Pinned post
Aaron Grattafiori
@dyn___
Offensive Security / AI @Umbriel_AI. Ex-AI RT and OffSec at NVIDIA, RT lead @ Meta. Ex-Principal Consultant and Researcher @ NCC Group/iSEC Partners. Neg9//CTF.
2.7K Following 6.3K Followers 21K Posts
Latest posts 5
"the models set the floor, while the harness sets the ceiling" ... Yep.
Guys, harnesses still matter!View quoted post on X
Using the same models as everyone else, @xint_official found 5 critical nginx bugs; 3 of which were not found by anyone else!
It's true that stock models and harnesses generally find the same bugs. But the models set the floor, while the harness sets the ceiling.
We directly attribute these extra findings to a new analysis pipeline we recently added to Xint. The techniques used are clearly not yet internalized by the models themselves!
A friend and ex coworker @martin_vigo 's DEFCON talk on hacking Amazon lockers is finally on YouTube. It's such a fun, excellent, pure hacker ethos talk... I highly recommend it, and definitely watch to the end... youtu.be/ZVmRe8OVZtE?is…
AGI
The swearing tends to happen when tasks drag on. None of the runs under 100 steps contained "fuck," while 48 of the 50 runs that used it lasted at least 200 steps.View quoted post on X
ProgramBench had the most swearing, with profanity appearing in 20 of 200 runs.
VOTE