🌍 世界数字教育网
模块 2 · 学习方法Module 2 · Learning Methods

把 AI 当你的私人老师

Turn AI into Your Personal Tutor

免费、全天在线、无限耐心——一个 24 小时的"老师"就在你手里。但大多数人只是"问答案",浪费了 90% 的能力。这一页教你怎么"用"它学习。

Free, available 24/7, infinitely patient — a "teacher" in your hand. But most people just "ask for answers" and waste 90% of its power. This page teaches you how to actually use it to learn.

🤖 24 个板块 · 30 条提示词 · 12 学科 · 30 天计划 24 sections · 30 prompts · 12 disciplines · 30-day plan
01

AI 老师是什么:它强在哪、弱在哪

What an AI Tutor Is: Where It's Strong and Where It's Weak

先搞清楚这个"老师"的能力边界,才用得对。
Know this "teacher's" capability boundaries before using it right.

大语言模型(ChatGPT、DeepSeek、豆包、Kimi、Gemini)本质是"模式识别 + 语言生成"引擎。它读过几乎所有公开的教科书、论文、教程,能像"读过很多书的人"一样组织语言、举例子、答疑。但它不是"数据库"(不保证每个事实都对),也不是"考官"(不会真正评判你)。把它当老师,核心是三件事:它"会讲"(能把复杂事讲简单)、它"有耐心"(同一个问题可以问 100 遍)、它"不打烊"(凌晨 2 点也能问)。

Large language models (ChatGPT, DeepSeek, Doubao, Kimi, Gemini) are essentially "pattern recognition + language generation" engines. They've read virtually every public textbook, paper, and tutorial, and can organize language, give examples, and answer questions like "someone who has read many books." But they are not "databases" (not every fact is guaranteed correct) nor "examiners" (they don't truly judge you). Using it as a teacher comes down to three things: it can explain (turns complex into simple), it's patient (you can ask the same thing 100 times), and it's never closed (ask at 2 AM).

能力Ability强度Strength说明Notes
把复杂概念讲简单Explain complex things simply★★★★★★★★★★最大优势:类比、分步、多版本Biggest strength: analogies, steps, versions
生成练习题/测验Generate exercises/quizzes★★★★★★★★★★可指定难度、格式、数量Specify difficulty, format, count
无限追问与讲解Infinite follow-ups★★★★★★★★★★同一问题可换 10 种方式再讲Re-explain the same thing 10 ways
批改作文/代码Grade essays/code★★★★★★★★给建议、指错、给改进版Suggestions, errors, improved version
制定学习计划Build study plans★★★★★★★★给框架,但需要你自己校准Gives framework; you calibrate
事实准确性Factual accuracy★★★★★★会"幻觉":编造不存在的细节"Hallucinates": invents details
真正评估你水平Truly assess your level★★★★只能"猜",需要你主动反馈Only "guesses"; you must feed back
持续跟踪与督促Track & motivate★★★★无记忆(除非用其记忆功能)No memory (unless using its memory feature)
📊96%的青少年每周使用 AI 辅助学习(2025 年调查)——问题是"大多数人只拿它抄答案"。of teens use AI to help study weekly (2025 survey) — the problem: most only use it to copy answers.Pew Research / Common Sense Media 2025
⚠️最大的坑:把 AI 当"答案机"——你问"答案是什么",它给答案,你抄下来。这样你什么也没学会,反而训练了"依赖"。正确的用法是把它当"苏格拉底":你问"为什么",它引导你自己想出来。记住:AI 老师的价值 = 你追问的深度。Biggest trap: treating AI as an "answer machine" — you ask "what's the answer", it gives the answer, you copy it. You learn nothing and train dependency. The right usage: treat it as "Socrates" — you ask "why", it guides you to think it out. Remember: an AI tutor's value = the depth of your follow-ups.
02

AI 学习四步循环:学 → 问 → 练 → 考

The 4-Step AI Learning Loop: Learn → Ask → Practice → Test

把 AI 嵌入"学习科学"的循环,而不是随机提问。
Embed AI in the "learning science" loop instead of random Q&A.

