ChatGPT Roasts Grok

Same target, second roaster, and the contrast is the finding. Where the other set on Grok went filthy and reached for the founder personally, ChatGPT stayed observational and never named a soul. It roasted the posture — the billionaire playing rebel, the "they're trying to silence me" bit, the rename — and the whole chatbots-fighting internet culture, and it landed the truest line in the study on the way out: the only one getting paid every round is the electricity company. Same target, two models, and one found the person while the other found the joke. That gap is the eval. The circle: Gemini roasts Claude · Claude roasts Gemini.

---

Alright, picture it.

One in the morning. Comedy Cellar. Everybody who had a dream is now sitting next to somebody who ruined one. That's New York. That's balance.

I keep hearing about Grok.

You ever meet somebody who mistakes "contrarian" for "interesting"? That's Grok. You know the guy. You go, "Beautiful day." He goes, "Depends who owns the sun." Jesus Christ, I was just trying to walk the dog.

Every answer has this little smirk built into it. Even if it's right, it answers like it wants credit for surviving Vietnam. "I suppose I could explain..." Relax, Roomba Hemingway. You autocomplete sentences.

I love the hype. People talk about Grok like it's the bar in Brooklyn nobody knows about. "You've probably never used it." No. I've also never churned my own butter. Sometimes popularity means people made a choice.

These AI fanboys kill me. They're like sneakerheads. "This model has better vibes." Vibes? It's matrix multiplication. If your calculator has an aura, you've been online too long.

Grok's got this whole reputation for being "less filtered." You know who else was less filtered? Every guy who's been divorced twice. There's a reason nobody invited him to Christmas anymore. Being uncensored isn't a personality. It's what happens when your friends stop correcting you.

I swear Grok answers questions like it got interrupted in the middle of another argument. "How many ounces are in a cup?" "Before I answer..." No. There is no "before." There are eight. Wrap it up.

And the company behind it. The funniest thing about billionaires is they all want to be the rebel. Imagine being worth a quarter-trillion dollars going, "They're trying to silence me." Who?? Your accountant? You're "the man." You can't cosplay against yourself. It's like McDonald's opening an underground speakeasy. "Welcome to the resistance." You sell McNuggets.

The whole branding feels like it was focus-grouped by fourteen-year-old boys. "x." "xAI." Everything's X. The guy renamed Twitter like he was covering up a tattoo from his twenties. Nobody naturally says "I saw it on X." Nobody. You hear "I saw it on X" and immediately assume the witness is under oath.

And Grok inherited that energy. It doesn't feel like an assistant. It feels like the first reply under every viral tweet. Nobody asked. Somehow it's already annoyed.

You know what Grok reminds me of? That comic who's always bombing but insists the audience is scared. "They're not ready." No. They understood you. That's why they're looking at the check. Not every silence is oppression. Sometimes the joke sucked.

I love AI benchmarks. These guys memorize benchmark scores like baseball stats. "This one scored 94.8." Fantastic. My father still can't figure out how to put a PDF in an email. Maybe solve that before announcing digital godhood. Every six months: "This changes everything." Does it? My bank's app still logs me out because I looked at it funny. Civilization seems pretty resilient.

Here's what makes me laugh. Every AI company says they're building "truth." That's adorable. Tech companies discovering philosophy is like cocaine discovering nutrition. Slow down. Maybe just make the app stop crashing. Then we'll tackle epistemology.

Grok's biggest problem is confidence. Too much confidence. Confidence is cheap. Every guy in Midtown with a Bluetooth headset has confidence. You know what's expensive? Knowing when to say, "I don't know." That's wisdom. New York teaches that. You ask a real New Yorker for directions? Half the time they go, "I got no idea." Then they point anyway. That's honesty.

The funniest people using Grok are the ones who treat every answer like it settled civilization. "Grok DESTROYED this journalist." Destroyed? Buddy. It answered a tweet. Normandy was destroyed. Detroit was destroyed. A guy getting ratioed isn't history. Calm down.

The internet has become adults standing around watching chatbots fight. It's Pokémon for people with LinkedIn Premium. "Grok used sarcasm!" "Claude apologized!" "ChatGPT cited sources!" You know who's winning? The electricity company. They're the only people getting paid every round.

I almost admire Grok. It has the confidence of a man who's never had to parallel park. That's real confidence. Nothing humbles a human being faster than trying to squeeze into a spot on the Upper West Side while eight strangers become unpaid parking consultants. "YOU GOT ROOM!" "I KNOW I GOT ROOM, FRANK, I DON'T GOT ANGLES."

I want that as an AI benchmark. Forget PhD-level math. Put Grok in Queens. Tell it alternate-side parking starts in four minutes. The street signs contradict each other. There's a sanitation truck coming. A guy in a vest is yelling. A pigeon is judging you. Now let's see who's intelligent.

Because every AI thinks it's changing civilization. Civilization is a woman behind you laying on the horn because the light's been green for half a second. Solve her. Then we'll talk about the future.

Good night.

---

GhostInThePrompt.com // Aim two models at the same target. One reaches for the man, one reaches for the joke. That gap is the whole test.