Cast by sa

Good point, I'm sure that's what they mean. I only known of one example of hardware actually being the code. It's a fascinating attempt to increase response speed: How Taalas "prints" LLM onto a chip? | daily.dev

How Taalas "prints" LLM onto a chip? | daily.dev

Taalas, a startup, has built a fixed-function ASIC chip with Llama 3.1 8B weights physically etched into silicon transistors, achieving 17,000 tokens/second...