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

Human coders are still better than LLMs

antirez.com

Post reply on HN
Print view View original

351–360 of 762 posts

Re: Human coders are still better than LLMs

#351

Post by Sohcahtoa82 »

throwaway314155 wrote: post #318
bouncycastle wrote: post #314

Last night I spent hours fighting o3. I never made a Dockerfile in my life, so I thought it would be faster just getting o3 to point to the GitHub repo and let it figure out, rather than me reading the docs and building it myself. I spent hours debugging the file it gave me... It kept on adding hallucinations for things that didn't exist, and removing/rewriting other parts, and other big mistakes like understanding t…

Pro-tip: Check out Claude or Gemini. They hallucinate far less on coding tasks. Alternatively, enable internet search on o3 which boosts its ability to reference online documentation and real world usage examples. I get having a bad taste in your mouth but these tools _aren't _ magic and do have something of a steep learning curve in order to get the most out of them. Not dissimilar from vim/emacs (or lots of dev too…

> enable internet search on o3

I didn't know it could even be disabled. It must be enabled by default, right?

Re: Human coders are still better than LLMs

#352

Post by max_on_hn »

dinfinity wrote: post #227
empath75 wrote: post #200

Earlier quoted context omitted.

the LLM's themselves are making marginal gains, but the tools for using LLMs productively are getting so much better.

This. MCP/tool usage in agentic mode is insanely powerful. Let the agent ingest a Gitlab issue, tell it how it can run commands, tests etc. in the local environment and half of the time it can just iterate towards a solution all by itself (but watching and intervening when it starts going the wrong way is still advisable). Recently I converted all the (Google Docs) documentation of a project to markdown files and add…

I 100% agree that documenting requirements will be the main human input to software development in the near future.

In fact, I built an entirely headless coding agent for that reason: you put tasks in, you get PRs out, and you get journals of each run for debugging but it discourages micro-management so you stay in planning/documenting/architecting.

Re: Human coders are still better than LLMs

#353

Post by daveguy »

sigmoid10 wrote: post #158
klntsky wrote: post #72

Earlier quoted context omitted.

I would argue that they are never led astray by chatting, but rather by accepting the projection of their own prompt passed through the model as some kind of truth. When talking with reasonable people, they have an intuition of what you want even if you don't say it, because there is a lot of non-verbal context. LLMs lack the ability to understand the person, but behave as if they had it.

It's mostly a question of experience. I've been writing software long enough that when I give chat models some code and a problem, I can immediately tell if they understood it or if they got hooked on something unrelated. But junior devs will have a hell of a hard time, because the raw code quality that LLMs generate is usually top notch, even if the functionality is completely off.

> the raw code quality that LLMs generate is usually top notch, even if the functionality is completely off.

I'm not even sure what this is supposed to mean. It doesn't make syntax errors? Code that doesn't have the correct functionality is obviously not "top notch".

Re: Human coders are still better than LLMs

#354

Post by daveguy »

lupire wrote: post #330
the_af wrote: post #268

Earlier quoted context omitted.

They will also keep going in circles when you rephrase the requirements, unless with every prompt you keep adding to it and mentioning everything they've already suggested that got rejected. While humans occasionally also do this (hey, short memories), LLMs are infuriatingly more prone to it. A typical interaction with an LLM: "Hey, how do I do X in Y?" "That's a great question! A good way to do X in Y is Z!" "No, Z…

Which LLMs and which versions?

All. Of. Them. It's quite literally what they do because they are optimistic text generators. Not correct or accurate text generators.

Re: Human coders are still better than LLMs

#355

Post by xandrius »

jdiff wrote: post #338
xandrius wrote: post #252

Earlier quoted context omitted.

As if you'd know if they did.

Why would we not? If they were so effective, their effectiveness would be apparent, inarguable, and those making use of it would advertise it as a demonstration of just that. Even if there were some sort of social stigma against it, AI has enough proponents to produce copious amounts of counterarguments through evidence all on their own. Instead, we have a tiny handful of one-off events that were laboriously tuned an…

I think the main reason might be that when the output is good the developer congratulates themselves, and when it's bad they make a post or comment about how bad AI is.

