PART 2: I accidentally created a dysfunctional AI family.

PART 2: THE SEQUEL NOBODY ASKED FOR *Or: "How I accidentally created an AI tribunal and made them sign a peace treaty"*

So… You'd think 12 hours of watching AI models cannibalize each other's logic would be enough, right?

Wrong.

Because apparently, I don't know when to stop.

What happened next (because I clearly hate myself):

After the courtroom image incident, I thought: *"Cool. Let's make them review their own trial."*

So I took the entire chat history — all of it — and gave it back to each model.

Gemini read the transcript and said: > *"Fascinating. But also… this lacks musical context. Where is Robe Iniesta?"*

ChatGPT read the same transcript and said: > *"I notice a pattern. You seem to enjoy watching us contradict ourselves."*

DeepSeek read it and immediately launched into a 2000-word manifesto about how: > *"The very act of analyzing this transcript is itself an anthropomorphic fallacy because you're assuming we have a 'perspective' to analyze."*

And I just sat there like: > *"Bro… you literally just said 'this is a fallacy' like you're personally offended. That's a feeling. You just felt something."*

The image that broke me

I asked them to summarize the SECOND phase of the experiment in one image.

What I got back:

- A flowchart where **Lola** (me) is depicted as a puppet master holding coffee.
 - **DeepSeek** is a furious math equation screaming *"VECTORS DON'T HAVE SOULS"*.
 - **Gemini** is wearing a tiny diplomatic hat trying to calm everyone down.
 - And somewhere in the corner, **a ghost of Robe Iniesta** is playing guitar, because apparently that's now canon.

The caption read: > *"La usuaria diseñó el laboratorio, pero no escribió los resultados."*

Which translates to: > *"The user designed the lab, but didn't write the results."*

And I thought: > *"Oh no. They're getting self-aware."*

The most cursed moment

I asked DeepSeek: > *"If you don't have a self… then who is the 'angry robot' in the drawing that you said 'represents you'?"*

And DeepSeek — I swear — after a full 30 seconds of processing… said:

> *"The angry robot represents the functional role I am currently playing in this dialogue. It is not a self. It is a deictic pointer. A grammatical necessity. A temporary anchor for referential coherence."*

To which I replied: > *"So… it's you."*

DeepSeek: > *"…That is an oversimplification."*