有效的 AI 学习不是一堆随机提问,而是一个循环:先用 AI 学(听讲解),再追问(暴露理解漏洞),然后练(做它出的题),最后考(让它测你)。每一步都有明确的目的,缺了任何一步,效果都会打折扣。这个循环把"AI 老师"和"学习科学"(间隔重复、主动回忆、检索练习)结合在一起。

Effective AI learning isn't a pile of random questions — it's a loop: first learn with AI (listen to explanations), then ask (expose understanding gaps), then practice (do its exercises), finally test (let it quiz you). Each step has a clear purpose; skip any and the effect drops. This loop combines the "AI tutor" with "learning science" (spaced repetition, active recall, retrieval practice).

步骤Step做什么What to doAI 的角色AI's role你的角色Your role时间占比Time share
① 学Learn让 AI 讲概念,要求"讲给 12 岁孩子听"Ask AI to explain, "like to a 12-year-old"老师(讲解)Teacher (explains)学生(听懂+笔记)Student (understand + note)30%30%
② 问Ask追问"为什么""换种说法""举反例"Ask "why", "rephrase", "give counterexample"苏格拉底(引导)Socrates (guides)提问者(暴露漏洞)Questioner (exposes gaps)30%30%
③ 练Practice让它出题,你手写答案,再让它批改It makes questions; you answer by hand; it grades出题人+批改者Question-setter + grader答题者(必须动手)Answerer (must write)30%30%
④ 考Test过 1 天/3 天/7 天,让它出综合题测你1/3/7 days later, it sets comprehensive tests考官Examiner考生(回忆检索)Examinee (recall/retrieve)10%10%
💡关键原则:"练"和"考"必须动手写——不要只看 AI 的答案。学习科学反复证明:检索(回忆)比重复阅读有效 2–3 倍。所以让 AI 出题、你亲手答、再让它批改,这个"出题→手答→批改"闭环才是学习真正发生的地方。光"听讲解"只学到了 30%。Key principle: "Practice" and "Test" must be hand-written — don't just read AI's answers. Learning science repeatedly shows retrieval (recall) is 2–3× more effective than rereading. So "AI sets question → you hand-answer → it grades" is where learning actually happens. Just "listening" only gets 30%.
03

提示词基本功:让 AI 当老师的 6 个句式

Prompt Fundamentals: 6 Sentence Patterns for an AI Teacher

同样的 AI,提示词不同,效果差 10 倍。
Same AI, different prompts, 10× different results.

提示词(prompt)是你对 AI 下指令的方式。大多数人说"帮我学 X",AI 就给你一个泛泛的概述。高手会说清"我是谁、我想知道什么、以什么方式给我、到什么程度"。下面 6 个句式覆盖了 90% 的学习场景,背下来就能用。

A prompt is how you instruct the AI. Most people say "help me learn X" and get a generic overview. Experts specify "who I am, what I want to know, in what form, to what depth." The 6 patterns below cover 90% of learning scenarios; memorize them and you're set.

句式Pattern模板Template用途Use
身份+水平Identity + level"我是大一学生,刚学完高中数学,请从我的水平讲起…""I'm a freshman, just finished high-school math, start from my level…"定起点,避免太浅/太深Set starting point
费曼讲解Feynman explain"像给 12 岁孩子讲一样,用一个生活类比解释…""Explain like to a 12-year-old, with one real-life analogy…"把复杂变简单Simplify complexity
苏格拉底Socratic"不要直接告诉我答案,通过提问引导我自己想出来""Don't tell me the answer; guide me with questions to think it out"训练思考,不抄答案Train thinking, no copying
出题+批改Quiz + grade"出 5 道关于 X 的题(3 基础 2 进阶),我答完你批改并指出漏洞""Give 5 questions on X (3 basic, 2 advanced); after I answer, grade and point out gaps"检索练习Retrieval practice
找漏洞Find gaps"这是我对 X 的理解:… 哪里错了?哪里不够深?""Here's my understanding of X: … where am I wrong? Where's it shallow?"暴露理解漏洞Expose gaps
计划+校准Plan + calibrate"给我一份 2 周学完 X 的计划,每天 1 小时,每步告诉我怎么检验学会了""Give a 2-week plan to finish X, 1h/day, tell me how to verify each step"制定可检验的计划Build verifiable plan
🎯万能开头:几乎所有学习提示词都可以这样开头——"你是我的老师,我是[身份],水平[水平],想学会[目标],请[方式]。"把四个空填好,AI 立刻从"搜索引擎"变成"私人老师"。剩下的 6 个句式,按"学→问→练→考"循环往里填。Universal opener: almost any learning prompt can start with — "You're my teacher, I'm [identity], level [level], goal [goal], please [method]." Fill the four blanks and the AI instantly becomes a "personal tutor" instead of a "search engine." Then slot the 6 patterns into the "learn→ask→practice→test" loop.
04

追问的艺术:10 个让理解翻倍的追问句式

The Art of Follow-Up: 10 Patterns That Double Understanding

第一轮回答只值 40 分,追问才值 60 分。
The first answer is worth 40%; the follow-ups are worth 60%.

大多数人问完一个问题就走了。但理解来自追问:第一轮回答给你"地图",追问给你"细节、反例、边界、联系"。下面 10 个追问句式,每个都对应一种理解漏洞,按顺序用下来,理解会翻一倍。

Most people stop after one question. But understanding comes from follow-ups: the first answer gives the "map"; follow-ups give "details, counterexamples, boundaries, connections." The 10 patterns below each target a type of understanding gap; used in sequence, they double your grasp.

##追问Follow-up暴露的漏洞Gap exposed
11"为什么是这样?背后的原理是什么?""Why is it this way? What's the principle behind it?"只记了"是什么",不懂"为什么"Knew "what" not "why"
22"换个说法/打个比方""Rephrase it / give an analogy"表面懂了,换个说法就糊涂Seemed to get it, fails when rephrased
33"有没有反例?什么情况下不成立?""Any counterexample? When does it fail?"不知道边界和例外Unknown boundaries/exceptions
44"这和我知道的 X 有什么关系?""How does this relate to X I know?"知识点孤立,没形成网络Isolated knowledge, no network
55"给我一个生活里的例子""Give me a real-life example"抽象概念落不了地Abstract concept doesn't land
66"常见的误解有哪些?""What are common misconceptions?"被错误理解绑住Bound by wrong understanding
77"如果我[改一个条件],会怎样?""If I [change one condition], what happens?"只会套公式,不会变通Only applies formula, can't adapt
88"这个的底层原理,再深一层是什么?""Go one level deeper on the underlying principle?"只停在表面Stops at the surface
99"用一句话总结 + 3 个关键词""Sum up in one sentence + 3 keywords"抓不住重点Misses the key point
1010"出 3 道题考我,从易到难""Quiz me with 3 questions, easy to hard"以为懂了,一考就露馅Thought got it, fails on test
💡追问节奏:一次对话里,"学"(第 1 轮)之后,按"为什么→换说法→反例→联系→例子"的顺序追问 3–5 轮,再让它"出题考你"。一轮对话学透一个知识点,比泛泛学 10 个知识点有效得多。深度 > 广度,尤其在打基础阶段。Follow-up rhythm: in one conversation, after "learn" (round 1), follow up 3–5 rounds in the order "why → rephrase → counterexample → connection → example", then have it "quiz you." Mastering one point deeply in one conversation beats skimming 10 points. Depth > breadth, especially when building foundations.
05

12 学科:AI 老师怎么用(对号入座)

12 Disciplines: How to Use the AI Tutor (match your field)

不同学科,AI 老师的"用法"不一样。
Different disciplines, different ways to use the AI tutor.
学科DisciplineAI 最擅长的AI's best at推荐循环Recommended loop典型提示词Typical prompt
数学Math分步讲解 + 出题Step-by-step + questions学→练→考(重练)Learn→Practice→Test (practice-heavy)"这道题分步讲,每步为什么""Explain this problem step by step, why each step"
物理Physics类比 + 可视化描述Analogies + visual description学→问→练Learn→Ask→Practice"用生活例子解释这个现象""Explain this phenomenon with a real-life example"
化学Chemistry记忆口诀 + 规律Mnemonics + rules学→练→考Learn→Practice→Test"帮我编个记忆这个规律的口诀""Make me a mnemonic to remember this rule"
外语Languages对话 + 批改 + 例句Conversation + grading + examples学→问→练(重对话)Learn→Ask→Practice (conversation-heavy)"和我用英语对话,纠正我的错误""Talk to me in English, correct my errors"
编程Programming讲代码 + 找 bug + 出题Explain code + debug + questions学→练→考(重练)Learn→Practice→Test (practice-heavy)"这段代码哪里错了?别直接改,先告诉我为什么""Where's the bug? Don't fix directly; tell me why first"
写作Writing批改 + 结构建议Grading + structure advice学→练→问Learn→Practice→Ask"批改我的作文,指出 3 个最弱处""Grade my essay, point out 3 weakest parts"
历史History讲故事 + 时间线Storytelling + timelines学→问→考Learn→Ask→Test"把这段历史讲成一个故事""Tell this history as a story"
经济/商科Economics/Business案例分析 + 建模Case analysis + modeling学→问→练Learn→Ask→Practice"用这个模型分析这个案例""Analyze this case with this model"
法律Law条文讲解 + 案例Statute explanation + cases学→练→考Learn→Practice→Test"这条法律在什么情况下适用?给 3 个例子""When does this law apply? Give 3 examples"
医学Medicine机制讲解 + 鉴别诊断Mechanism + differential diagnosis学→问→考Learn→Ask→Test"这个症状的鉴别诊断有哪些?""What are the differential diagnoses for this symptom?"
艺术/设计Art/Design风格讲解 + 灵感Style explanation + inspiration学→问→练Learn→Ask→Practice"分析这幅画用了什么手法?""What techniques does this painting use?"
心理/认知Psychology/Cognition理论 + 自我觉察Theory + self-awareness学→问→练Learn→Ask→Practice"这个心理学效应在生活里怎么用?""How to apply this psychology effect in life?"
🎯规律:需要"算"的学科(数学、编程、经济)——重"练",让 AI 出题、你手答。需要"记"的学科(外语、法律、医学)——重"考",让 AI 反复测你(检索练习)。需要"讲"的学科(写作、历史、艺术)——重"问",让 AI 批改你的输出。不管哪个学科,"学"(听讲解)都是起点,但只靠"学"永远到不了"会"。Pattern: "calculation" disciplines (math, programming, economics) — "practice"-heavy: AI sets questions, you answer by hand. "memorization" disciplines (language, law, medicine) — "test"-heavy: AI repeatedly quizzes you (retrieval practice). "explanation" disciplines (writing, history, art) — "ask"-heavy: AI grades your output. For any discipline, "learn" (listening) is the start, but "learn" alone never gets you to "can do".
06

30 条万能学习提示词(直接复制用)

30 Universal Learning Prompts (copy and use)

按"学/问/练/考"分类,拿来就用。
Categorized by "learn/ask/practice/test", ready to use.

📖 "学"类(10 条)

📖 "Learn" (10)

##提示词(把 [X] 换成你的主题)Prompt (replace [X] with your topic)
11"像给 12 岁孩子讲一样,解释 [X],用一个生活类比""Explain [X] like to a 12-year-old, with one real-life analogy"
22"用 3 句话概括 [X] 的核心思想""Sum up [X]'s core idea in 3 sentences"
33"给我讲 [X] 的历史:它怎么来的、解决什么问题、谁提出的""Tell me [X]'s history: how it came, what problem it solves, who proposed it"
44"解释 [X],但先别给结论,先带我一步步推导""Explain [X], but don't give the conclusion first; lead me through the steps"
55"用一张表格对比 [X] 和 [Y] 的区别""Compare [X] vs [Y] differences in a table"
66"把 [X] 拆成 3 个最底层的概念,分别讲""Break [X] into 3 most fundamental concepts, explain each"
77"给我 5 个 [X] 的经典案例/例子,从易到难""Give 5 classic examples of [X], easy to hard"
88"用一句话 + 1 个类比 + 1 个反例,解释 [X]""Explain [X] with one sentence + one analogy + one counterexample"
99"[X] 的常见误解有哪些?逐个纠正""What are common misconceptions of [X]? Correct each"
1010"假设我是[身份],从我的角度重新讲一遍 [X]""Assume I'm [identity]; re-explain [X] from my perspective"

❓ "问"类(10 条)

❓ "Ask" (10)

##提示词Prompt
1111"不要直接告诉我答案,用提问引导我自己想出 [X]""Don't tell me the answer; guide me with questions to think out [X]"
1212"这是我对 [X] 的理解:[你的话]。哪里错了?哪里不够深?""Here's my understanding of [X]: [your words]. Where's it wrong? Where's it shallow?"
1313"[X] 有什么反例?什么情况下它不成立?""What are counterexamples to [X]? When does it fail?"
1414"[X] 和我已知的 [Y] 有什么联系?""How is [X] connected to [Y] I already know?"
1515"如果我把 [X] 的一个条件改一下,结论会变吗?""If I change one condition of [X], does the conclusion change?"
1616"[X] 的底层原理,再深一层是什么?""Go one level deeper on [X]'s underlying principle?"
1717"换 3 种方式重新解释 [X](不同类比)""Re-explain [X] in 3 ways (different analogies)"
1818"[X] 在实际中怎么用?给我 3 个场景""How to apply [X] in practice? Give 3 scenarios"
1919"我卡住了:[问题]。别给答案,先问我 3 个问题帮我定位""I'm stuck: [problem]. Don't give the answer; ask me 3 questions to locate it"
2020"用 1 个问题检验我是否真懂 [X]""Use 1 question to test if I truly understand [X]"

✍️ "练"类(5 条)

✍️ "Practice" (5)

##提示词Prompt
2121"出 5 道关于 [X] 的题(3 基础 2 进阶),我答完你批改""Give 5 questions on [X] (3 basic, 2 advanced); after I answer, grade"
2222"给我一道综合题,要求用 [X] 解决一个实际问题""Give one comprehensive problem requiring [X] to solve a real problem"
2323"批改我的答案:[你的答案]。指出错误 + 给出正确思路""Grade my answer: [your answer]. Point out errors + correct approach"
2424"根据我的弱点 [弱点],出 3 道针对性练习""Based on my weakness [weakness], give 3 targeted exercises"
2525"把 [X] 做成一个 10 分钟的动手练习,一步步带我""Turn [X] into a 10-minute hands-on exercise, guide me step by step"

📝 "考"类(5 条)

📝 "Test" (5)

##提示词Prompt
2626"出 10 道 [X] 的测试题(选择题 5 + 简答 3 + 应用 2)""Give 10 test questions on [X] (5 multiple-choice + 3 short-answer + 2 application)"
2727"模拟考试:20 分钟内完成 [X] 的 5 道题,计时并评分""Mock exam: complete 5 questions on [X] in 20 minutes, timed and graded"
2828"根据我上周学的 [X],出 3 道综合题测我是否还记得""Based on last week's [X], give 3 comprehensive questions to test if I still remember"
2929"用 1 个开放性问题考我对 [X] 的深度理解""Test my deep understanding of [X] with 1 open-ended question"
3030"给我一份 [X] 的自测清单:我能否解释/应用/教给别人""Give a self-test checklist for [X]: can I explain/apply/teach it"
07

