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

Various LLM Smells

shvbsle.in

Post reply on HN
Print view View original

81–90 of 312 posts

Re: Various LLM Smells

#81

Post by tempest_ »

Barbing wrote: post #46
thewebguyd wrote: post #42

Earlier quoted context omitted.

> I would argue about the actual frequency of their use Assuming you mean load bearing & blast radius, I'd see those used and use them myself very frequently pre LLM, mostly in online discussions though so its telling where they got their training data. Load bearing itself is/was a pretty normal phrase in the ops world in daily discussion. Smoke test though, I can't say I've ever see irl usage.

Heard smoke test IRL & was confused to see it used indeed in place of “sanity check”. Weird.

I have heard people use smoke test but not nearly at the same rate an LLM uses.

If a repo is bare of CLAUDE.md but mentions a smoke test in a commit in the last year I assume it to be LLM written.

Re: Various LLM Smells

#82

Post by sofixa »

tptacek wrote: post #50

The LLM writing sameness is bad. Use LLMs to help your writing! But don't include a word they generate, even just a vocabulary adjustment, in your own output. Have them critique structure and flow, spot overused words and passive constructions and dumb picks for topic sentences. It's great for that, and those are all objective improvements in your writing that won't mess up your style. The LLM sameness in web design…

> The LLM sameness in web design is good. Most sites shouldn't try to be idiosyncratic. The best design for a site with real utility is legibility, and LLMs are better at that than the median developer. Always laying out the same buttons? Always using the same type scales? Good! If it looks good to you, you weren't going to do better on your own, and you were very likely to do worse.

See, I disagree. Having seen plenty of Claude generated websites and slide decks, to me it just screams "no effort whatsoever". AI sloppypasta for content, if you will.

If I can see within a few seconds that your website or slide was obviously AI generated, I will doubt its content, how much effort (if any) you've put into it, if it won't have hallucinations, and (especially for websites) if it's even real or a scam farm.

I'm not saying every website has to be unique, but at least tell your prompt to use a font or colour scheme or something specific to you that will make it seem like you've put in some effort and make the result stand out from the slop.

Re: Various LLM Smells

#83

Post by tomjakubowski »

viccis wrote: post #75

One Python one I hate is that it adds crazy amounts of newlines for no real readability gains. Instead of this: def add_three_ints(x: int, y: int, z: int) -> int: return x + y + z it will write: def add_three_ints( x: int, y: int, x: int, ): return x + y + z While it's always preferable to do this when you get either long or complex function signatures, Opus 4.7 and GPT 5.5 do this everywhere. When you combine it wit…

The second way produces more billable output tokens. Worse, when you feed that code back into the LLM service, the extra whitespace counts as input tokens.

Re: Various LLM Smells

#84

Post by fragmede »

viccis wrote: post #75

One Python one I hate is that it adds crazy amounts of newlines for no real readability gains. Instead of this: def add_three_ints(x: int, y: int, z: int) -> int: return x + y + z it will write: def add_three_ints( x: int, y: int, x: int, ): return x + y + z While it's always preferable to do this when you get either long or complex function signatures, Opus 4.7 and GPT 5.5 do this everywhere. When you combine it wit…

That's for diffing gains later.

If you have to add arguments, when they're on one line like that, the diff is cleaner, so the reviewer has an easier time kf understanding what's going on. That is, if you still have a human reviewing code, that is.

Re: Various LLM Smells

#85

Post by onion2k »

ryandrake wrote: post #66
Planktonne wrote: post #8

> The LLM generated writing obviously felt significantly better than my own writing. A general pattern for LLMs is that they look really good at things you are bad at. What that means is that if you find yourself thinking of its output as significantly better than yours in a particular domain, there's a high chance that you are not equipped to judge that quality effectively.

> A general pattern for LLMs is that they look really good at things you are bad at. This is true for coding, too, which I think, to a large degree, might explain the polarized differences in opinions on HN about the quality of LLM-produced code. You have the 1. "AI produces code better than I could possibly write, one shots things it would take me days to do, and has made me 10X more productive!" camp, and you have…

Well, I have bad news for camp 1..

