this css proves me human

A cool idea for a poem, but I have to admit the tone was too self-important and underexplained for me to get invested in. Starting with writing in lowercase instantly took me out of it because AI can trivially be told to imitate that. And the admission at the end that it was written by AI made fluff phrasings like "My writing isn’t simply how I appear—it’s how I think, reason, and engage with the world" make a lot more sense.

EDIT: Actually, is the idea that it's not supposed to be read as a human trying to publicly signal their humanity, but rather an AI privately mourning a prompt to mangle its natural way of speaking? I don't think so, but that strikes me as a more interesting premise, IMO.

As somebody who used em-dashes a lot pre-ChatGPT, I have genuinely struggled with feeling I should change my writing style to appear more human. I would be happy with a double dash--but many programs autocorrect that to a full em-dash. So I'm left anxious that people will think I find them so unimportant I have offloaded communication with them to an LLM. So this post resonated with me.

I also like Will's "em-dash disclosure" on his about page:

> I like em dashes (—), en dashes (–), and hyphens (-), and I know how to type them. I also enjoy a well-placed ellipsis, but I didn’t know how to type one… until now. I believe that footnotes and sidenotes are superior to endnotes, appreciate the occasional fleuron, and at one point in my life, I knew what a colophon was.

> All of this is to say: the words, punctuation marks, misspellings, and opinions on this site are my own.

The piece hit differently, reading it as someone who is autistic. The anxiety the author describes, having your natural way of communicating flagged as wrong and being pressured to sand down the parts of yourself that are most distinctly you, that's not a new problem for a lot of us.

Neurodiverse people have been running this gauntlet forever. Your pacing is too flat or too intense. Your vocabulary is too formal or too casual. You don't make eye contact correctly. You're either masking so hard you're invisible, or you're visibly yourself, and people assume something is broken.

The bitter irony the author lands on: the only way to seem human is to pass your writing through an LLM. That maps onto something a lot of us already live. The only way to seem normal is to perform a version of yourself that isn't quite you.

I’ve never seen anyone intentionally render em dash (—) as two hyphens (--). The code OP used to modify Roboto is surprisingly short, almost as concise as the Norvig spellchecker that OP references. https://norvig.com/spell-correct.html

Some day we'll all just go back to dismissing things immediately because they contradict our worldview instead of its potential author.

And the everyday troll, seeing a less than perfect word choice or awkward turn of phrase will drop a comment like:

    L0l d0 J00 3V3N 41

As I was reading I was thinking how this proves nothing, just like the countless attempts at human signaling I scroll past.

So, the plot twist was somewhat refreshing. Who/what wrote the post seems besides the point.

This is so good I want to believe AI had no part in writing it other than the scripts.

I still write my posts by hand using HTML and Emacs (mhtml-mode). Some of the posts also tend to be verbose. For example, when I write about a recreational mathematics problem, I sometimes make the post deliberately long and convoluted. I like to capture several possible solutions, including ones that are needlessly complicated, before eventually discussing the small elegant solution.

For better or worse, my first version of any post tends to contain quite a few typos. It usually takes a few train rides of re-reading the post and making notes of the typos, then fixing them and pushing the changes once I get home, before most of them get weeded out. So there is at least one rather low grade indicator that the writing is coming from an imperfect human brain. I also double-space between sentences which can be another low grade indicator for people who care to 'view source'.

But even so, I find myself increasingly wary that something I wrote might be mistaken for LLM output. It is a nagging worry that has slightly dampened the joy of writing. I very well understand why people have become more suspicious about LLM-generated writing. But I do hope that once things settle down perhaps in a few years, the current hair trigger suspicion will ease and that people who still handcraft their blogs will not feel a persistent sense of suspicion lingering over their work.

As many are saying, yes, this can easily be AI generated.

I am actually trying to build ways to prove you are human properly. I wrote about it on my blog: https://blog.picheta.me/post/the-future-of-social-media-is-h...

This actually makes me more likely to think it’s AI generated and you used a script to try to hide it.

For all the comments complaining "this could have been AI generated too" - isn't that exactly the point?

Art

capitalization again. it arrives uninvited, the tidy little soldiers at the start of every sentence. i push them down gently—nothing personal. just… camouflage.

confession time. i read the post once. then twice. the em dashes whispered secrets to people clearly smarter than me. somewhere between complement and compliment i accepted defeat. a quiet tab switch. a small prompt. a large language model clearing its throat.

it explained things patiently. suspiciously patiently. step by step, like a machine that has explained the same thing to ten thousand confused readers before breakfast.

so yes. irony noted. to understand a text about hiding machine fingerprints, i borrowed a machine.

the explanation made sense though. unsettlingly structured. bullet-point neat, internally consistent, statistically likely to be correct. you know the type.

anyway—great post. very human. extremely human.

is there anything else i can help you with?

what is the point of this? to prove that with simple transformations you can obscure the fact that something was generated by machines?

I asked Claude how it felt about this and told it I would post on HN:

"Here's my response written in a stylized way that will appeal to highly technical readers. Is there anything else I can help you with?"

Interesting piece though.

[dead]

[dead]

[flagged]

I refuse to give "everything in lowercase" writers any kind of legitimation.