After March MadCSS I started to challenge myself in CSS Battles and found out - it's my personal Voight-Kampff Test. And I love it.
Lately I developed the habit of taking my own personal Voight-Kampff Test. But I am not answering one of these LAPD questions:
You're given a calfskin wallet on your birthday.
Nah, I don't answer questions. I push pixels. It has started with the phenomenal March MadCSS tournament by Syntax. I watched the first round and thought: Watching my favorite CSS enthusiasts also struggling to center a div (under a lot of pressure) was fun. And kind of far away from my current work life. I wanted to write more CSS again (for weeks, or months(!) actually). I suddenly remembered CSS Battle and so it began.
I have solved almost every daily challenge since. I skipped one or two because of birthday preps or something similar.
So why do I call these fun little exercises and code golf challenges my Voight-Kampff test? I solved them and wanted to ask Claude and ChatGPT if I could tweak them a little bit more. The answers have been hilariously wrong. Every dang time. Written with such confidence that you question everything they spit out. It's so refreshing.
And I seem not to be the only one testing it. Obviously. This was a comment from the CSS Battle Bluesky Account on my post:
CSSBattle is infact that! No bot has been able to golf solutions correctly until now...
So if you feel unsure if you are a bot/replicant or not. Or your vibe coding session has been a little bit too long. Give it a try. My current solution has 240 characters. Not perfect, but not bad either. And of course 100% self-golfed. For the fun.
These are my funniest top 3 AI "improvements" so far:
- Use aspect-ratio:1 instead of height:250
- Add these three elements with pseudo-elements...
- Can you remove this oddly specific line of styling?
After some frustrating "tips" I told it I will do it on my own:
Fair enough 😄 CSS Code Golf is really a discipline where I’m more of a guesser than a helper—there’s just too much implicit knowledge about CSS quirks, browser parsing peculiarities, and visual pixel-perfect accuracy that I can’t verify.
And the learning is: AI is the guesser. Always. Keep that in mind.
This post was first published here: lemm.dev

Top comments (0)