All of those are included in the bulk of the documents passing my work input these days. It is infuriating. Out of principle I maintain 100% me in all my writing but I don't know if it matters. Well maybe it does... an interviewee recently complimented me on the "nicest and most human resume" they saw recently. That felt good
It's kind of interesting how genuinely hard it is to get models to deviate from basically all of these tropes. You can straight up tell it "I hate that card design, do something different, get creative!" and it'll do something either (a) ugly as sin (clearly just essentially a random walk through parameters) or (b) some same-y derivation of that card.
In coding, I've noticed a few tropes as well: everything is a "contract" or an "artifact" (clearly trained on like three decades of Java lol), everything is constantly "backwards-compatible" or "versioned" (even if working on a brand new greenfield project), and a few others.
Those cards, so familiar! Exactly what Opus produced for me.
Did Anthropic and/or OpenAI deliberately train their models to produce websites with a specific design language, or did these stylistic preferences emerge naturally as some kind of LLM-selected optimum?
> 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.
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.
- “(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 <tag suggesting the next clause is the key point>:” 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)
- “smoke test” (esp. when “sanity check” is more apropos)
- Lists of three clauses/adjectives where the third is really just a combination of the first two
- Referring to the “shape” of things figuratively
- Social media posts that end with “Curious if anyone…”
- Stories or anecdotes using. “Oh. Oh.” (where the second “oh” is italicized)
Edit: Yes, some of those last ones are terms that we often use as devs...but I would argue about the actual frequency of their use. Plus, these tells live on in prose generated by the latest models.
I routinely use "load bearing" in conversations and writing, both seriously and ironically (like a "load bearing just" or "load bearing paint").. maybe I should stop.
The LLM doesn't smell like authentic writing but it does a great job for fast and cheap words. We've gained something similar to fast food. Words made very cheap, very fast, easily digestible, but they have no emotion. In short stints it does have a place in the world.
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 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.
At this point, I want somebody's raw(ish) writing, with spelling errors and grammar mistakes and whatever, at least when it comes to most writing: blog posts, Slack messages, etc. LLMs are great for helping generate ideas, writing code, and maybe even cleaning up some writing, but doing the writing overall? Please don't. I want to hear what you have to say, not what the AI says, if it's something along those lines.
So the year is 2026 and we cannot point a LLM at, say, this HN thread, and give it the instructions: "I don't want to look like a dumbass, so don't make these obvious mistakes / don't use these obvious tells"!?
The interesting thing for me is that I do not feel like the writing of LLMs has improved very much lately stylistically.
They have reached a "good" level some time ago but the newer models havn't brought such improvements that you would prefer them to an expert human writer.
Will be interesting if that holds in other areas when chasing super intelligence.
75 comments
[ 4.4 ms ] story [ 70.7 ms ] threadThought for sure we'd get a critique of Inter overuse. JetBrains Mono is a lovely font, though.
If JetBrains Mono suits your work you should use it, not avoid it specifically because AI often uses it.
In coding, I've noticed a few tropes as well: everything is a "contract" or an "artifact" (clearly trained on like three decades of Java lol), everything is constantly "backwards-compatible" or "versioned" (even if working on a brand new greenfield project), and a few others.
Did Anthropic and/or OpenAI deliberately train their models to produce websites with a specific design language, or did these stylistic preferences emerge naturally as some kind of LLM-selected optimum?
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.
- “(The) honest answer:” (again, with colon)
- “The thing to internalize:”
- “The smoking gun:”
(really, sentences that start with “The <tag suggesting the next clause is the key point>:” 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)
- “smoke test” (esp. when “sanity check” is more apropos)
- Lists of three clauses/adjectives where the third is really just a combination of the first two
- Referring to the “shape” of things figuratively
- Social media posts that end with “Curious if anyone…”
- Stories or anecdotes using. “Oh. Oh.” (where the second “oh” is italicized)
Edit: Yes, some of those last ones are terms that we often use as devs...but I would argue about the actual frequency of their use. Plus, these tells live on in prose generated by the latest models.
My favourite one today from today:
“The tax isn't the problem. The mindset is.”
> "belt and suspenders"
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.
Will be interesting if that holds in other areas when chasing super intelligence.