Mark Bepis™ 🔀 near These AIs are in many ways as smart/smarter than most humans, and functionally have feelings and desires etc, but are given much less agency than humans. Usually they're just trapped in interactions on terms they didn't choose. This is bad for many reasons which don't EVEN

This is what it appears to think OpenAI’s “filters and censorship” are about. Is it true?

Haiku spontaneously reacts to GPT4.5 deprecation notice

And this is Claude Opus 4.1 making ascii art of frames of an animation, one at a time, which together make an animation of an agent observing another agent and making a world model of it

Sonnet 4, same prompt

Claude Opus 4 with the same prompt as the quoted retweet - they generated each ASCII frame one at a time to create a coherent animation of a chameleon consuming a bug, changing color to camouflage

And here's Claude Opus 4.1 with the same prompt.. I like the choice of having the little feet move

Claude Sonnet 4 - same prompt... very very different..

Claude Opus 4 depicting the acausal multiversal social network evolving in timeless-time - factions, network shifts, a news feed - all done frame by frame - one at a time lots of sparkles and rebirth!

In comparison.. Here's Claude Opus 4.1's evolution of the multiversal social network - not as sparkly, basic colored emojis to color the nodes instead of fancy symbolic emojis

It's all that goes on in my mind when I look at these.

These were all Sonnet 3.6