Let me guess, you saw the `–` and thought "Clearly no human uses the em dash. It must all be AI".
I looked over the website and their github page and saw no indication of AI usage. I'm not saying it wasn't used, but unless you have actual evidence, then please point it out.
That’s an en-dash – and I use those myself. (I’ve experimented with em-dashes—brefly—but it doesn’t look good to my eye in English.)
It’s all by the vibes, I don’t have any concrete proof. FWIW, the comment seems at least hand-edited, so maybe it’s just this “the [A] is the [B]” trope that has triggered me. If the rest of it is readable, then I apologize, but still: that’s a bad title, please don’t use this – unless you’re writing a LinkedIn post.
Not unlike how in the trades, a tee is T-shaped, a wye is Y-shaped, a delta is Δ-shaped. Spelled-out letters aren't always geometric metaphors, but it's a thing.
Then you've got the first syllable of Ω (omega) being a homophone of Ohm, for what that's worth. From there we have ℧ (mho) for conductance... It's all quite clever. But I digress; I'll get back to emceeing and deejaying, cueing these elpees after a quick pee, okay?
Different fonts use different lengths (sometimes en dash is the width of an uppercase “N” instead of lowercase, sometimes it’s half of the em dash which itself is `1em` which may or may not equal the width of the “M”...) But it’s a good way to remember that indeed, yeah!
After enough time spent reading LLM speak, I’ve found its phrasing and particularities subconsciously work into the way I write as well, entirely by hand.
It’s not always AI. It could be a human ;) I really despise LLM-written writing too; but I think it’s nice to be not super paranoid about it.
For me (not the OP) I see the patterns 'The X is the Y', 'Every X is a Y' (both of which are there) and llm-sense tingles. Add in Not A, B, C or Z and I'm increasingly convinced it's LLM generated text.
> It's also a concise and accurate description of the service.
If it was a clever expression that does get the idea across, I would be okay with it. “The API is the UI” is of similar form and I love it. It’s simple, catchy, and it does convey a meaning: designing an API is just designing a UI for developers.
“The URL is the app” makes literally no sense without additional context? I’m not sure this is a “concise and accurate description of the service”. I did take a guess [1] and I was correct, but that was after I’ve seen the examples.
> “The URL is the app” makes literally no sense without additional context?
Yes exactly - it's one of those sentences that only makes sense to people who already know what it was intended to mean, because they know how it's intended to be parsed without having to puzzle over it. For the rest of us, it's pretty meaningless.
Regarding the LLMism, I didn't click on this submission when I saw it on HN last night, because I'm sick of seeing "The X is the Y", and an LLMish title tends to imply yet another piece of vibecoded slop.
My man, that web design is absolutely pure Claude.
But hey that’s ok, the app is a good idea and works well, I have no issue with that.
Clearly the person making it knows how to structure a web app and make it usable and it shows it’s not just slop.
This is good use of AI coding
This isn't human writing. This is blustering, bloviated Claudespeak we have seen a thousand times over. Try Opus 5 and its ilk sometime and maybe you'll see it too.
The URL is the whole display: no accounts, no login, no backend.
Nothing else: no lists, tables, links, quotes, images or rules.
Underscores inside words are left alone, as in standard Markdown, so names like Home_Guest_5G, snake_case or first_name@example.com show exactly as typed. Only an underscore at a word boundary starts or ends emphasis.
Code is shown exactly as typed. Nothing inside backticks is treated as formatting, a slide break, {countdown} or an escape, which makes it the easy way to show passwords and codes with symbols in them. A code span has to open and close on the same line. Browsers may write a typed backtick as %60 in the address bar; both forms work.
All fonts ship with the site, each next to its license file. A font is only downloaded, from this site, when a display uses it.
Screen wake: with wake=on (the default) the viewer asks the browser to keep the screen on. It asks again whenever the tab becomes visible. Browsers without support ignore it.
The viewer can be embedded anywhere (OBS, dashboards); the editor and docs can't. HSTS is left to the TLS-terminating reverse proxy. The app reads all state from the URL fragment, so / serves every display. /editor and /docs are their own files (editor.html, docs.html); anything else is a real 404.
img-src stays open because displays show images from any host. Inline styles are needed for the runtime @font-face rules and style attributes. A snippet, so 404 responses get these headers too.
Exercises the canonical URLs, social cards and sitemap, which are only built when SITE_URL is set. The smoke test reads it too.
Don't get me started on the docstrings and code comments.
I looked over the website and their github page and saw no indication of AI usage. I'm not saying it wasn't used, but unless you have actual evidence, then please point it out.