Bringing a new computing approach from research to hardware that developers can use usually takes years, sometimes more than a decade. Sometimes it never makes it to market.
But we did it in about a year.
Today, with a team of three and the support of @fdotinc, we've shipped our first few dev kits with 10,000 synthetic neurons each, running computation based on oscillations and neural dynamics directly on hardware (paper out tomorrow).
Not an algorithm running on a CPU. Not a GPU. Synthetic neurons on hardware!
We believe in creating an open ecosystem and making the technology accessible to anyone who wants to experiment with synthetic neurons and build the foundations of biomimetic intelligence.
Together, we can create a community that moves away from the belief that brute-forcing data and scaling computing will solve robotics. We can create a community that takes the opposite approach and follows the cheat code nature handed to us.
A few years from now, I'm sure we'll look back and laugh at the approaches we took to build intelligent robots.
Seven dev kits sold last week to true pioneers, and counting. We're just getting started.
In the photo: our first dev kit, Prima-Lux, with a capacity for up to 10,000 synthetic neurons is delivered and sits where it belongs! Thank you deeply @FurqanR , our first believer!
Robotics looks like a fashion show switching up trends at a moments notice
But they are even shorter lived here
We started with VLAs
Moved on to World Models
With the release of Astra we have moved on to "agentic robotics" (whatever that means) in a matter of months
If you gave me a robot with supra human inteligence (please do), I'd tell you that's not even half the path.
Bottlenecks in 6 months will be:
- convincing factory owners
- robotics economy of scale
- finding the right revenue structure (not everything will be RaaS)
The best time to enter an industry is when consensus is low
That's robotics rn
Everyone has a slightly different opinion and there are not a lot of hard facts around
Start working on atoms instead of bits!
@joaopanizzutti If you go far back enough, the root of llms are neural networks, aka neurons
Also, look up the inspiration for winglets
There's a balance to be had