but it's still literally just looking at the text and calculating what's the most likely thing to follow, that's fundamentally how LLMs work