I drive a truck for a living and I’m not worried about AI taking over my role. What I do think about is what size population I may be serving in the future. My local university has stopped taking on new staff and it’s asking for voluntary redundancies. The period of time when AI tools will assist desk workers will be relatively short lived. I can’t forsee a world where professions like law, finance, accounting will flourish among populations that have little to do except consume. What role will the opinions of enlightened but static sages have in a world where they’re thoroughly dependent on supply chains putting products on the shelves regardless. During Covid the UK population of 65 million was serviced by 3.5 million ‘essential’ key workers. How many of those 3.5 million workers cared what anyone said or thought about the goings on then or now.
One thing worth flagging for founders reading this: the 'billion machines' framing is compelling but the actual bottleneck in physical AI has rarely been the agentic layer—it's been sensor fusion and edge compute constraints that don't show up until you're past the demo stage. I've seen a few robotics teams get a slick simulation-to-deployment pipeline working, then hit a wall when real-world latency requirements didn't match what the platform assumed. Does Dana address the sim-to-real transfer problem directly, or is that still left to the hardware partner?
The real bottleneck in physical AI is no longer intelligence alone. It‘s about getting intelligence through simulation, validation, regulation and hardware without losing traceability along the way.
That is why the move from tool → infrastructure → operating system → agent matters. Dana is not just another coding layer. It’s trying to compress the entire development loop while preserving the chain of evidence required when software starts moving several tons of metal near human beings.
Models may become abundant. Deployment memory, evaluation infrastructure and trust will not. That is where the durable moat probably sits.
This analogy is amazing, because I remember how Mosaic-Netscape changed my life in 1993:
The web browser offers a historical corollary. It didn’t “create” the underlying internet, but it did make the net usable by the vast majority of people who weren’t nerdy hobbyists, and trillions of dollars of economic activity moved toward what that access allowed them to build and do. Applied Intuition’s progression from simulator to data infrastructure to operating system followed the work its customers were already doing. As more of that work moved into Applied’s products, the software became part of how those companies built machines rather than a tool used for one stage of development.
I'm so very glad America has so many people way smarter than I am to move us into the future and to recognize these entrepreneurs by investing in their ideas.
I drive a truck for a living and I’m not worried about AI taking over my role. What I do think about is what size population I may be serving in the future. My local university has stopped taking on new staff and it’s asking for voluntary redundancies. The period of time when AI tools will assist desk workers will be relatively short lived. I can’t forsee a world where professions like law, finance, accounting will flourish among populations that have little to do except consume. What role will the opinions of enlightened but static sages have in a world where they’re thoroughly dependent on supply chains putting products on the shelves regardless. During Covid the UK population of 65 million was serviced by 3.5 million ‘essential’ key workers. How many of those 3.5 million workers cared what anyone said or thought about the goings on then or now.
Hey Marc,
Writing about private companies weekly - would love to get your thoughts and feedback on any of my posts.
Written on everything from defense tech (Anduril, Saronic) to AI (OpenAI, Anthropic).
Thanks and all the best.
Joseph
Marc - love your insights but how can we make European industry great again?
Marc is back!
That’s why we’re building the first Robot Native Insurance 🦾🦅🇺🇸
One thing worth flagging for founders reading this: the 'billion machines' framing is compelling but the actual bottleneck in physical AI has rarely been the agentic layer—it's been sensor fusion and edge compute constraints that don't show up until you're past the demo stage. I've seen a few robotics teams get a slick simulation-to-deployment pipeline working, then hit a wall when real-world latency requirements didn't match what the platform assumed. Does Dana address the sim-to-real transfer problem directly, or is that still left to the hardware partner?
The ritual of engineers rebuilding the same tools at every new company is quietly repeating right now with agent evals.
The real bottleneck in physical AI is no longer intelligence alone. It‘s about getting intelligence through simulation, validation, regulation and hardware without losing traceability along the way.
That is why the move from tool → infrastructure → operating system → agent matters. Dana is not just another coding layer. It’s trying to compress the entire development loop while preserving the chain of evidence required when software starts moving several tons of metal near human beings.
Models may become abundant. Deployment memory, evaluation infrastructure and trust will not. That is where the durable moat probably sits.
Making a soulless army of automatons
everyone should have their own private robot army
this is true abundance
Sounds interesting but there is so little information on the underlying technology of Dana, so it’s hard to evaluate how realistic it is
This analogy is amazing, because I remember how Mosaic-Netscape changed my life in 1993:
The web browser offers a historical corollary. It didn’t “create” the underlying internet, but it did make the net usable by the vast majority of people who weren’t nerdy hobbyists, and trillions of dollars of economic activity moved toward what that access allowed them to build and do. Applied Intuition’s progression from simulator to data infrastructure to operating system followed the work its customers were already doing. As more of that work moved into Applied’s products, the software became part of how those companies built machines rather than a tool used for one stage of development.
I'm so very glad America has so many people way smarter than I am to move us into the future and to recognize these entrepreneurs by investing in their ideas.
Any interest in remote pilot patents in case they need it for regulatory reasons? https://patents.google.com/patent/US10775783B2/