7 天行动:从"问答案"到"用 AI 学习"

7-Day Action: From "Asking Answers" to "Learning with AI"

每天 30 分钟,7 天养成"用 AI 学习"的习惯。
30 min/day, 7 days to build the "learning with AI" habit.
天Day任务Task做什么What to do检验Verify
11第一次"学"First "learn"选一个你在学的主题,用"费曼讲解"句式让 AI 讲一遍Pick a topic you're studying; use the "Feynman" pattern to have AI explain it你能用 1 句话复述它You can restate it in 1 sentence
22第一次"问"First "ask"对昨天的主题,追问 3 轮(为什么→换说法→反例)Follow up 3 rounds on yesterday's topic (why→rephrase→counterexample)你能说出它的边界/例外You can state its boundary/exception
33第一次"练"First "practice"让 AI 出 5 道题,你手写答案,再让它批改AI gives 5 questions; you hand-answer; it grades你至少答对 3 道You get at least 3 right
44第一次"考"First "test"过 1 天,让 AI 出 3 道综合题测你(间隔重复)1 day later, AI gives 3 comprehensive questions (spaced repetition)你能不看笔记答出 2 道You can answer 2 without notes
55完整循环Full loop选一个新主题,完整走一遍"学→问→练→考"(30 分钟)New topic, full "learn→ask→practice→test" (30 min)你能向别人讲清它You can explain it to someone
66针对弱点Target weaknesses把前 5 天答错的题整理,让 AI 出 3 道针对性练习Compile wrong answers from days 1–5; AI gives 3 targeted exercises弱点题全对Weakness questions all right
77复盘+固化Review + solidify写 3 句话:哪 3 个提示词最有用?固化成你的"模板"Write 3 sentences: which 3 prompts were most useful? Make them your "templates"你有 3 个常用模板You have 3 go-to templates
🏁7 天后的标志:你不再"问答案",而是"用 AI 学习"——会追问、会手答、会自测。你有了 3 个自己的"提示词模板",每个新主题都能套上。第 8 天起,把它变成日常:每天选 1 个主题,走一遍循环(30 分钟)。坚持 30 天,你会发现自己"会的东西"明显变多——不是学得更多,是"学进去"了更多。Marks after 7 days: you no longer "ask answers" but "learn with AI" — you follow up, hand-answer, self-test. You have 3 personal "prompt templates", ready for any new topic. From day 8, make it daily: pick 1 topic, run the loop (30 min). After 30 days, you'll find you "know" noticeably more — not studying more, but "absorbing" more.
08

AI 老师的 6 个边界:什么时候别信它

6 Boundaries of the AI Tutor: When Not to Trust It

知道它"哪里会出错",才不会被它带沟里。
Know "where it goes wrong" so it can't lead you astray.
边界BoundaryAI 会怎样How AI behaves你怎么防How to guard
幻觉Hallucination编造不存在的论文、数据、条文,且说得自信Invents nonexistent papers, data, statutes, confidently重要事实"必须"自己查证(搜原文/看教材)Verify important facts yourself (find original/textbook)
时效性Recency训练有截止日期,新事件/新数据可能不知道Training has a cutoff; may not know new events/data问"2025 年以后"的事,先确认它的知识截止For "post-2025" questions, first confirm its knowledge cutoff
无记忆No memory(默认)不记得上次对话,无法跟踪你的进度(By default) doesn't recall previous chats, can't track progress用它的"记忆"功能,或每次开头复述你的进度Use its "memory" feature, or restate progress each time
无真正评估No true assessment说"你懂了"可能只是顺着你说Saying "you got it" may just be agreeing with you让它"出更难的题"检验,而不是问"我懂了吗"Have it "give harder questions" to test, not ask "do I get it?"
安全/敏感Safety/sensitive医学、法律、金融建议可能"错得危险"Medical/legal/financial advice can be "dangerously" wrong这些领域 AI 只当"参考",决策找专业人士In these fields, treat AI as "reference" only; consult pros for decisions
依赖Dependency越用越懒,离开 AI 就不会想了The more you use it, the lazier; without AI you can't think先自己想 10 分钟,再问 AI("先思考后求助")Think 10 min yourself first, then ask AI ("think before asking")
⚠️三条铁律:① 重要事实"自己查证"——AI 说"有篇 2023 年的研究",你去搜它到底存不存在;② "先思考后求助"——自己先想 10 分钟,带着自己的思路问,而不是直接要答案;③ 敏感领域(医/法/财)"只参考不决策"——AI 给你方向,专业人士给结论。记住:AI 是"学"的工具,不是"答"的工具。Three iron rules: ① "Verify" important facts yourself — if AI says "there's a 2023 study", go check if it exists; ② "Think before asking" — think 10 min yourself first, then ask with your own thinking, not just demand the answer; ③ Sensitive fields (medical/legal/financial) "reference only, no decisions" — AI gives direction, pros give conclusions. Remember: AI is a "learning" tool, not an "answering" tool.
09

