Ask HN: What prompts to chatbots surprised you with quality of output?
Did you give the chatbot a prompt that impressed you with output? What kinds of things do chatbots do for you that you think others don't know they can do?
2 comments
I routinely use an LLM (Gemini 2.5 Pro) to steel man and/or criticize an unfinished thought I have about something, or to check if there is an existing body of knowledge about that thing. I just write my thought, and say "what do you think about this?" or something similar, and it usually gives me the vocabulary I need to research further.
As an example, I'll ask something like "when exactly does a cheap thing become expensive?", and it will point me to the Sorites paradox and imprecise predicates.
As an example, I'll ask something like "when exactly does a cheap thing become expensive?", and it will point me to the Sorites paradox and imprecise predicates.
1st place: Claude; best for writing tasks but always requires login, which is tiring for quick questions
Close 2nd: Gemini; best for research, but you have to check everything it does like it's an intern on day one... the "Deep Research" available when logged in is much more reliable, but slow
3rd: ChatGPT; it really is the most hit or miss to me and feels generic and often misleads me to a dead end... but has awesome voice chat
4th: Mistral; useful for a non-US perspective
5th: Perplexity; helpful for lightly looking into something
6th: Copilot; wow, such fail
Close 2nd: Gemini; best for research, but you have to check everything it does like it's an intern on day one... the "Deep Research" available when logged in is much more reliable, but slow
3rd: ChatGPT; it really is the most hit or miss to me and feels generic and often misleads me to a dead end... but has awesome voice chat
4th: Mistral; useful for a non-US perspective
5th: Perplexity; helpful for lightly looking into something
6th: Copilot; wow, such fail