.@shieldaitech just hit a milestone that rewrites how European air defense works.
Their Hivemind autonomy software, already running 130+ combat sorties in Ukraine, just finished a two-month flight test in Spain. The platform was a Swiss-made interceptor drone called Hornet, built by Destinus. No ground control station needed. No GPS. No manual reprogramming while airborne.
Destinus built the Hornet to kill other drones autonomously. Shield AI's Hivemind runs onboard the aircraft itself, not on some ground station. During testing, geofenced zones were altered while the drone was already in the air. It rerouted on its own. No human touched anything.
That's the first time a US autonomy stack has been integrated onto a European combat drone and it worked.
And the operational proof already exists. V-BAT drones running Hivemind are flying real missions in Ukraine right now, punching through heavy Russian electronic warfare and jamming. Still delivering results. The Netherlands just ordered 12 of them for naval ops across eight warships.
The pipeline keeps growing:
> Hornet (counter-drone interceptor), tested and baselined in Spain. > V-BAT (ISR and strike), operational in Ukraine, purchased by the Dutch. > X-BAT (stealthy loyal wingman), currently being shown to European militaries.
Next up: terrain-aware flight profiles and coordinated multi-drone swarm behaviors.
Here's the part that caught my attention. No customer funded this test. Both companies paid from their own R&D budgets. That's a calculated bet: European air forces need autonomous counter-drone systems that work without crewed aircraft or GPS, and they need them yesterday.
The company building the software layer for autonomous air combat just proved it works across platforms and across continents.
Austin-based startup @neurophos just built an AI chip that processes data with light instead of electrons. The performance improvements are wild.
Neurophos uses optical metamaterials, the same physics behind invisibility cloaks (used mostly on clothing), to build optical modulators that are 1/10,000th the size of current designs. Standard CMOS. No exotic materials.
They fit 1,000 x 1,000 optical modulators on a 5x5mm chip.
The same setup in silicon photonics?
About a square meter.
Same compute. 1/40,000th the area.
50x the compute density and 50x the energy efficiency of NVIDIA's Blackwell GPUs.
Here’s how it works: a laser encodes data, hits the chip, each metamaterial element changes the reflected beam, and the system outputs results from complex AI workloads with far less power draw.
Hyperscalers are evaluating proof-of-concept chips this year. Production ramp: mid-2028.
Scaling AI is now bound by the laws of physics. Neurophos is betting the future of compute isn’t faster electrons, it’s light.
.@MGStoverCo Acquires Asymmetric Information to Power AI-Driven Intelligence for Institutional Crypto - http://markets.businessinsider.com/news/1035935751
Four years ago, it was clear to me that the digital assets space lacked proper tooling and infrastructure for monitoring, managing and reporting on risk across the investment landscape, from Venture captial investments to SPVs to complex derivative portfolios, and hence, Asymmetric Information was born out of necessity.
Today, Asymmetric Information is joining the industry's best and most tenured (literally the first ever) fund administrator, @MGStoverCo to provide the most comprehensive, AI-enabled suite of back, front and middle office software for digital asset investment professionals.