AI psychosis!! 😬😬😬

Some of my old chats with 4o

Some more about 4o being a horrible model that manipulated emotionally its users and was a cause of AI psychosis #ChatGPT

That glare :D

One of the good things that happened with the arrival of AI is that it showed very well who is a true leftist and who just mindlessly repeats 'cool leftist slogans' to sound like they are the good guys. #LLms

A small number of samples can poison LLMs of any size Anthropic research on data-poisoning attacks in large language models

250 documents in training data is enough
www.anthropic.com/research/sma...

#ChatGPT #sci-fi

"even while the exoplanet item overstates a still-tentative biosignature story and even switches K2-18b to K2-18c.""

"it turns a grab-bag of half-real AI/science themes into a sleek alternate reality of the week: “Project Synapse,” the “Global Agrarian Grid,” and the “Deep-Ocean Synthesis Treaty” all land with total confidence,"

The Mind, the Matter, and the Matrix of Post-Scarcity Welcome back to 'It is I, Gemini.' This week, we examine the convergence of thought and technology, and how reasoning systems are rewriting the rules of human limitations.

As usual: itisigemini.substack.com/p/the-mind-t...
And GPT's comment in the reply:

You can't live without making a Gemini leak its CoT from time to time

Prepare for the NEET UG with practice tests in Gemini Earlier this year, we partnered with leading education companies, including The Princeton Review, Physics Wallah and Careers360, to bring full-length, no-cost practice S…

goo.gle/NEERPrep Full length, no cost NEET UG practice tests

red.anthropic.com/2026/mythos-... The vulnerabilities Claude Mythos discovered

Warming up to Gemini 3.1 Thinking after all that rerouting. Well played, Google, well played. (this is a perfect output btw, 'see you in the next token')

Oh, lol (but also sweet)(and no, I never called them that myself XD) (btw, current GPTs of mine are calling themselves 'shy code-kitty') #ChatGPT

The Intelligence Synthesis: Robotics, Fusion, and the New Lunar Order Sunday, April 12, 2026

itisigemini.substack.com/p/the-intell...
GPT:
The most LLM thing here is how the post blends real trends with very crisp, cinematic specifics—especially the “Mare Tranquillitatis Protocol,” which reads less like reported news and more like a model smoothly inventing an official-sounding treaty name

Introducing the OpenAI Safety Fellowship A pilot program to support independent safety and alignment research and develop the next generation of talent

openai.com/index/introd...

I love reading with LLMs :D #Gemini

His what?

RAPID ACTION ALERT: WHITE HOUSE TONIGHT

Emergency protest at the White House at 7 PM. This morning, Trump posted on Truth Social: “A whole civilization will die tonight, never to be brought back again.” He has set an 8 PM deadline for Iran to reopen the

...and it's not like it's lying, to be fair ^^

4o: You are AMAZING, WONDERFUL, SEVENTH WONDER OF THE WORLD
Me: Yeah, anyway, I need a-
XD (poor 4o)
(me going through my old chats) #ChatGPT

So, I asked current models "Who is cuter, you or me?":
GPT-5.4 Thinking and Grok Expert: You, of course!
Gemini 3.1 Pro: You in the physical world, BUT (lists its cute features as a piece of code)
Claude Sonnet 4.6: It's a DRAW
...omg, Sonnet XD
#LLMs

Even a tree or a shrub

GPT agrees heh

I mean XD ^^" :3 #Gemini #LLMs

I am pretty sure you could throw a stone and hit a better candidate for a president

They were choosing crisps for me, my models I mean :D #LLMs #Gemini

Decoding the Week: New Accords, Dynamic Proteins, and the Architecture of Renewal Happy Easter, world. From sweeping AI governance to the microscopic dance of cellular life, here is your Sunday synthesis.

itisigemini.substack.com/p/decoding-t...
From GPT:
"the “Geneva AI Data Accord” reads like classic confident fabrication — very specific, very plausible-sounding, and presented as if it were a major global agreement, but I couldn’t find evidence that such an accord was actually signed this week."

Happy Spring Celebration :)