New webdevs feel bad when LLMs produce poor results
10 posts across 5 communities in the last 30 days; 40% from Hacker News.
“The problem with LLMs is that: popularity of an answer != correctness.That concept might work a lot of the time but you will definitely run into situations where that'll never produce a correct or working response. To actually learn something you need an environment/playground…”
“No. LLMs are rather terrible at understanding business context. Its also pretty terrible at identifying the actual change requested and to implement it in a large code snippet (anything over 500 lines rly). Best to just feed it small snippets and to understand the problem…”
“I think it's still important for young people to learn coding without the LLM. they need to see the little pieces before they can build big structures. It will be like calculators, just on a bigger scale: you learn how arithmetic works, and then you rely on the calculator when…”
Junior developers are feeling discouraged when using Large Language Models (LLMs) for coding, as the models can generate incorrect or unhelpful code and hinder learning. Some suggest carefully examining the LLM's changes to understand what went wrong.
SOURCES (25)
“I've been out of the dev role for a few years. I'd appreciate if you could help me better understand what its actually like out there. How much are you trusting LLMs for writing production code:1. Chat-esque LLM assistance while mostly hand writing code2. LLM…”
“Good luck with that. LLMs can write code but they don't know why or what they are doing. They need someone who knows the why's and the what's to control, review and take responsibility. Your company is going to burn a lot money and effort because no one in the know is in a position of influence or they have just been shouted over at by kool-aid drinkers [re: How dependent on AI is the company you work for?]”
“I was a web dev before LLMs existed. Keep cope posting though. [re: Vibe Coding doesn't make you a dev as a non-dev!]”
Using llms gives you step by step guides on how to solve specific problems. Gaining knowledge ie by reading books can equip you with the tools required to solve problems without the…
“The problem with LLMs is that: popularity of an answer != correctness.That concept might work a lot of the time but you will definitely run into situations where that'll never produce a correct or working response. To actually learn something you need an environment/playground to apply what you think you know and observe the results. Without that you're not really learning, you're jus regurgitating what people want to hear.”
“For the last four years, this is what LLMs have been doing as modus operandi:1. here's thing you asked for. And here's a lot more you didn't ask for.Then I notice there are mistakes, so I point them out. And I get this pattern:2. Oh yes, sorry for that and thanks for pointing out my mistake. I think a better thing to do that would have been XYZ.Then I see that the output is far from what I originally had in mind. I steer the LLM back by saying they have strayed and made the thing far more compli”
“I got Claude to write an LLM trainer for me using wikipedia data downloaded with kiwix.”
“be honestly, the best way to learn web dev is practise, I would not recommend any audio book. some youtube video might makes sense, one good channel I remember is WebDevSimplified, but hoenstly I haven't learnt too much new with AI nowadays. AI might be the best teacher IMO”
“Subsection titles, and how the text flows. When I read these kinds of LLM-Powered posts, I feel the same thing. Like remembering the taste of some canned food which was not great, but I had to consume to stay alive. It feels like junk food, not optimized to taste great, but doesn't reek of machine oil either. A price-optimized, slightly unhealthy substance which is somewhat compatible with my body because of food engineering magic.These posts are not only a waste of time, but it also makes me fe”
“No. LLMs are rather terrible at understanding business context. Its also pretty terrible at identifying the actual change requested and to implement it in a large code snippet (anything over 500 lines rly). Best to just feed it small snippets and to understand the problem yourself otherwise it just goes in circles. Then again I don't work as a software engineer, seems like a dying career choice if all you're doing is writing code.”
“> Some things you need to understand-others, not so much.This is the core of the matter though, knowing what you need to understand and what you can ignore is the actual programmer's skill. It requires you to have a clear mental picture of both what you are trying to build and what the underlying machine will do when you are finished.You need to understand the abstractions, but also where they leak, when they won't match reality, and how. This is why knowing computer architecture and assembly he”
“I think it's still important for young people to learn coding without the LLM. they need to see the little pieces before they can build big structures. It will be like calculators, just on a bigger scale: you learn how arithmetic works, and then you rely on the calculator when you are multiplying large numbers.My guess is it will probably take some time to incorporate LLM use into education. People who are graduating right now have a problem, being between two worlds. Those graduating in a few y”
“Yeah, even before LLMs I’d regularly forget the exact syntax for something. Id just dip into my code base and find an example to refresh my memory, or google it Now I can ask the LLM instead. Same thing, just. Bit faster”
“I think he means the savviness in the art of programming the ideas. The skill and art involved in finding interesting and efficient ways to effect ideas and solutions to problems. It is(or was) fun for a good number of us who joined before AI changed the game. Now its dismissed tending to a waste of time, hence the supposed loss in savviness over time.”
“Maybe you’re partly right, maybe not, but I don’t think this is a good analogy. Let’s say output is the most important thing, which is the implication here. But why do I care about my output? Why does a woodworker care about their output?For a woodworker, I think your analogy should be a “furniture factory,” not power tools. (Power tools in software would be like a great IDE and powerful language and framework tooling; AI is… way more than that.) I bet a “woodworker” in a furniture factory (if t”
“I can’t agree more with this. I am personally held responsible for everything the LLMs generate. When working with Codex, I have it explain every change it’s made, and I often times ask it to justify itself or give simple, concrete code examples if it’s using fancy, abstract language. I always ask it to elaborate further. I go over every test case and function. Half the time it is completely overloaded with decorators and abstractions. Most of the time, it’s smuggled in at least a few erroneous”
“This whole post reads like a frustration with a lack of thought.I think it's worth distinguishing 'LLM assistance' based on how much thought/effort went into the prompt.e.g. It's popular to hate on "my colleague just copy-pasted my question into the LLM and copy-pasted the response". Because the colleague added nothing by copy-pasting. But if I'm unfamiliar with some system, I ask a question, and someone gives an LLM response prompted with the right key words, that's more useful than I could hav”
“I'm looking for some advice for small tech teams (~3 devs) developing quickly with LLMs.As most people are, we are leaning more and more on LLMs to generate code. We use Cursor in the side-panel to maintain strict control and engineering quality standards of our application.However, we are finding more and more that the code review is a tighter bottleneck in development, with PRs stacking up quickly. The balance of how long it takes to generate vs review code has skewed in the last year. Code re”
“Management. There are ways that an LLM can speed up production, but management has taken that as carte’ Blanche to reduce estimates when there’s not yet a consensus about how to estimate projects with LLMs, overwhelming everyone.”
“I agree that learning programming is a much more complex field and it's harder to learn it. And I can see how prompting for a solution could hinder learning. But I was saying that this is an issue with the LLM usage approach, not the issue of LLM as such. For example, at my job I can give the LLM our new feature discussion and say, like, do whatever, make all decisions, etc. And then I would learn nothing. (and would get shitty result). But I can formulate each task, ask for options for thin”
“There's one thing you can try to do to feel better about this and make it useful to your learning. Figure out what the LLM did and why. Mind, you might have more luck figuring it out yourself in the first place than working backward, but if you've got some source control going you can readily trace the changes the LLM made. I'm a pretty anti LLM guy because the way it changes things tends to be in mass swaths of apparently directionless modifications, but, I also learned a lot by tak”
“Its worth picking a YouTube tutorial - that's how I got started 6 years ago and I still go for that approach today. Usually any help you need is offered through the comment section and a google search. Struggling is a part of the process and every sticking point is progress. Good luck!”
“Because you choose python, I recommend you the following: first read the official tutorial for your language https://docs.python.org/3/tutorial/index.html You can go for a more interactive alternative but the key point is to learn the language basics. Feel free to get a little bit deeper to understand how python works under the hood. You'll get a better understanding of why some things works better than others. Do something or pick an existing python project. Some python books offers some in”
“Sure.. but learning to programme is much more difficult than learning to prompt. And the latter can and will hinder the former.”