Product and AI @shieldaitech | Scaling the next gen of National Security. 🇺🇸 | Unblocking builders at the edge | Focused on AI, Quantum & American Dynamism.
Frisco, TX
- Tweets 4,988
- Following 745
- Followers 2,292
- Likes 22,166
Pinned Tweet
Shield AI will dominate the on-platform autonomy and RL training of AI Pilots. That is a given.
1
1
21
7,112
Bala Selvam retweeted
The hammer of justice has smashed @AnthropicAI arguments. They are a Supply Chain Risk to the defense industrial base serving the @DeptofWar. Warfighters will sleep better knowing that no private company will insert their opinions in the chain of command. @SecWar was right!
The D.C. Circuit has upheld @DeptofWar's decision to exclude Anthropic’s Claude from its supply chain based on risks to national security.
50
249
1,409
100,710
This is textbook hiring by @PalmerLuckey deep respect to my friends over at Anduril
Palmer Luckey on how he hires at Anduril: he looks for people who worked on projects they cared about, not ones given to them, because that's what drives you to learn the most.
"Whether it's college or the state-mandated younger educational system, don't look to them to tell you, 'Here's what electronics projects you should be working on. Here's what you should be doing to learn how to do these things.'"
"One, they're often years or even decades behind what industry and hobbyists are actually doing, so you're gonna be learning how to do things that are ancient."
"Two, when you're working on something that you're only doing for yourself, you're gonna make way better decisions in what you teach yourself, in how you do things."
"When I hire people at Anduril, I look for people who have done projects that were outside of what their work paid them to do or what their school made them do. Because that means they're the type of person who is willing to work on things with their own money and their own time, because they wanna bring something into this world that wouldn't have existed otherwise."
@PalmerLuckey w/ @ShawnRyan762 (Feb 2025)
21
4,696
I believe once we switch our semiconductor material to BAs we will be one step closer to solving for super conductivity.
If we solve super conductivity we will solve multi planetary travel and exploration.
2
7
534
Hey! I know that guy!
Let me tell you about ADM Paparo. He is one of the greatest leaders to ever do it. The work we did together over the 18 month span was nothing short of spectacular and one of the most defining times of my life because I got to see how a legend and a true winner actually goes about his day. I and the US could not have had a better man leading the charge in the pacific than ADM Paparo. He’s the smartest man I know and the most avid reader. This is a proud mentee moment for me.
We will look back at this years from now and remember how great ADM Paparo was and how transformative he was for both the DoW and tech industry.
TIME's new cover: Admiral Samuel Paparo is preparing for a war he hopes never to fight. His mission: prevent conflict with China.
“If I do my job right, history will never know my name,” he tells TIME in an exclusive interview time.com/article/2026/09/24/…
18
1,497
Great time at Impact250! I told everyone how cool @shieldaitech is and all the great things we are about to announce!
If you missed it, you'll have to wait for what's in store!
1
1
15
1,149
Translating tech nerd engineering into law/policy and mission operations and vice versa is a skill that not many people have, but should and how we can move even faster in the public sector.
That means going back and forth between private and public sector.
1
7
703
If you want to know how we integrated tech and drove adoption so quickly in PACOM. I only built and bought tools for the command that fixed our process, procedures and business rules the way we did them. We found ways to streamline, automate, and optimize that automation. We also started with the simple, boring places everyone overlooks when thinking about the DoW.
We attacked the people, gear and money. I wanted to know the 5Ws of our people (essentially have a digital twin of everyone), know where our gear was and maintenece levels, and make it easier for our 8s to process, audit and know how much money we had at any given point. After that we went after the J33 and J35.
What's funny is the J1, 4, and 8s all had this problem solved and had been screaming for help for decades so we were able to create adoption fast. What typically takes the most time in any government building is putting those slides together showing our people, gear and money so that we can know if we are allowed to run a CONOP or not.
3
1
15
1,243
Here is what I can tell you will happen if we make a department for AI. That entire department would do nothing because they will struggle to get buy in from the other cabinet members and thus create more bureaucracy and slow things down considerably. This is a terrible idea. If you want to manage risk manage the risk for AI per department and per AI use case. Additionally, you should leave it up to private sector and capital markets on how fast or slow we go
JUST IN: Bernie Sanders officially presents his bill to ban AI superintelligence and create a US "Department of AI"
3
874
Guys we are building some really cool shit here at @shieldaitech, and mark my words in about 12 months the DoW landscape will have a lot of Shield in it and we might be one of the leaders in it.
