This CSS proves me human and that says something ugly about the web

One of the strangest things happening on the web right now is that people are starting to sand down their own polish on purpose. Lowercase everything. Misspell a word just enough. Avoid the punctuation that now feels machine-adjacent. Leave fingerprints on the sentence so nobody mistakes you for a model.

That is what makes Will Keleher’s short essay this css proves me human worth more attention than its size suggests. It uses CSS as a joke and a metaphor at the same time. The page is styled in lowercase, the author toys with manufactured mistakes, and the whole thing builds toward a line that lands like a slap at the end of a chat window.

Keleher ends with the sentence, “here’s your blog post written in a stylized way that will appeal to highly technical readers.”

This is not really about CSS

The title is clever because CSS is only the surface layer. The real subject is identity. Keleher keeps asking how much of a writer’s style can be changed before the voice stops feeling like theirs. The essay treats capitalization, typography, spelling, and punctuation as the visible signals we now use to guess whether something came from a person or a machine.

I like the piece because it gets at a nasty little truth about online writing in 2026. We do not only judge ideas anymore. We judge smoothness. If a paragraph feels too balanced, too polished, or too eager to please, people start sniffing for model residue. That changes how humans write too. Once a style becomes associated with AI, actual people start avoiding it like a contaminated accent.

The final line is the real knife twist

The ending is what makes the essay stick. After all the anxiety about preserving a human voice, the piece ends by slipping into sterile assistant language anyway. It is funny, but it is also bleak. The joke is that once you spend enough time optimizing your prose to look human, you can still end up sounding like a helpdesk model wearing an indie-blog costume.

That is why the essay worked on Hacker News. The thread was less about front-end tricks and more about authorship, taste, and how people now detect synthetic writing by vibe alone. One commenter said the piece felt too self-important until the ending reframed it. Another said it reads better if you imagine an AI privately mourning the prompt that told it to deform its natural style. That second reading is the one I cannot shake.

One of the shortest HN defenses was also the best: “It’s proper art. It’s a mirror. It makes us reflect.”

The web is training people to cosplay themselves

That is my real takeaway. We spent years teaching models to mimic human writing, and now humans are adapting in the opposite direction to avoid being mistaken for the models. It is a miserable little feedback loop. The cleaner the models get, the more “real” people may feel pressured to leave in rough edges, private references, typos, strange punctuation choices, and other signals that say, no really, a person was here.

I do not think that pressure stays confined to blogs. It spills into job applications, student essays, support emails, indie websites, and maybe even product copy. Once authenticity becomes legible as style, style turns into camouflage. The whole idea of a natural voice starts feeling unstable.

That instability is what makes the essay more than a neat literary bit. People already trade folk wisdom about what “sounds AI” in classrooms, hiring loops, and comment threads. Some of those instincts are useful. A lot of them are lazy. Either way, once they spread, they start shaping what honest human writing is allowed to sound like in public.

That is why the CSS framing works. CSS changes presentation without touching the underlying content. The essay keeps asking whether our new anti-AI habits are actually changing what we mean, or just changing how we decorate it so the machine suspicion detector leaves us alone. The answer, unfortunately, is probably both.

Hacker News was arguing about art, but the web problem is bigger

The HN thread kept circling the question of whether the piece is good art or self-indulgent art. I think that debate misses the more interesting part. Even if you hate the tone, the essay still captures something real. A lot of people now have a mental blacklist of tells they associate with machine prose. Some of those tells are fair. Some are cargo cult. All of them change behavior.

And once behavior changes, the internet changes with it. Maybe more sites go intentionally rough. Maybe more writing becomes jagged on purpose. Maybe polished sincerity becomes harder to trust. None of that is a technical problem in the narrow sense, but it is still the kind of thing technical culture creates and then has to live inside.

My take

I think the essay lands because it refuses the easy ending. It does not conclude that one quirky CSS trick can save human writing. It shows how flimsy those surface markers are, then twists the knife by ending in exactly the kind of assistant tone it was trying to escape.

That is the part worth keeping. The future of human writing online is not going to be secured by lowercase text or strategically ugly phrasing. If anything, those tricks will get absorbed too. What survives is not a set of typography hacks. It is point of view, risk, specificity, and the kind of awkward conviction that still feels expensive to fake well. That is harder than changing CSS, which is probably why the essay works. The web is going to keep rewarding surface signals for a while, but I do not think surface is where human writing wins in the end.

Sources and follow-up reading

• Original essay by Will Keleher
• Hacker News discussion