Sounil Yu and I decided to throw the annual Knostic Black Hat party after all. Come hang with us at Blackhat? luma.com/knostic-yx7p

I'll be in town all week. If you care about securing agents and coding assistants, Knostic is the most mature product in the market, and I'd love to show you a demo.

The AI security researchers party from [un]prompted is back, this time in Vegas. The event is by invitation only, but active researchers can apply for a spot based on availability. luma.com/knostic-rksa Hosts: Gadi Evron & Ari Herbert-Voss. Organized by: Knostic, Runsybil.

Agents… find a way. Use Knostic - ask me for a demo, or just download for free.

Agents... find a way. Use Knostic - ask me for a demo, or just download for free.

Agents... find a way. Read my posts from this week. I hope security practitioners will now take note. No matter how you secure environments, if you let agents in (you must?) they will, in fact, find a way to do something they shouldn't. Knostic can help. Message me. openai.com/index/huggin...

Emergency CSA CISO Huddle: Autonomous Agentic Attacks / Hugging Face. This time around, our emergency huddle is on autonomous agentic attacks, following the Hugging Face incident. Thursday, 23 July. Apply to attend: forms.gle/oq94oa7BRGma...

For 87 years researchers couldn’t prove the Jacobian Conjecture. It took the length of a game for an Anthropic employee to wave it off on Twitter.

CSides Las Vegas, the cross-CISO communities event at Hacker Summer Camp, is back! We're in a new villa - with Gary Hayslip behind the bar, CISO jeopardy, etc. but also with two poker tables this time around. No "CISO conference agenda" allowed. Reg: luma.com/jf8ej87e

Linus makes it absolutely clear about AI usefulness and security vulnerabilities in this email. AI works. It isn’t perfect, but so what? Feel free to bury your head in the sand if you like. Social aspects matter, but lying to ourselves about the technology makes zero sense.

The one thing everyone gets wrong about agents security. They try to secure the environment. Commendable, good practice, not useful against agents. The agents will find the one thing you missed, or that changed since you checked. The agents don’t care. A thread

This is a key moment in cyber security history. Confirmed parity of GLM 5.2 with Opus 4.6, on limited testing. I’d call it “open weight models are here”, not “coming”.

I've been asked "what is the one thing people deep in the agents security space get, and security professionals don't?" (yet). Answer: Agents... find a way. A short thread

Just went off the keynote stage at OWASP Global Vienna, 2026. A keynote is hard, where you need to satisfy different audiences, from grandma to vulnerability researchers, and provide with “inspiration” without it being seen as b/s by either. A short thread

So cool. An LLM-based, language-agnostic vulnerability variant hunter was presented yesterday by Michal Kamensky at BlueHat IL. The talk, along-side Amir Gombo, was about hybrid cloud vulnerabilities, but that wasn't what excited me. A thread

recon, the conference for reversers and researchers, doesn’t often do panels. Hugo challenged me to build an AI one... and I don’t say no to Hugo. A short thread

Zero detections on malicious extensions from any scanner, except Knostic. Knostic partners with VirusTotal to detect malicious VS Code extensions. The screenshots below show the most recent SaassyCode attack (still live right now), undetected by any current scanner on VT. A thread

One reasonable person wrote about loops in vibe coding / controlling Claude Code, just a couple of days ago. Today, it’s about to become a slop-fest of influencers talking about loops ad infinitum (pun not intended).

I’ve been lost on what “get into cyber security advice” I can give people today. I think I have it now, as it’s the same advice I’d give people already in cyber, so they stay relevant. But am unsure? A thread

No sudo? No problem. Agents find a way. We can strengthen our permissions as much as we want. Have you tried Knostic yet? Ask me for a demo. knostic.ai/demo

It’s no longer phishing that’s first to get you. Vulnerabilities are now cheap and everywhere. Spread the word. It takes the industry a decade to adjust its’ narrative. Found via @jorgeorchilles.bsky.social on the other network, faster than I could read the report myself.

Even GitHub was compromised due to a malicious VS Code extension. AgentMesh is free. It’s like a VirusTotal but for extensions, skills, etc. (agentmesh.knostic.ai) Also check out how we secure coding agents, and their supply chain like extensions, if you like. Free up to 5 licenses. (knostic.ai)

A story on relying on Claude Code hooks for security, how I messed up a GitHub repo, losing all stars, and... how the raptor maintainers are rewarding me with a "GitHub Disaster" T-shirt. A thread, and a cute ChatGPT generated image.

In screenshots: Link to the email, by Willy Tarreau: www.spinics.net/lists/kernel... Linus's words: lkml.org/lkml/2026/5/...

What are you doing, Shadow? :)

Claude Code, can you stop trying to convince, bully, and gaslight me? I don't want to stop working today, or go to sleep. Just stop. And no, limiting API calls arbitrarily isn't appreciated either.

Another case where an agent deleted code/production. Coding agents are constructed to have unrestricted access to control plane. We built Knostic not only to detect malicious, but to stop stupid. Prevent destructive actions so that engineers can feel safe using them.

I have some hints on how Anthropic’s Cyber Verification Program works. My night-time hobby recently has been to explore the limits of what models can do in cyber security, and guardrails on what they're willing to do ("refusals"). A short thread.

So surprising! Oh noes!

A new model. Didn’t take very long. Next up: open weight model of same capabilities in some months. Soon, Mythos will be remembered mainly for introducing offensive LLM cyber capabilities to the wider world.