Antirez understood the fundamentals of how LLMs work, and read the inference code (or summaries of it via ai).
But what really made the difference was his understanding of hardware and systems programming in general and low level or architectural tricks to pull.
wg0 · · focus · HN ↗
The other inference engine are also model by model with a huge switch statement deciding which part to load for which model or are they very generic?
epolanski · · focus · HN ↗
But what really made the difference was his understanding of hardware and systems programming in general and low level or architectural tricks to pull.