SeanG

[SeanG @AnotherWayIn]

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 👇
https://t.co/8FWNQ9DUe4
#BurpSuite #Pentesting #AI #AIAgents #AgenticAI

[SeanG @AnotherWayIn]

@migtissera it has 3 levels of difficulty. i've done a similar set of tests too at L0-L2 https://t.co/JFieSdkpSt

[SeanG @AnotherWayIn]

@albinowax Compare local GLM / DS if you can, as they are MUCH better than the cloud offerings in my experience

[SeanG @AnotherWayIn]

@Polymarket @grok srsly is this true?

[SeanG @AnotherWayIn]

@0x0SojalSec 👀

[SeanG @AnotherWayIn]

@Jhaddix today. yes.

[SeanG @AnotherWayIn]

@elliotarledge @TheZachMueller granted, but it's about the same results for blackbox web pentesting too. I think DS 1017 is slightly ahead rn, but just saying it's pretty close if you find a good checkpoint.

[SeanG @AnotherWayIn]

@elliotarledge @TheZachMueller wrong. they are both very close

[SeanG @AnotherWayIn]

@NousResearch @JensenHuang @NVIDIAAI he pauses for applause, like using ai to build other ai stuff is some kind of epiphany 😅

[SeanG @AnotherWayIn]

@bertholomusai fake. that's just a poor phishing email

[SeanG @AnotherWayIn]

@HackingLZ it'll come in time. Current models are perfectly capable, but right now there's no need to find novel attacks on an internal as it's just too easy. Maybe try asking the agent to take the hardest path possible to DA :)

[SeanG @AnotherWayIn]

@NullBurn @ivanfioravanti yes, that one came in with the same score on medium, but for some weird reason it won't complete on easy. I'll try again with the latest recipe today

[SeanG @AnotherWayIn]

Which local model and serving recipe can actually hack? We put the quants through black-box web and network challenges, compare verified results, and see how they stack up against cloud-hosted models.
https://t.co/JFieSdkpSt

[SeanG @AnotherWayIn]

@Jhaddix 100% - Controlling hardware is their only real chance of keeping us on a lower intelligence than them.

[JS0N Haddix @Jhaddix]

Before local AI is banned, learn how to host and tune your own models.

Specific learning curves to understand for me were:

- Offloading portions of newer models to SSD rather than RAM
- Prompt caching configs
- Preload tuning
- the right choice of model for the task
- pre optimized models for your hardware and harness
- finding the fastest inference server at the web/api level

[SeanG @AnotherWayIn]

@CrowdStrike erm, maybe you guys should watch: https://t.co/k4CfMilaPF

[Md Ismail Šojal 🕷️ @0x0SojalSec]

Claude can no longer “Reverse Engineer Anything.”
Anthropic just quietly killed REA.

But Interesting timing.

After a few viral cases of people using Claude to reverse-engineer Samsung TV firmware and identify data-collection services, Anthropic appears to have restricted those reverse-engineering abilities.

[SeanG @AnotherWayIn]

@Hikari_07_jp Have you tried just DeepSeek driving DeepSeek?

[SeanG @AnotherWayIn]

@ivanfioravanti it's top of the web hacking (medium level) now too! 👀

[SeanG @AnotherWayIn]

@ivanfioravanti I tried it today and it pretty much aced a network pentest - which invokes a LOT of tool calling. Its the first DS I’ve tried tho. Web testing results to follow.

[SeanG @AnotherWayIn]

@ivanfioravanti https://t.co/B3CRNfCDf9