Prompt Iteration Beats Token Counting | Generated by AI
Question: How do I change “提高 AI 应用水平” to better express “the ability of using AI or LLM” rather than “making AI applications”?
Answer:
既然”天才是百分之一的灵感，百分之九十九的汗水”，那我先来谈谈这汗水的部分吧。有人问我，提高驾驭 AI 与大模型的能力最有效的办法是什么？我想了很久，终于发现最有效的办法，其实是反反复复地修改和推敲输入给大模型的提示词（Prompt）与上下文（Context）。
有些人喜欢炫耀自己用过多少个 AI 工具，或者使用了多少 Token，仿佛使用的数量是衡量驾驭 AI 能力的标准。然而，如果你总是匆匆写出一个提示词，却从来不回头去推敲、修改和提炼提示词与上下文，其实是不能真正提高驾驭 AI 与大模型的能力的。你只会制造出越来越多平庸甚至糟糕的输出结果。
“使用经验”跟提示词和上下文的质量，其实不一定成正比。如果有多年使用 AI 工具的经历，却从来不回头去提炼和反思自己的 Prompt 与 Context，那么他也许还不如一个只有一两个月经验、却喜欢反复打磨、仔细领悟模型行为的人。
English Version:
Since “genius is one percent inspiration and ninety-nine percent perspiration,” let me talk about the perspiration part first. People often ask me: what is the most effective way to improve your ability to wield AI and large language models (LLMs)? After much thought, I’ve come to realize the answer is iterating relentlessly on both the prompts and the context you provide to LLMs.
Some people like to boast about how many AI tools they’ve tried, or how many Tokens they’ve consumed — as if sheer volume is the measure of one’s ability to wield AI. But if you hastily write a prompt and never go back to refine, revise, and improve both the prompt and the surrounding context, you won’t actually grow. You’ll just keep producing mediocre or poor outputs at scale.
“Years of experience” with AI tools does not necessarily equal quality. Someone with years of usage but no habit of reflecting on their prompts and context may still fall short of a person with just a few months of experience who obsessively iterates, experiments, and deeply understands how models respond to different inputs and contextual setups.
The real skill is not in how many Tokens you’ve consumed — it’s in how carefully you’ve refined both the prompt and the context around it.
Note on the Chinese phrasing: “驾驭 AI 与大模型的能力” literally means “the ability to command/master AI and large models” — which clearly conveys using AI skillfully, not building AI applications. Other natural alternatives include: 「善用 AI 与大模型的能力」(the ability to make good use of AI/LLMs) or 「用好 AI 与大模型的能力」(the ability to use AI/LLMs well).