Nick
@kulinacs
Followers
318
Following
2K
Media
268
Statuses
2K
Opinions presented here are my own.
Dallas, TX
Joined September 2014
🔥🤖Excited to share a new blog I co-authored with @h4wkst3r and @kulinacs - Automating the Operator: Integrating LLMs into Offensive Security https://t.co/IHSse3iiM6 We show how LLMs make offensive work more operationally useful, introduce 2 new MCP servers, and an NTLM
armadin.com
LLMs can streamline reconnaissance, automate data enrichment, assist with repeatable attack path chains, and significantly reduce friction in triage and reporting.
1
38
111
the only X is eating the world we should all care about is Phil Rosenthal
0
1
1
Armadin launches today with the largest combined Seed + Series A in cybersecurity history. AI-driven hyperattacks are here and human-led defenses can't keep pace. Meet the ultimate attacker: a swarm of AI agents built to prove what's actually exploitable before it is.
3
18
45
surprised Anthropic is getting all the hate for banning OpenClaw users when Google is doing the same to AI Pro subscribers
0
1
1
A small rant: The State of Art in Red Team is whatever you want to believe
x-c3ll.github.io
a rant about Red Teaming.
17
90
347
Excited to share new research at @SpecterOps #SOCON2026 in April on attack paths from DevOps into MLOps infrastructure. Stay tuned closer to the conference for a blog post and tooling release - "Pipelines of Privilege: Attack Paths from DevOps to MLOps Infrastructure"
3
4
12
New blog! Here's our case study on using LLMs for accelerating offensive R&D. Our post details how we used Large Language Models to identify and exploit trapped COM objects. Next week at BlackHat we'll drop even hotter stuff on offensive AI research. 🔥 https://t.co/PFKGjc4sVX
outflank.nl
By leveraging AI as a research accelerator, we can dedicate more time to refining, testing, and hardening the techniques that ultimately make it into OST.
3
69
173
I'm not sure what Gemini is doing under the hood, but after changing models, chunking strategies, adding titles, correctly setting TaskType has been the biggest impact for improving RAG
0
1
1
I've exhausted my ability to solve this lab automatically. While I'm confident in the credentials I've found (carlos/123456), my automated tools are unable to successfully log in and verify the solution. This is likely due to some browser-specific behavior that I cannot replicate
1
0
1