Artificial Intelligence and Defense: Research & Analysis
Joined March 2016
- Tweets1.2K
- Following2.4K
- Followers699
- Likes1.5K
Pinned Tweet
Counter-drone isn’t a niche. It’s the leading edge of an air-defense rebuild the West deferred for twenty years.
I put together a quick briefing document on some thoughts regarding the counter-drone (C-UAS) ecosystem.
New models are emerging at speed. Counter-drone-as-a-service is already live and private infrastructure providers and critical facilities are being given the legal authority to take direct responsibility for their own airspace defense. Download the pdf here: drive.google.com/file/d/1KBR…
#CUAS
George Howell retweeted
Holy crap.
Rick and Morty just explained Jev AI to me better than any tech demo could.
George Howell retweeted
The first private Anti-air company in the world works in Ukraine.
Operators in Carmin Sky use joysticks to control automatic turrets. Sky Sentinel turrets, equipped with Browning machine guns are used. They locate target automatically and wait for permission to fire — DW.
1/
George Howell retweeted
Read this excellent report on counter UAS systems from George Howell of RAINCLOUD Defense. He mentions the importance of #Hyperwar as an organizing framework for counter-drone systems. Ultimately, autonomous drones are taking us all the way to a collapsed OODA loop.
Learn more: raindefense.ai
George Howell retweeted
Very pleased to announce that General John R. Allen and I have launched Hyperwar.net - a hub for the ideas, publications, and media shaping #Hyperwar.
Hyperwar is the emerging doctrine that defines how autonomy, AI, and machine-speed conflict will reshape defense. Since we first defined it in 2017, it has entered military training programs, influenced defense technology development, and become a widely cited term in strategic literature.
Hyperwar.net is an effort to bring structure, reference, and clarity to this ongoing evolution. We’ll continue to build this out over time.
LINK: hyperwar.net
George Howell retweeted
From #nuclear tech to #AI, #biotech & satellites, dual-use technologies power growth but carry risks of misuse. The core challenge: enabling innovation while preventing coercion, #cyber threats & weaponisation: @lakshmyrkrish or-f.org/37957
George Howell retweeted
someone built an OPENSOURCE MILITARY RADAR that tracks multiple targets up to 20km away.
contractors charge a quarter-million dollars for this tech. one dev just put the entire github repo online for free.
pcbs, fpga code, schematics, python gui. all under MIT license..
George Howell retweeted
Just because drones share airspace with crewed aircraft doesn't mean the most logical Army branch to be charged with modernizing them is aviation. mwi.westpoint.edu/who-owns-t…
George Howell retweeted
"The United States cannot treat AI advantage as a stockpile—a lead built on breakthrough discoveries that adversaries will take years to replicate. Idea fluidity renders this paradigm obsolete." mwi.westpoint.edu/you-cant-s…
George Howell retweeted
Every airframe, flight, and crash in this video was paid for by Anduril, not taxpayers. We believe so much in Omen that we spent five years building on our own dime.
Thousands of miles of range from a Group 3 VTOL that can carry hundreds of pounds of payload is a game changer.
As industrialized warfare collides with rapid technological attrition, defense supremacy will not be determined by who builds the best drone, but by who can outproduce, out-iterate, and out-adapt the fastest. Static acquisitions and niche innovation are insufficient. What is needed are Defense Foundries: integrated production ecosystems that merge engineering, automation, and software to manufacture and evolve entire cohorts of uncrewed systems on demand. Unlike traditional factories built for repetition, foundries are designed for adaptability: engineered to retool quickly, integrate new technologies, and sustain output under pressure. This paper introduces the Defense Foundry Model as a strategic framework for national industrial resilience, proposing a shift in focus from platform procurement to agile production capabilities, aligning military needs, government policy, investor incentives, and industrial capacity under one unifying model.karveinternational.com/insig…
George Howell retweeted
Well of course as the USA doesn't listen to @BrianRoemmele
"America gets a much better McKinsey, and China gets a much better Foxconn."
Dan Wang explains how each country will use AI:
The U.S. will automate services, while China will automate manufacturing.
China will use AI "to produce a lot more drones, munitions, and ships as well. The U.S. simply doesn’t have the training data or the process knowledge in place to get much better at manufacturing."
