I'm not a coder, but my job requires a bunch of menial, boring coding. I do numerical simulations. After mathematically understanding the numerical method, it's basically half a step above data entry. There's also a bunch of legacy fortran code I have to build on that has zero documentation and three letter variables. This would be one of the few actually good applications of text generating machine learning imo.

[–] Tartas1995@discuss.tchncs.de 3 points 9 hours ago (1 children)

Or your employer would invest some money in a proper tool for your job.

[–] Droggelbecher@lemmy.world 0 points 9 hours ago (1 children)

Nobody has built a tool that executes a mathematical method that I have developed or at least adapted, at least not before I publish the method.

[–] marsza@lemmy.cafe 2 points 5 hours ago (1 children)

So why haven’t you published the method?

[–] Droggelbecher@lemmy.world 2 points 5 hours ago (1 children)

Because I'm in academia and it's a slow process to get things published in a way that 'counts' to the university and scientific community. I often need to implement stuff first to check a few things, whether it's viable etc.

[–] marsza@lemmy.cafe 1 points 5 hours ago (1 children)

That’s not how it works. Put it on GitHub like the rest of us and stop making excuses.

[–] Droggelbecher@lemmy.world 3 points 5 hours ago (1 children)

I'm sorry to tell you this but people do not, in fact, publish mathematical proofs on GitHub routinely. You publish them on arxive once the paper is done. And then in a journal. The solvers themselves aren't even what it's about at all, they're just to do numerical experiments with to have some examples. They aren't immediately useful for any applications outside niche research.

[–] marsza@lemmy.cafe 1 points 4 hours ago (1 children)

You’re right that mathematical proofs are usually published on arXiv and then in journals. But since you mentioned code: sharing code on GitHub is actually very normal in research. Even if it’s just a solver or scripts for experiments, putting it on GitHub helps with reproducibility, gives others a chance to learn from or build on your work, and makes it easy to cite. There’s no obligation to polish it perfectly—lots of research repos are just “as is” snapshots to support a paper.

[–] Droggelbecher@lemmy.world 1 points 4 hours ago (1 children)

To be very honest, I'm also a tad embarrassed to share my code. I guess I'll ask my professor about this.

[–] marsza@lemmy.cafe 1 points 4 hours ago

Now, that is completely understandable. This is also a reason I don’t publish most of my things. They work, they work well, but… Some of it is kind of nasty. However, other developers are going to understand. Just mention this in the read me file. Or better yet, use this as an opportunity to refactor code. An LLM could be very helpful For that process.

If you are not familiar with the process of using git or GitHub, i’m sure many of us, including myself, would be more than happy to help you.

[–] calcopiritus@lemmy.world 2 points 10 hours ago (1 children)

If it has three letter variables, chances are it was also written by someone that doesn't want to code either

[–] Droggelbecher@lemmy.world 2 points 9 hours ago (1 children)

It's so old that it was for purposes of saving memory.

[–] Patches@ttrpg.network 1 points 3 hours ago

Back when optimization was Black Magic. Now we just tell the customers they need better hardware.

Someone probably had to argue hard to get 3 letter variables. Guarantee there was some one arguing for 1 letter variables.