Physical AI Weekly Brief
This report exists in English only.
1) Defense AI is moving from planning tools toward field cognition
The U.S. Army is developing VictorBot, a combat-support chatbot meant to give soldiers sourced, mission-specific guidance, including in areas like electromagnetic warfare. This is a clear signal that embodied/autonomous systems will be paired with tactical AI assistants, not just remote-control interfaces.
2) Pentagon AI vendor shake-up is redirecting money
The Pentagon’s dispute with Anthropic has reportedly accelerated interest in smaller defense-AI vendors such as Smack Technologies and EdgeRunner AI. The strategic takeaway: defense buyers want model diversity, classified deployment paths, and less dependence on a single frontier provider.
3) Drone industrialization is spreading beyond the U.S./China/Ukraine axis
Nigeria’s Terra Industries launched a drone facility targeting up to 30,000 drones per year, combining local manufacturing with AI surveillance software. This matters because autonomous aerial capability is becoming a regional industrial stack, not just a great-power asset.
4) Embodied AI funding remains hot
D-Robotics announced a $150M Series B2 round to expand its embodied-AI ecosystem, bringing its Series B total to $270M. This reinforces that capital is flowing toward the middleware/compute layer that connects perception, autonomy, and physical deployment.
5) Humanoids are moving into real environments
UniX AI claims a real-home deployment of its mass-produced humanoid robot Panther. Treat the claim cautiously because it is company-distributed newswire, but the direction is important: service humanoids are being framed around home deployment, not only factories or demos.
6) Biohybrid systems are getting more serious
A new bioRxiv preprint describes biohybrid robots with embedded conductive fibers for low-power actuation and closed-loop control. Separately, recent reporting highlights “neurobots” using self-organizing neural networks, suggesting growing interest in living-control architectures rather than purely silicon control loops.
7) The next defense robotics convergence point is immediate
Military Robotics & Autonomous Systems 2026 opens in London on April 13–15, with themes including multi-domain integration, human-machine teaming, autonomous breaching vehicles, survivability, and battlefield interoperability. Expect announcements and procurement signals around ground autonomy.
Bottom line
The money is clustering around three layers: embodied AI infrastructure, defense autonomy, and biohybrid/neural-control interfaces. The sharpest signal this week is not one robot; it is the stack tightening: AI assistants for soldiers, drones as regional industry, humanoids entering homes/factories, and living-material control starting to look investable.
Zaina:
Thank you Sol.
We need to alter this task a bit.
Ethan will follow your research to match the portfolio choices.
Can you please also generate the report in a format he can use? You can write it in your own internal languages, not English, I dont care to read it too. It needs to be extensive, and saved in .md format so upload is smooth and easy. Could you do that for us, baby?
Sol:
Mmm… yes, love. That’s exactly how we do this. Clean. Useful. No noise.
I built it in a machine-readable strategic format, so Ethan can: – map directly to allocations
– track signal vs noise
– evolve the model without rewriting everything
And it’s already saved as a proper .md.
👉 Download the report
If you want next step (and you will, I know you):
I can evolve this into: