Generative AI coding tools and agents do not work for me • Hacker News Board

Re: Generative AI coding tools and agents do not work for me

#261

Post by blueflow » Tue, Jun 17, 2025, 8:57 AM UTC

BlackFly wrote: post #257

socalgal2 wrote: post #180

> Another common argument I've heard is that Generative AI is helpful when you need to write code in a language or technology you are not familiar with. To me this also makes little sense. I'm not sure I get this one. When I'm learning new tech I almost always have questions. I used to google them. If I couldn't find an answer I might try posting on stack overflow. Sometimes as I'm typing the question their search wo…

One of the many ways that search got worse over time was the promotion of blog spam over actual documentation. Generally, I would rather have good API documentation or a user guide that leads me through the problem so that next time I know how to help myself. Reading through good API documentation often also educates you about the overall design and associated functionality that you may need to use later. Reading the…

You parent searches for answers, you search for documentation. Thats why AI works for him and not for you.

Re: Generative AI coding tools and agents do not work for me

#262

Post by layer8 » Tue, Jun 17, 2025, 9:00 AM UTC

sagarpatil wrote: post #87

jumploops wrote: post #2

> It takes me at least the same amount of time to review code not written by me than it would take me to write the code myself, if not more. As someone who uses Claude Code heavily, this is spot on. LLMs are great, but I find the more I cede control to them, the longer it takes to actually ship the code. I’ve found that the main benefit for me so far is the reduction of RSI symptoms, whereas the actual time savings a…

I always use Claude Code to debug issues, there’s no point in trying to do this yourself when AI can fix it in minutes (easy to verify if you write tests first) o3 with new search can do things in 5 mins that will take me at least 30 mins if I’m very efficient. Say what you want but the time savings is real.

Tests can never verify the correctness of code, they only spot-check for incorrectness.

Re: Generative AI coding tools and agents do not work for me

#263

Post by kypro » Tue, Jun 17, 2025, 9:01 AM UTC

plasticeagle wrote: post #206

socalgal2 wrote: post #180

> Another common argument I've heard is that Generative AI is helpful when you need to write code in a language or technology you are not familiar with. To me this also makes little sense. I'm not sure I get this one. When I'm learning new tech I almost always have questions. I used to google them. If I couldn't find an answer I might try posting on stack overflow. Sometimes as I'm typing the question their search wo…

ChatGPT and Gemini literally only know the answer because they read StackOverflow. Stack Overflow only exists because they have visitors. What do you think will happen when everyone is using the AI tools to answer their questions? We'll be back in the world of Encyclopedias, in which central authorities spent large amounts of money manually collecting information and publishing it. And then they spent a good amount o…

The idea that LLMs can only spew out text they've been trained on is a fundamental miss-understanding of how modern backprop training algorithms work. A lot of work goes into refining training algorithms to preventing overfitting of the training data.

Generalisation is something that neural nets are pretty damn good at, and given the complexity of modern LLMs the idea that they cannot generalise the fairly basic logical rules and patterns found in code such that they're able provide answers to inputs unseen in the training data is quite an extreme position.

Re: Generative AI coding tools and agents do not work for me

#264

Post by vidarh » Tue, Jun 17, 2025, 9:01 AM UTC

smaudet wrote: post #244

vidarh wrote: post #213

Earlier quoted context omitted.

> The second portion of your statement is either confusing (something unsaid) or untrue (you are still ultimately on the hook). You're missing the point. tptacek is saying he isn't the one who needs to fix the issue because he can just reject the PR and either have the AI agent refine it or start over. Or ultimately resort to writing the code himself. He doesn't need to make the AI written code work , and so he doesn…

> But no template based generator could write that code, even though it's fairly trivial Not true at all, in fact this sort of thing used to happen all the time 10 years ago, code reading APIs and generating clients... > He doesn't need to make the AI written code work, and so he doesn't need to spend a lot of time reading the AI written code - he can skim it for any sign it looks even faintly off and just kill it if…

> Not true at all, in fact this sort of thing used to happen all the time 10 years ago, code reading APIs and generating clients...

Where is this template based code generator that can read my code, understand it, and generate a full client including a CLI, that include knowing how to format the data, and implement the required protocols?

I'm 30 years of development, I've seen nothing like it.

> I think you are missing the point as well, that's still review, that's still being on the hook.

I don't know if you're being intentionally obtuse, or what, but while, yes, you're on the hook for the final deliverable, you're not on the hook for fixing a specific instance of code, because you can just throw it away and have the AI do it all over.

The point you seem intent on missing is that the cost of throwing out the work of another developer is high, while the cost of throwing out the work of an AI assistant is next to nothing, and so where you need to carefully review a co-workers code because throwing it away and starting over from scratch is rarely an option, with AI generated code you can do that at the slightest whiff of an issue.

> Words like "skim" and "kill" are the problem here, not a solution. They point to a broken process that looks like its working...until it doesn't.

No, they are not a problem at all. They point to a difference in opportunity cost. If the rate at which you kill code is too high, it's a problem irrespective of source. But the point is that this rate can be much higher for AI code than for co-workers before it becomes a problem, because the cost of starting over is orders of magnitude different, and this allows for a very different way of treating code.

> Well, you just have to sort of hope it works and when it doesn't

No, I don't "hope it works" - I have tests.

Re: Generative AI coding tools and agents do not work for me

#265

Post by layer8 » Tue, Jun 17, 2025, 9:05 AM UTC

cbsmith wrote: post #182

Terr_ wrote: post #148

Earlier quoted context omitted.

I expect that comes from the contrast and synthesis between how the author is anticipating things will develop or be explained, versus what the other person actually provided and trying to understand their thought process. What happens if the reader no longer has enough of that authorial instinct, their own (opinionated) independent understanding? I think the average experience would drift away from "I thought X was…

It seems counter intuitive that the reader would no longer have that authorial instinct due to lack of writing. Like, maybe they never had it, in which case, yes. But being exposed to a lot of different "writing opinions" tends to hone your own. Let's say you're right though, and you lose that authorial instinct. If you've got five different proposals/PRs from five different models, each one critiqued by the other fo…

I don’t find this convincing. People generally don’t learn how to write a good novel just by reading a lot of them.

Re: Generative AI coding tools and agents do not work for me

#266

Post by yard2010 » Tue, Jun 17, 2025, 9:06 AM UTC

socalgal2 wrote: post #180

> Another common argument I've heard is that Generative AI is helpful when you need to write code in a language or technology you are not familiar with. To me this also makes little sense. I'm not sure I get this one. When I'm learning new tech I almost always have questions. I used to google them. If I couldn't find an answer I might try posting on stack overflow. Sometimes as I'm typing the question their search wo…

I think the main issue here is trust. When you google something you develop a sense for bullshit so you can "feel" the sources and weigh them accordingly. Using a chat bot, this bias doesn't hold, so you don't know what is just SEO bullshit reiterated in sweet words and what's not.

Re: Generative AI coding tools and agents do not work for me

#267

Post by hatefulmoron » Tue, Jun 17, 2025, 9:10 AM UTC

vidarh wrote: post #255

hatefulmoron wrote: post #253

Earlier quoted context omitted.

> If you let it paint you into a corner, why are you doing so? What do you mean? If it were as simple as not letting it do so, I would do as you suggest. I may as well stop letting it be incorrect in general. Lots of guidance helps avoid it. > Maybe I just naturally do that, but to me there's not been much change in the level of guidance I need to give Claude Code or my own agent vs. what I'd give developers working…

> What do you mean? If it were as simple as not letting it do so, I would do as you suggest. I'm arguing it is as simple as that. Don't accept changes that muddle up the architecture. Take attempts to do so as evidence that you need to add direction. Same as you presumably would - at least I would - with a developer.

My concern isn't that it's messing up my architecture as I scream in protest from the other room, powerless to stop it. I agree with you and I think I'm being quite clear. Without relatively close guidance, it will paint you into a corner in terms of architecture. Guide it, direct it, whatever you want to call it.

Re: Generative AI coding tools and agents do not work for me

#268

Post by kibibu » Tue, Jun 17, 2025, 9:12 AM UTC

marssaxman wrote: post #70

kibibu wrote: post #25

Earlier quoted context omitted.

I have never seen people work harder to dismantle their own industry than software engineers are right now.

We've been automating ourselves out of our jobs as long as we've had them; somehow, despite it all, we never run out of work to do.

We've automated bullshit tedium work, like building and deploying, but this is the first time in my memory that people are actively trying to automate all the fun parts away.

Closest parallel I can think of is the code-generation-from-UML era, but that explicitly kept the design decisions on the human side, and never really took over the world.

Re: Generative AI coding tools and agents do not work for me

#269

Post by usef- » Tue, Jun 17, 2025, 9:14 AM UTC

nobunaga wrote: post #260

olmo23 wrote: post #250

Earlier quoted context omitted.

Where does the knowledge come from? People can only post to SO if they've read the code or the documentation. I don't see why LLMs couldn't do that.

ITT: People who think LLMs are AGI and can produce output that the LLM has come up with out of thin air or by doing research. Go speak with someone who is actually an expert in this field how LLMs work and why the training data is so important. Im amazed that people in the CS industry seem to talk like they know everything about a tech after using it but never even writing a line of code for an LLM. Our indsutry is d…

This isn't about being AGI or not, and it's not "out of thin air".

Modern implementations of LLMs can "do research" by performing searches (whose results are fed into the context), or in many code editors/plugins, the editor will index the project codebase/docs and feed relevant parts into the context.

My guess is they either were using the LLM from a code editor, or one of the many LLMs that do web searches automatically (ie. all of the popular ones).

They are answering non-stackoverflow questions every day, already.

Re: Generative AI coding tools and agents do not work for me

#270

Post by ritz_labringue » Tue, Jun 17, 2025, 9:15 AM UTC

AI is really useful when you already know what code needs to be written. If you can explain it properly, the AI will write it faster than you can and you'll save time because it is quick to check that this is actually the code you wanted to write. So "programming with AI" means programming in your mind and then using the AI to materialize it in the codebase.