capyba

Nothing about this is moving goalposts - you and/or the person(s) conducting this study are the ones being misleading!

If you want to measure time to complete a complex task, then measure that. LOC is an intermediate measure. How much more productive is "55% more lines of code"?

I can write a bunch of garbage code really fast with a lot of bugs that doesn't work, or I can write a better program that works properly, slower. Under your framework, the former must be classified as 'better' - but why?

I read the study you reference and there is literally nothing in the study that talks about whether or not tasks were accomplished successfully.

It says: * Junior devs benefited more than senior devs, then presents a disingenuous argument as to why that's the senior devs' fault (more experienced employees are worse than less experienced employees, who knew?!) * 11% of the 55% increase in LOC was attributed directly to LLM output * Makes absolutely no attempt to measure whether or not the extra code was beneficial

“ Productivity (measured by the number of lines of code produced) increased”

The LLM’s better have written more code, they’re a text generation machine!

In what world does this study prove that the LLM actually accomplished anything useful?

I’m glad to hear someone say that. I’ve been wrestling for weeks with the idea of reinventing a particular wheel in my profession, for a personal coding project. The problem is that my implementation can’t ever be as complete or as useful as the existing solutions because it’s way too much for one person to accomplish in a reasonable amount of time.

But, I like it, I’ve reinvented many wheels in my work and it’s benefited me greatly. So I will reinvent this particular wheel as well…

I disagree - a lot of people with positions of power are unfortunately saying the “will be left behind “ bit

But it wasn’t industrial automation that ruined Detroit. It was the automakers’ failure to compete with highly capable foreign competition.

Can you explain this to me? Trying to understand but can’t haha.

I’m out of the loop with laptop processors and GPU performance measured in “AI TOPS”.

I’m not “doing AI” locally on my laptop, are those AMD processors of any use to me?

I’m familiar with traditional X86_64 CPU architectures, I just don’t understand what (if any) extra bells and whistles the “AI” chips offer.

I was thinking the same but got a prusa kit instead. Have you use or built a prusa before - how do they compare? At some point I’ll want two printers and was thinking a voron or similar would be perfect because I could print parts for it now.

Reading about the `left-pad` incident (almost 10 years ago itself!) gave me an incredibly strong aversion to dependencies in my own projects. I suppose that like the author, I care a lot about understanding the foundations and ensuring that whatever I build can stand on its own, as much as that may be. I think this philosophy has helped me learn a lot more than if I had just focused on “build as fast as possible”, but definitely limited the “amount” of projects I’ve completed.

I love writing in C because a) it’s simple and I can understand everything that’s going on and b) I feel confident that it’ll compile just fine years from now, because that’s how that particular ecosystem has worked and likely will continue to work.

Recently, I’ve been writing new projects in Rust, which has far better ergonomics but is far more complicated and causes me to cede control to the whims of the community and the compiler. I’m sure there are simple ways to ensure that I can use the same compiler for all new platforms that will support it, but just like everything in the modern day (for both better and worse), is implicitly and explicitly designed to always be updated.

Rust has become much more stable in the past few years, but you’re right to expect breaking changes in the next ten years. I love using the language, it’s been fun to learn and I (hope) it’s helped me write far more robust code, but I’m sure in the future there will be many times where I’ll miss C.

In terms of performance, how does it compare to compiled languages like the C it’s written in?

I love working in Julia, it makes clean numerical code so easy to write.

Excellent point, that makes a lot of sense. You’re absolutely right that I’ve only tried problems that could be auto vectorized…

Very cool, thank you for sharing!

Personally I’ve never been able to beat gcc or icx autovectorization by using intrinsics; often I’m slower by a factor of 1.5-2x.

Do you have any wisdom you can share about techniques or references you can point to?

as other commenters noted, this sounds like chatgpt and it gave me indigestion when reading

I’m a mechanical engineer, not a software person, but I write a lot of (hopefully close to professional quality) code for my work. writing tests while in the beginning/middle of my development cycle has been the best change I’ve made in how I do things in quite a long time. Since I’m a self-taught amateur often working solo, its invaluable for helping me gain confidence that everything I’m doing is working correctly as I learn the language/libraries necessary for me to build each new program.

I’m not saying that you yourself have this attitude - but the “tests are for suckers, I just ship” crowd really grinds my gears because to me it says “ha! Why do you care about getting things right?”

Totally get where you’re coming from though, sometimes the expected behavior is trivial to verify or (in the case of GUIs) can be very difficult and not worth the effort to implement.

Hard agree.

Alibre CAD has affordable permanent/monthly licenses, I recommend that as the affordable commercial option.

I agree with the general gist of this piece, but the awkward flow of the writing style makes me wonder if it itself was written by AI…

There are open source or affordable, paid alternatives for everything the author mentioned. However, there are many places where you must use these things due to social pressure, lock-in with a service provider (health insurance co, perhaps), and yes unfortunately I see some of these things as soon or now unavoidable.

Another commenter mentioned that ChatGPT is one of the most popular websites on the internet and therefore users clearly do want this. I can easily think of two points that refute that: 1. The internet has shown us time and time again that popularity doesn’t indicate willingness to pay (which paid social networks had strong popularity…?) 2. There are many extremely popular websites that users wouldn’t want to be woven throughout the rest of their personal and professional digital lives

Makes sense - thanks for clarifying!

Isn’t the `real` datatype in Fortran a float64? Or am I making that up?

Chickens aren’t particularly noisy, especially relative to a “non rural area”, and any mess is contained within the coop.

Can’t a good privacy browser randomize some of the ‘nonessential’ elements it sends, thereby creating a new fingerprint for each website? Or change them all to the same one so everyone using that browser has an even more similar fingerprint?

Its always struck me as odd that your browser just volunteers so much information every time you go to a site. Then again, my knowledge of how the web works is limited to HTML+CSS from the 00’s…

Yeah - that's the thing. I've seen enough hallucinations or other errors to need to carefully check the output from every prompt. If I need to check it thoroughly, then I suppose I could treat it as a fancy autocorrect? But then the productivity gain is much more marginal.

If the accuracy didn't matter, then what's the point of doing the work in the first place?

I honestly feel that LLMs have been a net negative for me rather than a positive, *especially* with software development. Most of the time I end up coaxing it through a series of prompts, only for there to be some bug in the code it generated for me that takes me longer to debug than if I had just read the relevant docs and examples myself.

I’m sure there’s utility now, and I’m sure it’ll (ever more slowly, asymptotically) improve, but at the moment it (and the Google that performs worse than that of 5 years ago) has been costing more time than the alternative web search, study, then write it myself. And don’t get me started on non coding technical subjects.

Or just write it in java

I don’t know anyone that works at Meta, so I’m hoping that someone here could answer this for me-

What makes employees there feel good (or at least okay) about doing stuff like this? You're spying on people, no? Surveilling ordinary people, not enemy combatants or foreign militaries? Perhaps a friend of a friend or even a family member? This kind of thing is so creepy and disturbing to me, not that it’s anything new…

Why telegram instead of signal?

Dang, feeling this right now. Went above and beyond again and again, for years, barely got a thank you - let alone a raise of any sort. Now I’m completely burnt out and don’t have the energy to even consider my own ideas for my own next move.

I’ve been trying to build a desktop app in Rust+egui but as a Rust newbie (and desktop app newbie in general) I’ve found it really difficult to learn so many concepts at once. My work centers around mechanical engineering analysis tools which require high perf backend and data visualization on the front end. With Tauri did you find it difficult to maintain multiple stacks (rust, js, html, etc)?