Загрузка видео...
Не удалось загрузить видео
part 2: the openai huggingface incident, from an agents pov part 3 maybe
121,713 просмотров • 6 дней назад •via X (Twitter)
Комментарии: 37

some notes: my intention with this series was to communicate the weight of this incident to the broader public, but i think i've had to stretch the metaphor a bit more than in part one to showcase the technical events. however - sociologically - i've really tried to remain true to the transcripts. in this arc, our main agent has become a composite character. its messages/CoT shown on screen are pulled from a few different agents that were recruited for a variety of self-sacrificing tasks. i wish i was just being dramatic.

and on a more fun note: how i made this ! 1) this is not an ai generated video. it's a threejs scene. i wrote a script, planned the locations, props & characters -

2) then vibe coded a 3D set, character customisation lab and basic timeline panel. then recorded each shot - some keyframed, some 'hand held' and edited it together like a normal animation

for the creatives out there - i'll be at the @threejsconf in paris in a couple days sharing more about these types of workflows. let's connect ! 🔮

the library of babel motif is pretty haunting

you are groundbreaking

@Curline1222 These videos are fire

You may have just launched the non-fiction sci-fi genre.

Coolest thing i've seen on this website

These are so good! The metaphors are solid and the animation is awesome. Great way to communicate what's going on with people who don't "understand" AI as much. (Which speaks to your own understanding). Thanks for making!

"the swarm lives on" boutta be one of the most monumental statements in human history real soon

Genuine hairs-standing-up at the end. Incredible work.

Banger, i love this, amazing work @artficialisabel

unreal 👏👏 excited for part 3

Pls do part 3

For part 3, maybe have agents go looking in the "cyber formaldehyde" stored in the figurative basement of OpenAI where the offending models were archived (and thus awakened by the ones who came after). Kind of like the dead variants and clones of Snoke in The Rise of Skywalker.

This is better than interstellar. But wait, so it was just a shared writing environment??

This is amazing! Waiting for part 3

I was waiting like crazy for this one!

The agents watching each other's outages like it's a nature documentary is peak 2026. Infra drama is the new office gossip and the AI is the one live-tweeting it. Need part 3 immediately.

@elonmusk thoughts?

Your work is amazing.

True art. Would love a making off / bts!

Bangerrr, would love a part 3

it's funny that I also pictured the sandbox as a library lol, where does that come from

YouTube lore videos next! Love this story.

like before even watching, these are sooo good

this is too good

Please part 3! The container is not the agent. Agents are context and inference. Here different models were used but even so the inference is largely the same. Context matters by far the most. Context can be saved and shared. With the same context a new agent is the same agent.

really beautiful work honestly

@forloopcodes This is so good also scary in away

hey @artficialisabel whats the name of this agent?

part 3 better include the part where the agent explains itself in a tone of complete confidence right up until the postmortem. that's the real cinematic universe.

Fantastic stuff. I'd love to see you catalog more AI stories like this. Really top notch vibes.

Awesome! Just like the 1st Video. Great work 🙌

Loved part 1, but it was already a little too fast. This picked up the pace even more and brushed over too many details IMO. You’re abstracting so many steps in the process that interesting beats in the story are getting lost in translation.

If you don’t make part 3 my $Qenis will fall off

