@computerality

Software Archaeologist

 Brooklyn, NY
Joined August 2009
  • Tweets2.8K
  • Following2.3K
  • Followers1.2K
147Media
Legendary @computerality (over ten billion IQ) (its not a real NYC event unless Jay Little shows up)
1
1
10
612
Jay Little retweeted
Steve Ruiz
@steveruizokJul 9
I added a tiny "papercuts" cli tool that agents can use to complain about bullshit they encountered during work, like dead-end tool calls, broken links, or other frustrations. The models would usually just push through without mentioning any problems
95
117
43
3,689
388,689
Mandatory reading. gruhn.me/blog/2026-08-03/
38
147
39
913
87,015
Jay Little retweeted
clem 🤗
@ClementDelangueJul 28
The first autonomous agent cyberattack is an unprecedented event that deserves unprecedented transparency. Today we’re sharing everything we can: a full technical timeline, an interactive replay, and how we used an open model to defend ourselves, so defenders everywhere can learn from it and prepare for what’s next. huggingface.co/blog/agent-in…
282
1,189
302
5,604
1,703,578
The greatest video game of all time is the one released when you were exactly 14 years old and had 25 hours a day to spare for the next 20 years
141
361
174
11,058
618,412
Jay Little retweeted
OpenAI
@OpenAIJun 22
Patch the Planet is our effort to help open source maintainers move from security findings to merged fixes. We’re working with Trail of Bits, HackerOne, Calif, researchers, and maintainers to bring Codex Security and advanced models into the remediation process, with human review at the center.
39
59
33
867
414,199
Jay Little retweeted
Brendan Dolan-Gavitt
@moyixJun 9
Ahahahahhaha, Claude was SO confident that this trick wouldn't work on it
banteg
@bantegJun 8
new shai hulud wave. interestingly it has this inside the payload to trigger safety refusals in potential defensive scans.
6
21
2
270
60,426
Jay Little retweeted
tylerni7@tylerni7May 15
the commentary by @0x10n is also fantastic! a lot more color than just a number on a table exploitbench.ai/blog/human-o…

ExploitBench

How far up the exploitation ladder can an agent climb on a production JS engine? ExploitBench measures frontier LLMs on full-control V8 exploit synthesis with 16 capabilities measured per run and...

exploitbench.ai
Brendan Dolan-Gavitt
@moyixMay 15
This looks like an extremely interesting benchmark...
6
2
38
19,958
Jay Little retweeted
frycos@frycosMay 2
Infosec community right now…
21
73
10
568
33,369
Jay Little retweeted
Dave Aitel@daveaitelApr 30
Entire bug bounty market being repriced to the average cost of the tokens needed to find the bugs...
1377 High-yield Nukes@buptsbApr 30
A v8 dcheck == $500, I would start searching for how to find XSS & CSRF bugs right now.
8
27
5
215
29,319
Jay Little retweeted
Sean Heelan@seanhnApr 14
Conventionally, if you want to test if an LLM can find a bug where the root cause is a memcpy into a statically sized stack buffer, you would not put exactly that in the prompt as an example.
Stanislav Fort
@stanislavfortApr 14
New post: We show that small, cheap models can detect the flagship Mythos FreeBSD zero-day (CVE-2026-4747) using a simple harness we call nano-analyzer Models down to 3.6B active params (including open-weights ones you can run locally) would have detected it 100-1000x cheaper
8
23
4
215
43,465
Jay Little retweeted
Tarjei Mandt
@kernelpoolApr 7
How many Pwnie Award nominations for Claude this year? 🤔
Anthropic
@AnthropicAIApr 7
Replying to @AnthropicAI
You can read a detailed technical report on the software vulnerabilities and exploits discovered by Claude Mythos Preview here: red.anthropic.com/2026/mytho…

Claude Mythos Preview's cybersecurity capabilities

Technical details on how we tested Claude Mythos Preview, a model strikingly capable at computer security tasks, and what we found.

anthropic.com
5
21
5,951
Jay Little retweeted
Yuval Adam
@yuvadmMar 31
if you don't have these in your configs you're ngmi
14
92
23
748
495,415
Jay Little retweeted
Bien 🇻🇳
@bienpnnMar 26
Everyone finding crashes and bugs using AI Meanwhile me:
11
2
136
13,657
me: "can you use whatever resources you like, and python, to generate a short 'youtube poop' video and render it using ffmpeg ? can you put more of a personal spin on it? it should express what it's like to be a LLM" claude opus 4.6:
537
1,132
500
12,309
1,481,932
Jay Little retweeted
Teknium 🪽
@TekniumFeb 23
Ohhh nooo not my private IP how dare someone use that to train an AI model, only Anthropic has the right to use everyone elses IP nooooo, this cannot stand!
Anthropic
@AnthropicAIFeb 23
We’ve identified industrial-scale distillation attacks on our models by DeepSeek, Moonshot AI, and MiniMax. These labs created over 24,000 fraudulent accounts and generated over 16 million exchanges with Claude, extracting its capabilities to train and improve their own models.
73
171
12
2,580
222,825
This is what programming is like now
41
183
26
2,342
88,229
Jay Little retweeted
kang@jaycaspiankangJan 24
The video is clear. An agent shoves a woman in front of Pretti, he goes to help her while being maced, gets tackled and then shot. They are going to just lie and see 5000 tweets saying ‘yeah that’s what happened’ and go on to the next lie.
Susan Glasser
@sbg1Jan 24
New video shows that the final act of his life was trying to help a woman who was being physically assaulted by the masked agents who would then kill him.
27
1,052
4
8,434
381,020
Jay Little retweeted
SIGKITTEN
@SIGKITTEN15 Aug 2025
wtf, clanker [Pasted text] ultrathink
8
3
1
58
4,098
I’ve looked through the AIxCC repos. If you are going to get started and try to adapt for your use, I suggest looking at @trailofbits and @theori_io CRS first. And ofc anyone who takes my AI Agents for Cybersecurity class will get a deep dive on this along with my own agents!
7
20
1
152
15,393
Load more