannot write the specification for you.

When I see a sloppy PR I remind people “AI didn’t write that. You wrote it. Your name is on the git blame.”

You should(n't) watch Quin69. He's currently "vibe-coding" a game with Claude. Already spent $3000 in tokens, and the game was in such a shit state, that a viewer had to intervene and push an update that dragged it to a "playable" state.

The game is at a level of a "my first godot game", that someone who's learning could've made over a weekend.

Are you saying the comment is fake, or the sentiment? This was actually posted to reddit: https://archive.is/U9ntj

That comes up in that sub occasionally and people offer it as a service. It's 2 different universes in there - people who are like giving a child a Harry Potter toy wand that think they're magic, and then a stage magician with 20 years of experience doing up close slight-of-hand magic that takes work to learn, telling the kid "you're not doing what you think you're doing here" and then the kid starts to cry and their friends come over and try to berate the stage magician and shout that he's wrong because Hagrid said Harry's a wizard and if you have the plastic wand that goes "bbbring!" you're Harry Potter.

Yeah, this is my nightmare scenario. Code reviews are always the worst part of a programming gig, and they must get exponentially worse when the junior devs can crank out 100s of lines of code per commit with an LLM.

"What do you do? "Oh, I work in AI Disaster Response"

It's a real post on Reddit. I don't know what combination of screenshotting/uploading tools leads to this kind of mangling, but I've seen it in screenshots from Android, too. The artifacts seem to run down in straight vertical lines, so maybe slight scaling with a nearest-neighbor algorithm (in 2025?!?) plus a couple levels of JPEG compression? It looks really weird.

I'm curious. If anyone knows, please enlighten me!