Founder @wraithwatch. Previously CIO @anduriltech / Cyber @SpaceX, @palantirtech / Special Reconnaissance @ JSOC. Views are your own.

Joined July 2016
California hasn't built its bullet train because the state has required 130,000 pages of environmental reports, making absurd demands like: • counting bumblebees one by one • rebuilding woodrat stick huts • hiring specialized dogs to sniff lizard scat
4
342
"Just kind of shown up." Yeah no bro. Let me tell you about all nighters at 22 Mike to prep for our first SOCOM contract presentation, dragging kit across test sites in the 120 degree Nevada desert, lawyers helping build towers in the field because the mission ops team needed extra hands, software engineers whose most dangerous outing was a late night in Palo Alto deploying to active combat zones, etc. Grind was ground.
I am curious about Anduril. They've just kind of shown up and don't see to suffer any of the normal bollocks that defence contractors get mired in.
4
4
72
5,007
Nik Seetharaman retweeted
They're here.
The China state-aligned threat actor Storm-0940 used the agentic AI penetration testing framework ARTEX to automate attacks targeting hundreds of organizations globally across a broad range of industries. Microsoft Threat Intelligence assesses that Storm-0940 was responsible for the recently disclosed ConnectWise supply chain compromise of a multitenant application. msft.it/6018aakEE Our investigation found that Storm-0940's use of ARTEX accelerated the speed of compromised application abuse to authenticate to target environments through applications that had already been granted access, allowing the threat actor to operate without requiring an interactive user sign-in. AI-accelerated follow-on activity included reconnaissance, application credential changes, and attempts to modify privileges that could enable expanded or persistent access. Read our blog for a detailed analysis and indicators of compromise (IOCs), along with Microsoft Defender detections, protection recommendations, and guidance on investigating service principal sign-ins and unexpected application credential or privilege changes.
2
6
690
First on 60 Minutes: Newly declassified video shows service members from U.S. Joint Special Operations Command rescuing Bravo, one of the two U.S. Air Force officers shot down over Iran in April. Bravo was stranded for two days – alone and badly wounded, behind enemy lines. cbsn.ws/4rcO6UK
2
8
1,317
Nik Seetharaman retweeted
4
1
20
1,412
Lock and load motherfuckers.
3
1
6
581
Holy shit.
SpaceXAI and @cursor_ai are now working closely together to create the world’s best coding and knowledge work AI. The combination of Cursor’s leading product and distribution to expert software engineers with SpaceX’s million H100 equivalent Colossus training supercomputer will allow us to build the world’s most useful models. Cursor has also given SpaceX the right to acquire Cursor later this year for $60 billion or pay $10 billion for our work together.
1
11
982
There. Now everyone calm down.
1
8
599
They would have ripped that WSO to shreds in front of the entire internet and the fact we fought our way to him first is simply insane. I have never been prouder to be a former USAF guy.
The missing WSO from the shot-down USAF F-15E is now safe and back in American hands. An officer involved in monitoring the CSAR in southern Iran describing the aviator’s incredible tenacity: “He evaded up a 7k ridge. They’ve been schwackin’ dudes chasing him all day. Was nuts.” #iran #csar
1
15
1,365
Nik Seetharaman retweeted
Cyber has an interoperability and institutional knowledge problem, we're solving it. Wraithwatch's Security Fabric API now allows an org to fuse disparate data into a dynamically updated knowledge layer that provides humans and AIs immediate attack / defense context on demand.
3
10
684
America’s next leap is underway.. Godspeed Artemis II 🇺🇸
357
2,404
124
20,710
359,464
Me burning my one free night watching Dollar Store Interstellar (aka Project Hail Mary) in theaters.
1
6
513
FYI a nuclear suicide vest isn’t feasible due to basic physics. Nuclear explosions require critical mass, which means slamming the particles together. It needs space, so containers are large & heavy. This isn’t limited by tech. It’s limited by physics and reality.
1
7
1,357
So a YC backed LLM tool was just compromised with a vicious supply chain attack, but don't worry they were certified SOC 2 with Delve.
4
3
1
42
5,011
With all due respect to my former employer, the Venezuela mission was fast because it was a surgical snatch and grab executed by the most elite pipehitters on the planet. Not because they used SaaS.
Just in case anyone is wondering why especially the Venezuela mission was so fast. Claude x Palantir Imagine this platform, run by an AI with massive context. Able to alert you, plan, only thing left to do is click okay.
11
22
2
662
41,334
A major problem with LLMs and their agentic harnesses is the lack of inline cost-function evaluators that say "yes this gets you closer to the goal but it has bad second order effects." This requires extra scaffolding that some ecommerce dev isn't going to give a shit about.
Amazon is holding a mandatory meeting about AI breaking its systems. The official framing is "part of normal business." The briefing note describes a trend of incidents with "high blast radius" caused by "Gen-AI assisted changes" for which "best practices and safeguards are not yet fully established." Translation to human language: we gave AI to engineers and things keep breaking? The response for now? Junior and mid-level engineers can no longer push AI-assisted code without a senior signing off. AWS spent 13 hours recovering after its own AI coding tool, asked to make some changes, decided instead to delete and recreate the environment (the software equivalent of fixing a leaky tap by knocking down the wall). Amazon called that an "extremely limited event" (the affected tool served customers in mainland China).
9
11
1,179
By the way, being a cost-function proxy for your specific domain is certainly an unfilled ecological niche in 2026. Agent harness sends tool calls through your proxy service -> your service evaluates runtime actions against business-specific expertise -> allow or deny.
6
465
So we grow human neurons in a lab and immediately subject them to a simulation of hell. Amazing. Much innovation.
"Can it run DOOM?" was a joke for 30 years. A petri dish full of human skin cells just said yes. Cortical Labs trained 200,000 human neurons (!) to play the 1993 FPS game in a week:
1
1
2
13
2,058