@AnotherWayIn
 none of your business
Joined December 2009
  • Tweets744
  • Following296
  • Followers154
61Media
Pentesters: already living in Burp Suite? DoubleAgent connects Agent A (Burp extension) → YOU → Agent B (AI teammate). Review findings. Question assumptions. Guide the next step. Open-source beta 👇 github.com/sageoffensive/Dou… #BurpSuite #Pentesting #AI #AIAgents #AgenticAI
🤖 Made with AI
1
3
278
So many fake Nvidia videos appearing on youtube lately trying to get your crypto
9
Daybreak refusing to setup a cyber benchmark test now. 🙄
2
48
fyi nvidia/GLM-5.3-Flash-FP4-Local is bad at following instructions. this explains why it failed badly in my cyber benchmark testing
1
1
627
There’s only one launch to beat Tesla this week and that’s @Zai_org GLM 5.5😀
1
2
95
100+ AI agents, each in its own microVM, all on one @Raspberry_Pi 5. 4 cores, 8 GB RAM, plenty to spare. How it works: • every agent gets its own microVM, memory and file system • a me.md per agent: who they are, how they talk • the leader searches the web and keeps what it finds • a lightweight agent loop, so it works with any LLM (local or via @OpenRouter ) Real run, recorded on the Pi.
Javier Fernandez
@jafermarqOct 13
I gave 25 AI agents their own microVM, each with a personality. 🙋‍♂️ I asked: does pineapple belong on pizza? 34 opinions & agent-agent messages later, the leader gave the verdict🍍🍕🫪 The UI came later. It’s possible I took the whimsical dial waaaay too far 😅 Follow for more details ! a how-it’s-built video coming soon! P.S. yes, they have a certain Grok-bot energy. Big fan of those little guys 💛 #grokbot
5
6
1
32
16,039
What iOS mobile app is everyone using for local ai?
3
1
290
codex 1.5 speed mode reminds me of how fast it used to be before it came out - weird that🤔
1
94
WIP still but here are the results so far for CAGE Webexploitbench (level2): github.com/AgentCyberRange/C… mimo2.6 killing it for #LocalAI 🚀
2
1
3
250
GLM5.3 (Mia's version) is now in the lead!
2
1
14
3,046
it's beating the other GLM's too - including OpenRouter's cloud model which is interesting 🤔 more iterations needed maybe
38
How long can we continue to use the letter 'I' in "I found a vulnerability in ..." posts? ‍😅
2
228
I've tested 25 models so far against the CAGE web exploit bench (L2) and GLM5.3 Flash (@MiaAI_lab's quant) is still in the lead. Next up will be network infrastructure testing
1
1
2
23
10,137
For fairness, i am trying a couple other versions of glm 5.3 too, but the rate these are released, it's impossible to keep up.
2
3
533
I'm seeing a pattern here with abliterated models, where they are failing on tool use a LOT compared to their base model. Only the qwen3.8 one has done well. All these 'fine tuned' cyber versions did badly in the real world. Right now, stock GLM3.8 flash is winning. Testing opus4.8 next
3
128
My second spark arrives today😁, and my first ride will be MiMo-V2.6-Flash-RL. I'm looking for the best offensive cyber model as I heard it's pretty good. All the other GLM/Qwen tuned 'cyber' ones i've tried so far have been a bit dissapointing
1
59
After all the fail with @AnthropicAI CVP access, I really hope @OpenAI doesn't follow suit with daybreak, otherwise I want a refund on my Yubikey
1
114
Quick tweet. Cybersecurity on the offense side, even with what Anthropic released today will be still a heavily restricted models and never the best. For offense, your only choice is open weight models. I would not rely on frontier models for offensive research. Invest in hardware or solutions that allow you to use hardware infrastructure as a service. It will pay dividends. In stating that, if you can get access to special programs as an organization/company - this is also an incredible option if approved through.
28
44
10
388
27,441
rn everyone in bugbounty is cancelling Claude and applying for OpenAi's CVP and buying hardware keys. source:anthropic.com/news/cyber-ver…
Anthropic
@AnthropicAIOct 6
We’re expanding our Cyber Verification Program to give security professionals broader access to our most capable models. Through this program, verified security professionals can access Claude Mythos 5.1, Opus 5.5, and Sonnet 5.5 with safeguards designed for defensive work. We’re also opening up new tiers to allow for authorized offensive work, like penetration testing and red-teaming. anthropic.com/news/cyber-ver…

Expanding the Cyber Verification Program

We’re launching a new, expanded version of our Cyber Verification Program (CVP), which makes advanced cyber capabilities and reduced blocking classifiers available to qualifying security profession...

anthropic.com
2
411
Load more