“Winston believes that while LLM performance is plateauing for consumer use cases, it will continue to improve for enterprise applications and especially for code generation.”