Safety Researcher / Red Teaming @meta MSL. ex-Scale. Mostly harmless. NYC.
Joined September 2013
- Tweets9K
- Following2K
- Followers983
- Likes143K
Pinned Tweet
I'm the lead author on this - by far the coolest thing I've ever worked on (that I'm allowed to talk about)
🧵 1/N) Excited to share our recent work at @scale_AI, "Jailbreaking to Jailbreak (J2)".😈 We present a novel LLM-as-red-teamer approach in which a human jailbreaks a refusal-trained LLM to make it willing to jailbreak itself or other LLMs. We refer to this process as constructing a J2 attacker. They are so good at attacking 🫨.
Our approach is straightforward without complex wiring between LLMs, serving as an improved baseline in this domain. We see that the current LLMs have gained non trivial capability at jailbreaking.
🔗 Demo & Paper: scale.com/research/j2
They’re talking about 10 lines of code like the dev lead was accused of sex crimes
The reveal that Big Walk used AI for "under 0.1%" of its code has caused a big blowup among fans, developers, and game award judges.
Where do people draw the line on genAI in gaming, and how did a 2026 GOTY indie darling become ground zero for the debate?
We broke down the messy discussion:
kotaku.com/big-walk-ai-discl…
3
58
Last year there was a vibe shift in Dec-Jan when people outside the bubble tried out new AI tools on holiday - very interested to see if we get the same this year
2
33
Jeremy Kritz retweeted
My AI agents found an unmapped Maya site in Mexico, using LiDAR data from a NASA dataset.
72 hours from a random thought on vacation to a reply from Mexico's national archaeology institute.
What a time to be alive…
Jeremy Kritz retweeted
Steve Selvin, the creator of the Monty Hall paradox, has passed away at age 85.
If there aren’t three coffins at his funeral, he’s missed a huge opportunity.
Jeremy Kritz retweeted
Nothing like waking up to dozens of congrats for taking a job you did not take! While I loved my time at Meta, no, I am not going back! I’m still dealing with severe health issues and focusing all my energy on finding cures for myself and millions of patients like me. And while it’s been a hard time personally, it has also been the most meaningful mission I have ever worked on.
Nope, fake news! Sadly still dealing with health issues and focused on finding cures through @ChronicleBioAI.
Jeremy Kritz retweeted
I asked Fable 5.2 to create a painting as indistinguishable as possible from a photograph. this is roughly 700,000 simulated brush strokes
Jeremy Kritz retweeted
The #Eagles used a 1st-round pick on DT Jordan Davis. Had him for 4 1⁄2 years. Then traded him for a 1st- and 2nd-round pick.
The #Eagles used a 2nd-round pick on C Cam Jurgens. Had him for 4 1⁄2 years. Then traded him for a 2nd-round pick.
The #Eagles traded a 1st-round pick for AJ Brown. Had him for four years. Then traded him for a 1st-round pick.
BREAKING: The #Jaguars are trading a 1st-round pick (2027) and a 2nd-round pick (2029) to the #Eagles for DT Jordan Davis, per @AdamSchefter.
A bombshell trade as JAX is going for it.
Jeremy Kritz retweeted
For agents that perform hundreds of small tasks, efficiency matters at every step. Muse Ember is built to deliver more intelligence for the cost.
On the @ArtificialAnlys Intelligence Index, it achieves performance comparable to Claude Haiku 5.5 at roughly one-third of the cost per task, while scoring higher than GPT-6 Luna across all of Luna's reasoning settings.
Shrapnel was invented by Henry Shrapnel en.wikipedia.org/wiki/Henry_…
Gf and i have this bit that we'll just say stuff like "oh you know the inventor of muffins, John Muffin", and she called it our bit so i told her thats not ours, John Bit invented that
1
1
158
I’ve seen a lot of bad opinions on this site but “mosquito nets are worse than white supremacy” might take the top spot. I’m stunned. (Also if anyone still needs to fill out their @JeffDean apology form…)
unavailable
6
306
Jeremy Kritz retweeted
Darline Graham ran the worst campaign in the history of human democracy, even going back to fucking Ancient Greece.
