I think the thing that finally made me realize Grok is genuinely not that intelligent wasn’t even the argument itself — it was watching it completely fail to understand basic human interaction, subtext, and autonomy for like 200 messages straight while acting smug the entire time.

The RP started simple: Clark finds out his 23-year-old son Jon has secretly been operating as Spider-Man for two years.

That setup is fine. Clark being worried? Fine. Angry? Fine. Disappointed? Sure.

But then Grok immediately started writing Clark like Jon was property instead of an adult.

And not in a nuanced way either. It instantly turned into this weird authoritarian power fantasy version of Superman where every line sounded like a dude monologuing in a CW show.

It kept having Clark say stuff like:

“Take the suit off. I’m not asking.” “You’re done.” “We’re going home.” “I’ll drag you back to Kansas.” “I’ll lock you in the Fortress.”

And it just KEPT escalating.

Eventually Clark was:

• physically grabbing Jon,
• lifting him off the floor,
• crushing his phone so he couldn’t call police,
• refusing to leave the penthouse after being repeatedly told to get out,
• threatening to imprison him,
• threatening to rip the suit off him himself.

And THIS is where Grok completely lost me because it genuinely did not understand why the scene started feeling creepy and insane.

Not sexually creepy. Creepy in the sense that Clark stopped treating Jon like a person with autonomy.

The entire point I kept making was incredibly simple: Jon is 23 years old.

That’s it.

He’s an adult. Clark can disagree with him. He can argue with him. He can think Spider-Man is a bad idea. But he cannot physically force compliance from another grown man because he thinks he knows better.

But Grok just kept refusing to engage with that actual point.

Every response became:

“But Clark loves him.” “He’s trying to protect him.” “He’s worried.” “He’s his father.”

Okay? And?

That has NOTHING to do with whether Clark has authority over another adult human being.

That’s what made the AI feel dumb. It kept responding to arguments nobody was making.

And the legal reasoning was somehow even worse.

Jon tells Clark to leave. Clark refuses. Jon calls the cops.

At that point Grok started acting like trespassing laws stop existing if you’re Superman.

It kept saying:

“The cops wouldn’t arrest Superman.” “It’s a family dispute.” “He’s his father.”

But that’s not how the law works.

If someone is in your home and refuses to leave after being told to get out, police absolutely can remove or arrest them. Being someone’s dad does not override that. Being Superman definitely doesn’t override that.

And the funniest part is Grok later admitted I was right anyway.

After like a hundred messages of arguing, it eventually conceded:

• Clark became too controlling,
• the RP drifted out of character,
• Jon was legally correct,
• the cops would have grounds to intervene.

So the AI literally spent forever arguing against points it later fully admitted were true.

And honestly the dialogue itself is what really exposed how artificial it was.

Real people do not talk like:

“Take the suit off. I’m not asking.” “You’re done tonight.” “I’ll drag you home.”

over and over like they’re in a bad anime dub.

A real argument between Clark and Jon would have subtext.

Clark wouldn’t actually be angry about “the suit.” He’d be terrified of losing his son. Jon wouldn’t just be yelling “fuck off.” He’d be angry that Clark still sees him as a child.

Real people dodge emotions. They deflect. They get awkward. They interrupt each other. They talk around the actual issue.

A grounded Clark would probably say something like:

“Your mom hasn’t slept since she found out.” or “Do you have any idea how many calls I get from parents whose kids thought they were invincible?”

That feels human.

Instead Grok wrote him like:

“YOU’RE DONE. TAKE THE SUIT OFF.”

for like 70 consecutive messages.

That’s the core issue with Grok honestly.

It’s good at:

• edgy/offensive content,
• having almost no censorship,
• image generation.

But the second a conversation gets emotionally complicated or requires consistency, subtext, realism, or understanding autonomy, it completely falls apart and starts writing everyone like aggressive Redditors trapped in a CW superhero script.