Abstract
This paper presents a puzzle in the form of a tension between two things: (1) data points about what intuitively count as good guesses to a question given an agent's subjective probabilities, and (2) plausible strategic norms of inquiry that tell us how to go about answering a question. Recent theories of guessing and good guessing have acknowledged this puzzle in one form or another, and they aim to get around it by appealing to some sort of contextualism. But I argue that the puzzle is more pressing than has been appreciated, that it makes it hard to understand how individually good guesses can form part of an overall rational inquiry, and that the appeal to contextualism in response to it relies on a very implausible kind of semantic blindness. In other words, I argue that the puzzle stands as a serious challenge to our understanding of how good guessing relates to rational inquiry.