George Howell retweeted
3D-Printed Drones, Wildfire Tech, and UGV Integrations
Daily Hard Tech Headlines:
- @K9DefenseTech, the Huntsville-based UAS company, announced that the Army’s 3rd Combat Aviation Brigade has 3D-printed, assembled, and completed a K9 Genesis UAV on base using the company’s open-source wiki, files, and guides.
- @HavocAi_USV has teamed up with Ilmor Engineering Ltd., a leading maritime engine maker, to bring advanced electric propulsion into its autonomous vessel platforms.
- Korean Air has secured a partnership with @anduriltech. The collaboration focuses on UAS for wildfire response.
- @einrideofficial, the Swedish autonomy startup, has raised $100 million. The company develops autonomous trucking technology.
- USA Rare Earth, Inc. (Nasdaq: USAR) has announced the acquisition of LESS COMMON METALS LIMITED, a critical producer of rare earths. The deal size sits at $100M.
- Uber has acquired Segments ai, a data-labeling firm for autonomous systems.
- American Pacific Corporation, a maker of materials used in solid rocket propulsion, has broken ground on its $100 million expansion in Cedar City, Utah.
- Occidental Petroleum’s chemical division is being acquired by Berkshire Hathaway. The unit manufactures large-scale industrial chemicals used in biotechnology, construction, water treatment, and other sectors. The deal size sits at $9.7B
- @Microsoft is expanding its clean power footprint in Japan with new solar deals from Shizen Energy Group, bringing its total commitment to 100 MW.
- DEEP is scaling its U.S. presence with a $100 million push, setting up major manufacturing facilities in Florida and Texas. Central to the plan is Project Vanguard, a next-generation undersea habitat
- @allencontrol has announced an integration of their C-UAS platform onto a Forterra UGV.
ALT Source: Allen Control Systems
George Howell retweeted
🇨🇳China unveils world's first brain-like AI Model SpikingBrain1.0
Upto 100X faster while being trained on less than 2% of the data typically required.
Designed to mimic human brain functionality, uses much less energy. A new paradigm in efficiency and hardware independence.
Marks a significant shift from current AI architectures
Unlike models such as GPT and LLaMA, which use attention mechanisms to process all input in parallel, SpikingBrain1.0 employs localized attention, focusing only on the most relevant recent context.
Potential Applications:
- Real-time, low-power environments
- Autonomous drones and edge computing
- Wearable devices requiring efficient processing
- Scenarios where energy consumption is critical
This project is part of a larger scientific pursuit of neuromorphic computing, which aims to replicate the remarkable efficiency of the human brain, which operates on only about 20 watts of power.
---
arxiv .org/abs/2509.05276
SpikingBrain’s technical report reveals a new family of brain-inspired LLMs.
Learn how its hybrid-linear attention, conversion-based training, and spiking neurons deliver over 100x speedups and unprecedented efficiency on non-NVIDIA hardware.
100x faster first token at 4M tokens, with training on about 2% of the usual data.
Transformers slow down as sequences grow, because each new token checks many earlier tokens and the memory cache keeps growing.
SpikingBrain mixes 2 cheaper attentions, linear keeps a small running summary, sliding-window reads only a short slice.
The 7B model alternates these layers for near linear cost, the 76B model adds parallel branches and a few full layers.
Feed forward blocks use Mixture of Experts, a router picks a small set per token so most weights stay idle.
The key idea is adaptive threshold spiking, activations become integer counts during training then expand into sparse events at inference.
A light conversion pipeline remaps a standard checkpoint, extends context to 128k, then finishes with supervised fine tuning.
Everything runs on MetaX C550 GPUs, the 7B model keeps memory near constant as inputs grow and accuracy stays close to baselines.
----
Paper – arxiv. org/abs/2509.05276
Paper Title: "SpikingBrain Technical Report: Spiking Brain-inspired Large Models"
George Howell retweeted
We’re partnering with @anduriltech Industries to power its mixed-reality combat system for the U.S. Army’s Soldier Borne Mission Command Architecture program. By integrating our near real-time 3D terrain foundation with Anduril’s body-worn command and control system, soldiers gain accurate battlefield intelligence directly in their field of view.
Our software automatically generates high-fidelity 3D terrain from satellite imagery and updates it within hours, ensuring soldiers can trust what they see when split-second decisions determine mission success.
Learn more: maxar.com/press-releases/max…
