Trinary AI

I have been coding since I was about 9 years old. My first computer was a ZX Spectrum 48K, then a Commodore 64, and I went on to the PC stacks of the 90s. I studied low-level programming at the University of Cape Town, then started building startups, hiring engineers, and stopped programming for 20+ years during my career. I missed it, and I regretted deviating from where I started in life and my passion for computing for many years. It pained me. AI has changed all that. Humans are the most inefficient coders today. Those of us with the skills to embrace AI quickly will become 10x+ more efficient at what we already do in all facets of our digital lives. In just a few months, @trinaryai, we recreated an entire (and complex) computing stack from scratch (and for a fraction of the cost and time - coordination between humans is now just between sub-agents), and this is just the beginning of our journey. This technical progress means teams within big companies will start to shrink, and productivity will rise. There will be massive layoffs, but these people will leave to start their own companies, because capital requirements are plummeting across industries too. It's always been expensive to hire and find engineering talent. AI benefits the entrepreneur now more than ever. It's a crazy world out there, but it's clear to me now that software is a solved problem - hardware is where the real challenge is. I'm bearish on software companies without hardware or network effects and bullish on hardware companies (across the stack) - going forward it's about atoms, not electrons. The wealth creation effects of AI have barely been unlocked yet but I think we're all going to be very surprised by what happens next...

Here's a quick explainer about what we're building @trinaryai - might seem crazy, until it's not...

Better memory just extends the binary AI stack as it hits bandwidth and power limits. The bigger opportunity is native ternary chips + software: build around −1, 0, +1 so ternary AI runs directly on matching hardware. Rethink the whole stack, not just the memory. @trinaryai