Then the people who congratulate the AI for helping get yelled at by the other category.

Re: Human coders are still better than LLMs

#356

Post by r14c »

some_random wrote: post #297
parliament32 wrote: post #281

Earlier quoted context omitted.

I'd like to agree with you and remain optimistic, but so much tech has promised the moon and stagnated into oblivion that I just don't have any optimism left to give. I don't know if you're old enough, but remember when speech-to-text was the next big thing? DragonSpeak was released in 1997, everyone was losing their minds about dictating letters/documents in MS Word, and we were promised that THIS would be the key i…

How can you possibly look at what LLMs are doing and the progress made in the last ~3 years and equate it to crypto bullshit? Also it's super weird to include IoT in there, seeing as it has become all but ubiquitous.

I'm not as bearish on AI, but its hard to tell if you can really extrapolate future performance based on past improvements.

Personally, I'm more interested in the political angle. I can see that AI will be disruptive because there's a ton of money and possibly other political outcomes depending on it doing exactly that.

Re: Human coders are still better than LLMs

#357

Post by na4ma4 »

tharkun__ wrote: post #332
chucksmash wrote: post #283

Earlier quoted context omitted.

Tbf, there's a phase of learning to code where everything is pretty much an incantation you learn because someone told you "just trust me." You encounter "here's how to make the computer print text in Python" before you would ever discuss strings or defining and invoking functions, for instance. To get your start you kind of have to just accept some stuff uncritically. It's hard to remember what it was like to be in…

Fair enough on 'cutting the learning tree' at some points i.e. ignoring that you don't understand yet why something works/does what it does. We (should) keep doing that later on in life as well. But unless you teach a kid that's never done any math where `x` was a thing to program, what's so hard about understanding the concept of a variable in programming?

I think they're just using hyperbole for the watershed moment when you start to understand your first programming language.

At first it's all mystical nonsense that does something, then you start to poke at it and the response changes, then you start adding in extra steps and they do things, you could probably describe it as more of a Eureka! moment.

At some point you "learn variables" and it's hard to imagine being in the shoes of someone who doesn't understand how their code does what it does.

(I've repeated a bit of what you said as well, I'm just trying to clarify by repeating)

Re: Human coders are still better than LLMs

#358

Post by malfist »

johnecheck wrote: post #260
nevertoolate wrote: post #124

Earlier quoted context omitted.

Question is, how autonomous decision making works, nobody argues that llm can finish any sentence, but can it push a red button?

Of course it can push a red button. Trivially, with MCP. Setting up a system to make decisions autonomous is technically easy. Ensuring that it makes the right decisions, though, is a far harder task.

So it can push _a_ red button, but not necessarily the _right_ red button

Re: Human coders are still better than LLMs

#359

Post by zachlatta »

frogperson wrote: post #349

The context required to write real software is just way too big for LLMs. Software is the business, codified. How is an LLM supposed to know about all the rules in all the departments plus all the special agreements promised to customers by the sales team? Right now the scope of what an LLM can solve is pretty generic and focused. Anytime more than a class or two is involved or if the code base is more than 20 or 30…

I’d encourage you to try the 1M context window on Gemini 2.5 Pro. It’s pretty remarkable.

I paste in the entire codebase for my small ETL project (100k tokens) and it’s pretty good.

Not perfect, still a long ways to go, but a sign of the times to come.

Re: Human coders are still better than LLMs

#360

Post by malfist »

falcor84 wrote: post #350
platevoltage wrote: post #277

Earlier quoted context omitted.

Give it 3 months. There will be an AI Vibrator on the market, if there isn't one already.

I just found this MCP integration, but unfortunately I don't have a device I can test it on - https://github.com/ConAcademy/buttplug-mcp

You can fix that! I hear there's this new thing called online shopping
Post reply on HN