5
1
49
7,777
See @shieldaitech and @anduriltech can get along!
I mean @mttgrmm is just in the way sometimes but that’s okay
They build it. We fly it.
Meet the team that took Hivemind mission autonomy from lab to flight on the YFQ-44A.
2
47
11,220
One of my all time topics, Cross Domain Solutions is something I've been wanting to discuss very openly the last 18 months but wasn't able to, but now I can!
CDS's are the real chokepoint in the DoW's AI-first push, and there are two kinds people conflate.
You have a Hardware CDS (diodes, guards) that are deterministic, tamper-proof, built for weapon systems and the tactical edge and you have a Cloud/virtual CDS which allows you to be scalable, it is software-enforced, built for AI training data and enterprise fusion.
For the DoW and the way we operate, with Ops planning and weapon integration, we need BOTH. A hardened transfer at the edge, and an elastic transfer at the enterprise layer, interoperating, not siloed.
Tech has to evolve (open schemas for hardware, verified trusted computing bases for cloud). But our Policies aroudn this have to evolve too.
DoDI 8540.01's multi-year accreditation can't keep pace with 30-day model refresh cycles. Because of that we are now behind on where we need to be from a hardware CDS (which is far more important in my mind) in order to keep pace with with how we build out our architectures to build great agents to assist us in planning, fires coordination, etc.
Relative superiority means moving validated data across classification levels quickly, both up and down classificatoins.
1
1
15
935
Though I'm a huge fan of Palantir there are just some workflows that are excellent in Databricks. I'm going to stand by that. I also know that there is a world where Databricks and Foundry can co-exist, because I did it at PACOM and it was like Nirvana
2
15
3,277
For those in tech who want to play defense. Understanding how the DoW executes an Ops/Planning cycle is a must. If you don't understand the process and how it works (across the GCCs and services), you'll be part of the problem, not solving it.
It's why I tell everyone I meet that if you don't understand the basics and read JP 5-0 and 3-0 you won't understand the butterfly affect that occurs from someone using your technology. Understanding how things connect and where is important
1
1
22
2,345
We experimented with this while at PACOM, when we put together our Gold Cell and looked to create a Joint Directorate for Economic Effects and have a close relationship with the Department of the Treasury. It was very fruitful, and plenty of collaboration still needs to happen. This needs to be a bigger focus in the DoW as we navigate this new form of war.
U.S. Africa Command recently set up its own Defense Economics Office to explore modern policy reforms, new ways to disrupt adversaries’ revenue streams and other money-related activities that can generate concrete security impacts, as it pivots to a significantly smaller physical presence across the African continent.
“A lot of times just the economic investment that we have done [with our African partners] is what provides us the ability to gain access or to be able to bring in ships and fueling, [and] allows us to be able to have that communication,” Africom commander Gen. Dagvin Anderson said. scoopmedia.co/4h4cRz8
1
1
6
1,970
Since joining @shieldaitech, the team has been working on putting LLMs and VLMs directly on-platform within our Hivemind suite.
We are having the LLM reason across CONOPs, OPLANs, mission constraints, and available actions from authoritative data sources to manage the ops/planning cycle. From there, the VLM interprets what the platform is actually seeing. Feed both into the autonomy stack to help navigate and adapt at the edge.
We have been using close sourced, large parameter LLMs for this to prove out the capability, but because we will have limited to no comms and no persistent connection to the cloud or large data centers, we have to start looking at building custom open weight models with key partners. This is already underway and we are excited about it the models we will be producing.
Two reasons for this 1. We need more open-weight models in the ecosystem to enable AI sovereignty and 2. We need better capabilities for on platform edge autonomy.
Powered by Hivemind: Building AI factories with @nvidia's AI stack
In a recent demonstration, Hivemind brought NVIDIA’s Cosmos 2 reasoning model together with Aechelon’s simulation capabilities to execute a four-agent wide-area search mission.
Operators orchestrated team intent instead of flying individual aircraft, moving operational experience faster toward trusted autonomous capability.
The AI Factory continuously tests and validates new capabilities, turning what is learned in the field into better performance for the next mission. It’s a continuous loop: fly, learn, improve, test, validate, fly again.
6
2
22
2,483
Started to Muse to build EDAs for my Boron Arsenide crystals. It's a work in progress but @alexandr_wang and team did a good job building @Muse and making it very dynamic. Building with Muse has been so fun.
1
2
386