Suppose you're using an LLM that allows up to a 4000-token input.
	Which of the following is good advice for asking that LLM questions?

	A. Aim for a 4000-word question: anything less wastes the LLM's inputs.
	B. Aim for a 3000-word question: use the full input, but leave room for non-word tokens.
	C. Aim for a 2000-word question: half the LLM's inputs will end up being its own output.
	D. Start with a short question, then ask follow-ups, aiming for a 4000-word conversation.
	E. Keep the question short; shorter questions use less energy.


















	Suppose you're using an LLM to do a complicated task, like programing or research.
	Which analogy below is most likely to guide your use of the LLM correctly?

	A. It's like an arrogant person: it lies when confused and never backs down.
	B. It's like a sycophant: if you tell it something it will agree without reservation.
	C. It's like a crowd: each answer might come from someone different, correcting one another.
	D. It's like a search engine: its responses are based on what you type and what it read.
	E. It's like a genie: its power is vast but hidden behind caprecious rules and secret phrases.

















	An agentic AI agent can be given a task and will then do many things along the way to completing it,
	including making notes for itself, trying things that fail and then trying something else, and so on.
	This process represents
	
	A. the AI stumbling around at random until it trips over the answer.
	B. the AI emulating its training data, which contains many humans doing processes indirectly.
	C. the AI asking an LLM how to get an AI to complete the task, then following that advice.
	D. the AI engaging in online learning, discovering more about the task as it goes.