Is prompting a lasting skill or a passing phase?

Part of prompting is a passing phase and part of it is a lasting skill. The wording tricks are the part most likely to fade. Knowing what you want, saying it clearly and asking a good next question is the part every one of these five voices expects to keep its value. Their apparent split mostly disappears once you ask which part they mean.

Amanda Askell, who leads work on Claude's character at Anthropic, treats serious prompts as careful work. She defines her terms, probes the model, looks for cases where it misunderstands, and revises, sometimes hundreds or thousands of times. She also says that for everyday tasks you should just ask. Heavy prompting matters for the last few percent of performance or for prompts at the core of a product. Her advice is about attitude: read your prompt as if you had none of the context in your head, and when the model errs, ask it why. Even inside the lab, a prompt is often a temporary fix. One line in Claude's instructions was dropped once training solved the problem it patched.

The founders of Cursor, the AI code editor, build an elaborate system that assembles prompts and ranks what goes into them. They do it so that their users never have to. Users should write naturally while the product finds the right context. They add that showing an example is sometimes easier than explaining, and that judgment and taste stay human.

Fei-Fei Li, the Stanford computer scientist, gives the strongest case for a lasting skill: "I would love to see our schools, K-12, teaching prompting." By prompting she means the Socratic method, seeking truth by asking questions. Her reason is agency, since people who learn the tool feel more in control.

Manolis Kellis, an MIT computational biologist, says we have always prompted each other, and that asking for step-by-step thinking works on his ten-year-old too. He warns students they will be replaced not by AI but by people who use it.

Aravind Srinivas, head of Perplexity, says prompt engineering won't last, because "a better product should be one that allows you to be more lazy, not less." He still thinks people are poor at asking questions, so his product suggests the next question for them.

The real disagreement is between Li and Srinivas: teach people, or fix the product. They mean different things. Srinivas predicts the end of hunting for wordings that work, and Li wants schools to teach inquiry. A difference remains. Li thinks the effort is good for the learner, and Srinivas thinks removing effort is the point. No one offers evidence on how fast the tricks will stop mattering.

What to do with it: for everyday use, ask in plain words. Before sending anything that matters, read it as a stranger would and add what is missing. Define your terms. When an answer is wrong, ask what instruction would have prevented it. Say who the answer is for and what form it should take. Show an example when that is easier than describing. Practice the follow-up question. Spend little time memorizing wordings and more on knowing your subject, writing clearly and judging results.