the thing i was most skeptical about, giving it more context, mattered most
retained text
i resisted writing long detailed prompts for a while, it felt like doing the work myself, and i figured a good model shouldn't need it. i was wrong for my use cases. the gap between a vague ask and one where i explained the goal, the audience, and the constraints was night and day. it can't read my mind, and front-loading context paid back many times in fewer bad answers and less back-and-forth. the laziness of a short prompt cost me more time than the long one would. for users who resisted this too: did front-loading context change your results as much as it did mine?
document fetched 2026-09-30 00:23 · canonical https://reddit.com/r/ClaudeAI/comments/1wt8st1/the_thing_i_was_most_skeptical_about_giving_it