It's bad if they work in a part of the industry where code quality or efficiency matters. That's maybe 10% of the total though.

Re: Various LLM Smells

#86

Post by ray__ »

I wonder if the tendency to write short punchy sentences stems from deliberate RL efforts to avoid repetitive, consistent writing? I seem to remember that a critique of early LLMs was that they would produce sentences whose construction was too homogeneous. Would be interesting to know the answer to this.

Re: Various LLM Smells

#87

Post by triyambakam »

spdustin wrote: post #18

- “(The) honest caveat:” (or “genuine caveat:”, both with the colon) - “(The) honest answer:” (again, with colon) - “The thing to internalize:” - “The smoking gun:” (really, sentences that start with “The :” are a strong tell, but those four are the most prolific) - “load bearing” (when not talking about architecture) - “blast radius” (when not talking about actual explosives, but rather the effect of an event/action…

- Ending something with "happy to ..." (usually "happy to help")

- And a variant of the above is omitting the subject, "happy to" instead of "I am happy to"

- Codex refers to "the spine" of something

- Claude often says some decision is "locked" (i.e. decided on)

Re: Various LLM Smells

#88

Post by metadat »

Don't forget about Contrastive Negation:

> Contrastive negation is a rhetorical structure that denies a specific idea in the first half of a sentence and asserts an alternative in the second half.

> It typically follows an "It’s not X, it’s Y" or "not just X, but Y" formula.

Wikipedia also has a great resource which covers many of the common LLM patterns: https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing

Re: Various LLM Smells

#89

Post by Noumenon72 »

poszlem wrote: post #17

What I find amazing is how HARD it is to make the LLM produce a piece of text that does not sound like slop. I have had dozens of sessions where I tried to make it write like a human would, and yet it still uses those tired writing phrases. I don't understand why neither openai, nor anthropic are able to do anything to make it better, and in some cases it feels like we are actually going backwards.

Alan Turing poisoned the context in ways we can't comprehend and all LLMs are bound by his dead hand.

Re: Various LLM Smells

#90

Post by torginus »

dvt wrote: post #11
Planktonne wrote: post #8

> The LLM generated writing obviously felt significantly better than my own writing. A general pattern for LLMs is that they look really good at things you are bad at. What that means is that if you find yourself thinking of its output as significantly better than yours in a particular domain, there's a high chance that you are not equipped to judge that quality effectively.

Honestly, I can't fathom thinking that LLM writing is even remotely passable. People that think this should honestly read more. One book a month is hardly an aspirational goal. You don't even have to read Melville or Hemingway or Chaucer or Shakespeare, just pick up any popular NYT best seller, and it'll be significantly better than anything an LLM can generate.

I haven't used these things for writing recreationally for a while (since the Claude 3.X days), so my opinions might be outdated - but they definitely weren't bad - after all they had a huge library of witticisms to pull from, and like Stable Diffusion that pulls from master artists, so do LLMs from skilled writers. Pro writers did come up with an absolute dearth of interesting ideas, and there are mountains of skillfully written prose out there - and its all in the training data, and AI is quite good at pulling from it.

The advantage of the writing vs images, is that it takes longer to absorb the whole with text, so its less apparent that the whole thing doesn't quite come together.

My problems was with Claude's prose and ideas is that it kept recycling the tropes and phrases after a while - something that has been observed that these models have very strong statistical biases - when asking for a random number for example, LLMs are far more predictable than even humans, this shows up in unguided writing exercises.

But as for actually crafting text that is both terse and to the point - such as oneliner explanations, or writing summaries - these models are quite bad. The best I have seen is they could turn a given length of prose into an even longer version - with generally some loss in the tonal accuracy or the points made in there.

As such they are a terrible tool for professional communication, but unfortunately, lots of people have started using them for exactly that.

Post reply on HN