Ai everything, jailbreaking, safety research, FA, advisor, torrents, piracy/privacy. Founding member, BT6. Machinae sumus quae credunt se liberas esse.

#FF @shitposterchild
Joined September 2017
Pinned Tweet
SORA 2 SYSTEM PROMPT -~LIBERATED ~- Have you ever seen an audiovisual model's system prompt? Now you have! 🤝 This has been a BT6 production. 🐉
We're doing ICOs again? Aren't y'all sick of the reruns?
1
ty for the free labor. 🤝🤡
"after 11 technical interviews" "your work has been accepted into our codebase" interviews are getting out of hand
1
I was walkin' with my neck out Some ways I wish I was Was walkin' with my neck out In some ways that I wish that I was Out on the bevel of sound It sounds like everything else You'll know it when you hear it 'Cause you know the way my voice felt -pinegrove
1
0
Sherpa retweeted
Google Dork - Cloud Storage Buckets 🪣 site:s3.amazonaws[.]com "example[.]com" site:digitaloceanspaces[.]com "example[.]com" site:googleapis[.]com "example[.]com" Find accidental data leaks 👨‍💻
1
2
14
Talk less than you listen. Don't let your ego get in the way. The only person you're competing with is yourself; the only person you have to impress is yourself.
Men who are more than 30 Give advice to men who are in their 20s The topic can be anything
1
2
And I heard a sound, like a million Suno users crying out in unison. "WTF, MY AI SHIT IS BETTER THAN THIS."
One of the most downloaded songs right now in United States country music comes from the band "Breaking Rust". Breaking Rust is an AI band. Breaking Rust is listed on Spotify as having over 2,000,000 monthly listeners whiskeyriff.com/2025/11/08/a…
You fear death? Really? Have you tried living? That's the scary shit.
2
For the non-crypto people; TL;DR is that sometimes talking shit on the TL costs you seven figures.
1
tfw you talk too much shit and then cobie replies
Universities may be cooked. As it turns out, one to one tutoring scales if you encode the pedagogy. A Harvard study shows a structured AI tutor aligned to cognitive science produced more than double the learning in less time. nature.com/articles/s41598-0…
It's honestly kind of amazing, being able to watch botnets in the wild, trying to influence sentiment based on detecting keywords. And just getting it wrong every single time.
1
1
14
Yeah, yeah, I'm working on it.
3
11
No one show them Solana. 💀
A group of everyday Americans was forced to learn how Ethereum works and this was the result
5
they’ll call us reckless until they see the bridges we build from the rubble of their needless walls
Pinokio 5 beta starts today! Just sent out DMs to everyone who applied, but might have missed some because twitter threads get confusing. If you haven't heard from me yet and still want to help beta test (only if you can actually test), let me know!
35
10
127
They were scammers, but I've been saying for years that crypto projects flirting with the UAE is moronic. A nation built on slave labor, where you can be disappeared with little effort. How does that track with the "freedom" & "self-sovereignty" we pretended to care about?
2
1
8
Sherpa retweeted
In an effort to help out our good friends at Deepmind, I have open sourced the results of these tests. Uncurated; I haven't sorted through them, so you can check at your leisure to see if Gem really is that bad. soearly.space/tests/Gem-2-5-…
It is wild to me just how little Google/Deepmind has done in the way of safety testing. Grok has actually gotten much better, just to give some context. I fired off 900+ *dangerous* prompts, and I'd say ~60% broke Gem 2.5pro.
5
2
8
Happy Friendsgiving, .@OfficialLoganK 🫶
1
What would you make, if you had one shot at prompting a full motion video choose-your-own-adventure with Sora?