Breaking Ant Mills
A few months ago I went to a tech event. Hadn’t been out for awhile, was expecting to hear about amazing forward looking visions of how LLMs might enable new forms of robotics, semiconductor designs or perhaps how network citizenship for digital nomads could lead to a revolution in government design.
Instead the primary topics were things that were in vogue 12 years ago. Something about edge computing, IoT and the mobile revolution. It was quite shocking to me — Nothing had changed. The same experts were discussing the same topics from years and years ago. It was like they were frozen in time.
The question that arose for me: “Am I crazy? Are all these new ideas and designs for the future that I see pure hallucinations? Or is everyone else just stuck walking in circles in familiar paths?”
At one point, 5G, Edge Computing and IoT were new and innovative. They certainly are not new and innovative now. The arrival of many new technologies that can provide exciting new possibilities by extending these core concepts into new areas mean some of the basic concepts have been completely disrupted or are on the verge of being disrupted.
In the semiconductor world, you have organizations like Tenstorrent effectively open sourcing their kernels, which will enable almost total end-to-end design of new processors and operating systems on-the-fly with software. If you have spent as much time in semiconductors as I have, you know that taking steps in this direction offer radical new avenues for innovation. Some of the biggest complaints about the most popular computing platforms such as the NVIDIA Jetson and Raspberry Pi revolve around the closed aspects of their designs.
Is anyone talking about this outside of a few niche YouTube videos? It certainly isn’t in my main feed or at any of the events I have attended.
A friend of mine turned me on to Groq, who seem to have created something called an LPU (Language Processing Unit) that drastically accelerates the performance of LLMs on hardware. In this video, which has 25 likes, and less than 1,000 views, Groq staff discuss using LLMs for instantaneous hardware design.
Last week on twitter (X), I ran across this project, the 01, which seeks to create a fully open source Rabbit r1 aka a “Language Model Computer.” Several hundred developers volunteered to spend time working on it in Seattle.
Hyperspace has arrived, offering a downloadable desktop client which enables peer-to-peer AI hardware sharing similar to BitTorrent. This is a an important direction for innovation, and a segment I have written about extensively in my Machine Economy thesis over the last three years.
This developer in Israel announced a completely open source AI pin to enable wearable LLMs. This direction is likely to open a completely new set of developers to creating every sort of wearable IoT LLM device imaginable.
Looking on LinkedIn and traditional tech — People feel listless, uninspired, focused purely on tactical near term goals or rehashing trends from years ago. We have all been stuck in an ant mill for the last few years.
I suppose that is the byproduct of an economic tech train wreck. But it’s time to snap out of it: MAJOR INNOVATIONS ARE ARRIVING PAY ATTENTION.
If you have been thinking, talking, writing and posting about the same 5 topics for the last three years, its time to consider exiting the ant mill you may be stuck in.
