Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Can LLMs write better code if you keep asking them to “write better code”?

minimaxir.com

Post reply on HN
Print view View original

111–120 of 461 posts

Re: Can LLMs write better code if you keep asking them to “write better code”?

#111

Post by pplonski86 »

dash2 wrote: post #5

Something major missing from the LLM toolkit at the moment is that it can't actually run (and e.g. test or benchmark) its own code. Without that, the LLM is flying blind. I guess there are big security risks involved in making this happen. I wonder if anyone has figured out what kind of sandbox could safely be handed to a LLM.

I think that there should be a guard to check the code before running it. It can be human or another LLM checking code based on its safety. I'm working on an AI assistant for data science tasks. It works in a Jupyter-like environment, and humans execute the final code by running a cell.

Re: Can LLMs write better code if you keep asking them to “write better code”?

#112

Post by emseetech »

babyent wrote: post #81
juujian wrote: post #55

I often run into LLMs writing "beginner code" that uses the most fundamental findings in really impractical ways. Trained on too many tutorials I assume. Usually, specifying the packages to use and asking for something less convoluted works really well. Problem is, how would you know if you have never learned to code without an LLM?

I used to really like Claude for code tasks but lately it has been a frustrating experience. I use it for writing UI components because I just don’t enjoy FE even though I have a lot of experience on it from back in the day. I tell it up front that I am using react-ts and mui. 80% of the time it will use tailwind classes which makes zero sense. It won’t use the sx prop and mui system. It is also outdated it seems. It…

I've stopped using LLMs to write code entirely. Instead, I use Claude and Qwen as "brilliant idiots" for rubber ducking. I never copy and paste code it gives me, I use it to brainstorm and get me unstuck.

I'm more comfortable using it this way.

Re: Can LLMs write better code if you keep asking them to “write better code”?

#113

Post by SparkyMcUnicorn »

babyent wrote: post #81
juujian wrote: post #55

I often run into LLMs writing "beginner code" that uses the most fundamental findings in really impractical ways. Trained on too many tutorials I assume. Usually, specifying the packages to use and asking for something less convoluted works really well. Problem is, how would you know if you have never learned to code without an LLM?

I used to really like Claude for code tasks but lately it has been a frustrating experience. I use it for writing UI components because I just don’t enjoy FE even though I have a lot of experience on it from back in the day. I tell it up front that I am using react-ts and mui. 80% of the time it will use tailwind classes which makes zero sense. It won’t use the sx prop and mui system. It is also outdated it seems. It…

To each their own, and everyone's experience seems to vary, but I have a hard time picturing people using Claude/ChatGPT web UIs for any serious developmen. It seems like so much time would he wasted recreating good context, copy/pasting, etc.

We have tools like Aider (which has copy/paste mode if you don't have API access for some reason), Cline, CoPilot edit mode, and more. Things like having a conventions file and exposing the dependencies list and easy additional of files into context seem essential to me in order to make LLMs productive, and I always spend more time steering results when easy consistent context isn't at my fingertips.

Re: Can LLMs write better code if you keep asking them to “write better code”?

#114

Post by Shakahs »

babyent wrote: post #81
juujian wrote: post #55

I often run into LLMs writing "beginner code" that uses the most fundamental findings in really impractical ways. Trained on too many tutorials I assume. Usually, specifying the packages to use and asking for something less convoluted works really well. Problem is, how would you know if you have never learned to code without an LLM?

I used to really like Claude for code tasks but lately it has been a frustrating experience. I use it for writing UI components because I just don’t enjoy FE even though I have a lot of experience on it from back in the day. I tell it up front that I am using react-ts and mui. 80% of the time it will use tailwind classes which makes zero sense. It won’t use the sx prop and mui system. It is also outdated it seems. It…

Both these issues can be resolved by adding some sample code to context to influence the LLM to do the desired thing.

As the op says, LLMs are going to be biased towards doing the "average" thing based on their training data. There's more old backend code on the internet than new backend code, and Tailwind is pretty dominant for frontend styling these days, so that's where the average lands.

Re: Can LLMs write better code if you keep asking them to “write better code”?

#116

Post by freehorse »

vykthur wrote: post #107

I find that it is IMPORTANT to never start these coding sessions with "write X code". Instead, begin with a "open plan" - something the author does allude to (he calls it prompt engineering, I find it also works as the start of the interaction). Half the time, the LLM will make massive assumptions about your code and problem (e.g., about data types, about the behaviors of imported functions, about unnecessary optimiz…

> I find that it is IMPORTANT to never start these coding sessions with "write X code". Instead, begin with a "open plan"

Most llms that I use nowadays usually make a plan first on their own by default without need to be especially prompted. This was definitely not the case a year ago or so. I assume new llms have been trained accordingly in the meantime.

Re: Can LLMs write better code if you keep asking them to “write better code”?

#117

Post by cjbgkagh »

_andrei_ wrote: post #106
dgacmu wrote: post #80

I'm amused that neither the LLM or the author identified one of the simplest and most effective optimizations for this code: Test if the number is max _before_ doing the digit sum. It's a free 5.5x speedup that renders some of the other optimizations, like trying to memoize digit sums, unnecessary. On an m1 macbook pro, using numpy to generate the random numbers, using mod/div to do digit sum: Base: 55ms Test before…

That's a prompting issue though.

Do you have an example prompt that works?

Re: Can LLMs write better code if you keep asking them to “write better code”?

#118

Post by scosman »

By iterating it 5 times the author is using ~5x the compute. It’s kinda a strange chain of thought.

Also: premature optimization is evil. I like the first iteration most. It’s not “beginner code”, it’s simple. Tell sonnet to optimize it IF benchmarks show it’s a pref problem. But a codebase full of code like this, even when unnecessary, would be a nightmare.

Re: Can LLMs write better code if you keep asking them to “write better code”?

#119

Post by ben_w »

b112 wrote: post #82
ben_w wrote: post #65

Earlier quoted context omitted.

Even as someone with plenty of experience, this can still be a problem: I use them for stuff outside my domain, but where I can still debug the results. In my case, this means I use it for python and web frontend, where my professional experience has been iOS since 2010. ChatGPT has, for several generations, generally made stuff that works, but the libraries it gives me are often not the most appropriate, and are som…

I wish people would understand what a large language model is. There is no thinking. No comprehension. No decisions. Instead, think of your queries as super human friendly SQL. The database? Massive amounts of data boiled down to unique entries with probabilities. This is a simplistic, but accurate way to think of LLMs. So how much code is on the web for a particular problem solve? 10k blog entries, stackoverflow res…

> There is no thinking. No comprehension. No decisions.

Re-reading my own comment, I am unclear why you think it necessary to say those specific examples — my descriptions were "results, made, disagree, right/wrong, struggle": tools make things, have results; engines struggle; search engines can be right or wrong; words can be disagreed with regardless of authorship.

While I am curious what it would mean for a system to "think" or "comprehend", every time I have looked at such discussions I have been disappointed that it's pre-paradigmatic. The closest we have is examples such as Turing 1950[0] saying essentially (to paraphrase) "if it quacks like a duck, it's a duck" vs. Searle 1980[1] which says, to quote the abstract itself, "no program by itself is sufficient for thinking".

> I guess my point is, when you use LLMs for tasks, you're getting whatever other humans have said.

All of maths can be derived from the axioms of maths. All chess moves derive from the rules of the game. This kind of process has a lot of legs, regardless of if you want to think of the models as "thinking" or not.

Me? I don't worry too much if they can actually think, not because there's no important philosophical questions about what that even means, but because other things have a more immediate impact: even if they are "just" a better search engine, they're a mechanism that somehow managed to squeeze almost all of the important technical info on the internet into something that fits into RAM on a top-end laptop.

The models may indeed be cargo-cult golems — I'd assume that by default, there's so much we don't yet know — but whatever is or isn't going on inside, they still do a good job of quacking like a duck.

[0] Turing, A. M. (1950). Computing machinery and intelligence. Mind, 59, 433–460. https://doi.org/10.1093/mind/LIX.236.433

[1] Searle, J. R. (1980). Minds, brains, and programs. Behavioral and Brain Sciences, 3(3), 417–424. https://doi.org/10.1017/S0140525X00005756

Re: Can LLMs write better code if you keep asking them to “write better code”?

#120

Post by seba_dos1 »

alkonaut wrote: post #89
moffkalast wrote: post #48

Earlier quoted context omitted.

Don't they? It highly depends on the errors. Could range from anything like a simple syntax error to a library version mismatch or functionality deprecation that requires some genuine work to resolve and would require at least some opinion input from the user. Furthermore LLMs make those kinds of "simple" errors less and less, especially if the environment is well defined. "Write a python script" can go horribly wron…

They could, but if the LLM can iterate and solve it then the user might not need to know. So when the user input is needed, at least it's not merely to do what I do know: feed the compiler messages or test failures back to ChatGPT who then gives me a slightly modified version. But of course it will fail and that will need manual intervention. I often find that ChatGPT often reasons itself to a better solution (perhap…

> The problem with this workflow is that it's like having a bad intern on the phone pair programming.

Even worse than that - an intern has a chance to learn from this experience, get better and become a senior one day.

Post reply on HN