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

LLMs work best when the user defines their acceptance criteria first

blog.katanaquant.com

Post reply on HN
Print view View original

381–390 of 460 posts

Re: LLMs work best when the user defines their acceptance criteria first

#381

Post by kccqzy »

2god3 wrote: post #64
kccqzy wrote: post #60

Earlier quoted context omitted.

Yeah right. A LLM in the hands of a junior engineer produces a lot of code that looks like they are written by juniors. A LLM in the hands of a senior engineer produces code that looks like they are written by seniors. The difference is the quality of the prompt, as well as the human judgement to reject the LLM code and follow-up prompts to tell the LLM what to write instead.

Lol what. The difference is that the senior... is a senior. Ask yourself what characteristics comprises a senior vs junior... You're glossing over so much stuff. Moreover, how does the Junior grow and become the senior with those characteristics, if their starting point is LLMs?

I’m not glossing over anything. You and I are talking about the exact same thing phrased differently. How does a senior know when to reject some LLM code and start over? Experience. I don’t disagree with you but your tone is aggravating.

Re: LLMs work best when the user defines their acceptance criteria first

#383

Post by otterley »

treetalker wrote: post #244

This is my experience with how LLMs "draft" legal arguments: at first glance, it's plausible — but may be, and often is, invalid, unsound, and/or ill-advised. The catch is that many judges lack the time, energy, or willingness to not only read the documents in detail, but also roll up their sleeves and dig into the arguments and cited authorities. (Some lack the skills, but those are extreme cases.) So the plausible…

As an attorney, I’m interested in this theory. Do you have any examples that illustrate the phenomenon you describe?

Re: LLMs work best when the user defines their acceptance criteria first

#384

Post by tovej »

andai wrote: post #324
tovej wrote: post #299

Earlier quoted context omitted.

Because they don't. The chain-of-reasoning feature is really just a way to get the LLM to prompt more. The fact that it generates these "thinking" steps does not mean it is using them for reasoning. It's most useful effect is making it seem to a human that there is a reasoning process.

Is this position axiomatic or falsifiable? What would it take to change your mind?

It doesn't have to be either because the burden of proof is not on me. It's on whoever claims that chaining multiple prompts together produces thinking, even though a single prompt is just predicting n-grams.

The chain does not change the token generation process, it just artificially lengthens it.

Re: LLMs work best when the user defines their acceptance criteria first

#385

Post by tovej »

satvikpendem wrote: post #307
tovej wrote: post #298

Earlier quoted context omitted.

Could you please cite these papers. If by AI you mean LLMs, that is not supported by what I know. If you mean a theoretical world-model-based AI, that's just a tautological statement.

https://arxiv.org/abs/2305.11169 https://arxiv.org/abs/2506.02996

One conference proceeding paper and one preprint, about LLMs encoding either relative geometric information of objects or simple 2D paths.

One of the papers call this "programming language semantics", but it is using a 2D grid navigation DSL. The semantics of that language are nothing like actual programming language semantics.

These are not the same as the concept being discussed here, a human "world model" of a computer system, through which to interpret the semantics of a program.

Re: LLMs work best when the user defines their acceptance criteria first

#386

Post by tovej »

seanmcdirmid wrote: post #305
tovej wrote: post #295

Earlier quoted context omitted.

Assuming this is not a rhetorical question: no, it is not. The only "goal" is to maximize plausibility.

Again, how is that different from humans? I’m not going around trying to prove my code correct when I write it manually.

I write code to solve a problem. Not code that looks like it solves the problem if a non-technical client squints at it.

And if you don't prove your code, do you not design at all then? Do you never draw state diagrams?

Every design is an informal proof of the solution. Rarely I write formal proofs. Most of the time I write down enough for myself to be convinced that the desing solves the problem.

Re: LLMs work best when the user defines their acceptance criteria first

#387

Post by jbergqvist »

Producing the most plausible code is literally encoded into the cross entropy loss function and is fundamental to the pre-training. I suppose post training methods like RLVR are supposed to correct for this by optimizing correctness instead of plausibility, but there are probably many artifacts like these still lurking in the model's reasoning and outputs. To me it seems at least possible that the AI labs will find ways to improve the reward engineering to encourage better solutions in the coming years though.

Re: LLMs work best when the user defines their acceptance criteria first

#388

Post by 112233 »

Foobar8568 wrote: post #113
unlikelytomato wrote: post #70

Earlier quoted context omitted.

This is why I'm confused when people say it isn't ready to replace most of the programmer workforce.

LLM code is higher quality than any codes I have seen in my 20 years in F500. So yeah you need to "guide" it, and ensure that it will not bypass all the security guidance for ex...But at least you are in control, although the cognitive load is much higher as well than just "blind trust of what is delivered". But I can see the carnage with offshoring+LLM, or "most employees", including so call software engineer + LLM.

Uhuh. Let me present you Rudolph. For the next 15 minutes, he will paste pieces of top rated SO answers and top starred GH repos. Then he will suffer complete amnesia. He might not understand your question or remember what he just did, but the code he pastes is higher quality than any codes you have seen in your 20 years in F500! For 20$ a month, he's all yours, he just needs a 4 hour break every 5 hours. But he runs on money, like gumball machine, so you can wake him with a donation. Oh, you are responsible for giving him precise instructions, that he often ignores in favour of other instructions from uncle Sam. No, you can't see them.

Re: LLMs work best when the user defines their acceptance criteria first

#389

Post by geysersam »

There's also such a thing as being too ambitious. 99% of developers can not rewrite SQLite in rust even if they spent the rest or their lifetime doing it.

Expecting an AI do to a good job vibe-coding a Sqllite clone over a few weekends just isn't realistic. Despite that, it's useful technology.

Re: LLMs work best when the user defines their acceptance criteria first

#390

Post by satvikpendem »

tovej wrote: post #385
satvikpendem wrote: post #307

Earlier quoted context omitted.

https://arxiv.org/abs/2305.11169 https://arxiv.org/abs/2506.02996

One conference proceeding paper and one preprint, about LLMs encoding either relative geometric information of objects or simple 2D paths. One of the papers call this "programming language semantics", but it is using a 2D grid navigation DSL. The semantics of that language are nothing like actual programming language semantics. These are not the same as the concept being discussed here, a human "world model" of a com…

Well I didn't find any papers off the bat for code world models but if they can create a world model for the task given, such as geometric manipulation, I don't see why they wouldn't in terms of code.
Post reply on HN