I have always believed what the article more or less states. But you have to remember, the primary and maybe only source of duplication in software is situational dependency (the other word escapes me for this). If there was a universal tree of software functions that could be accessed over a network no function would ever be duplicated and every function would be reused from a central tree. When you put 2+2 inside a method or function body you just duplicated code. or any code inside a method or function body.
This is why we have to have programs that duplicate code by doing anything like adding two numbers together or complex logic that is easy to create bugs when someone wrote it 40 years ago better. Because code reuse is mostly done on a very small scale.
Given thats the case when you start on a new React project as an example you are not reusing application code you are duplicating the react framework so you can duplicate every other web app in every sense except maybe the visual.
There is no such thing as full reuse and until we get to a universal network invocable function tree that can be extended only when its truly unique we never will. Maybe AI will do this. People cannot.
At the end of the day code duplication needs to exist to optimize for local correctness (or incorrectness) and speed and abstractions goal is not to provide pure reuse. Its to provide a place to "put your logic" that may be similar and has access to typical state that some kind of widget might typically need.
It was definitely an interesting way to allocate pointers.
I did once have a very large project where devs didnt understand this and resolved hundreds or more off by one and memory overwrites in C due to this feature.
But at the same time, I think blaming the software was kind of a cop out. Devs were in a hurry and simply didnt respect the rules. Given todays software engineer at large. Nerfing programming languages so they cant destroy things might not be a bad idea. But AI will nerf everything.
why is AI gonna nerf everything? sure it could be used as the easy button, but I just spent two hours this morning learning about the neuroscience of how memory works in the brain that I didn't mean to and now I want to run studies on how memory works.
Why do you assume that AI is gonna nerf everything?
See, AI was trained on existing data - on all that existing C code out there (sure, and also on all the papers and articles saying what was wrong with that C code). Those bugs are in the training data, and often not marked as bugs. So when AI generates C code, is it going to avoid making the mistakes that human code made? No, it's going to generate the kind of code it was trained on. How could it be otherwise?
> See, AI was trained on existing data - on all that existing C code out there (sure, and also on all the papers and articles saying what was wrong with that C code). Those bugs are in the training data, and often not marked as bugs. So when AI generates C code, is it going to avoid making the mistakes that human code made? No, it's going to generate the kind of code it was trained on. How could it be otherwise?
The generalization of this is why I think all these AI companies writing blog posts where the marketing department is just jer—ranting endlessly about how AI will improve itself into the singularity is just crazy talk. They generate a random statistically likely output, and the most statistically likely output is mid. Exceptional outputs — the ones that wow us or move the needle are exactly that, unlikely. AGI is sci-fi, and LLMs will not change that.
You can see the same effect when AI emits bash, too, and especially so since most bash is terrible, and most users of bash do not put in the effort to learn bash and its foibles. So it outputs what most people write, which is not great.
In far from an AI fanatic, but I would argue training it on GitHub PRs and general software patches already provides that. Instead of just seeing the static snapshot it sees “this code was replaced by this (hopefully better) code”
Would you even know? Serious question. The volume of code the models can produce, the subtle ways these bugs can manifest (or even only manifest when under attack), it seems like they would be easy to overlook.
I have a habit of getting GPT 5.5 to review everything Opus writes for me, and vice versa. The model in the reviewer role frequently finds things I overlooked myself. Occasionally in parts of the code I wrote.
No modern LLM has found any buffer overflow bugs in parts of my code that originated from another LLM. Again, though, they have found one or two that were my fault.
I use Zig, which has slices, so so far none. But man, it can't get ref counting right to save its life. There have been remarkably few times it's gotten it right on the first try. My codebase considers OOM recoverable, so it keeps forgetting to clean up memory when OOM is raised. Even in the happy path though it still messes up ref counting. I use Kimi k2.6.