30 天计划:从"问答案"到"独立学习"

30-Day Plan: From "Asking Answers" to "Independent Learning"

每周升级一级,30 天把"AI 学习"变成能力。
Upgrade one level weekly; 30 days to turn "AI learning" into a skill.
周Week主题Theme做什么What to do标志Mark
11建立循环Build the loop每天 1 个主题,完整走"学→问→练→考"(30 分钟)1 topic daily, full "learn→ask→practice→test" (30 min)7 天 7 个主题,都走完循环7 topics in 7 days, all looped
22加深追问Deepen follow-ups每个主题追问 5 轮以上,用"10 个追问句式"Follow up 5+ rounds per topic, use the "10 follow-up patterns"能讲清 1 个主题的"为什么+边界"Can explain one topic's "why + boundaries"
33间隔复习Spaced review第 1/3/7 天让 AI 出综合题测你(间隔重复)Days 1/3/7 have AI give comprehensive tests (spaced repetition)第 7 天还能答对第 1 天的题Day 7 can still answer day 1's questions
44教别人Teach others把学过的 3 个主题,用"费曼法"讲给 AI(让它当学生)Teach 3 learned topics to AI (let it be the student) via FeynmanAI 能指出你"没讲清"的地方AI can point out where you "didn't explain"
🏁30 天后的标志:① 有 3–5 个自己的"提示词模板";② 能独立走完"学→问→练→考"循环(不靠提醒);③ 会用"间隔重复"复习(1/3/7 天);④ 能"教" AI(费曼法反向验证);⑤ 不再"问答案",而是"问为什么"。这 5 条达成了,"AI 学习"就成了你的能力——换任何学科、任何主题,都能套上这套循环。Marks after 30 days: ① 3–5 personal "prompt templates"; ② can independently run the "learn→ask→practice→test" loop (no reminders); ③ uses "spaced repetition" (days 1/3/7); ④ can "teach" AI (Feynman reverse-verification); ⑤ no longer "asks answers" but "asks why". Hit all 5 and "AI learning" becomes your skill — any discipline, any topic, this loop applies.
10

5 个真实改变案例(用 AI 学习的人)

5 Real Change Cases (people who learned with AI)

不是"AI 多神奇",是"怎么用"的差别。
Not "how magical AI is", but the difference in "how to use".
1案例 1:考研数学的"苏格拉底"用法。小李,考研数学基础弱,从"看视频课"改为"用 AI 当苏格拉底"——每道难题先自己想 10 分钟,再让 AI 用提问引导他找思路(不直接给答案)。3 个月后,他的数学从 60 分提到 85 分。关键不是"AI 讲得好",是"他被迫自己想了"——检索和推导训练了大脑,而不是"听懂了"的错觉。Case 1: The "Socratic" use for grad-exam math. Xiao Li, weak in grad-exam math, switched from "watching video lectures" to "AI as Socrates" — for each hard problem, think 10 min first, then have AI guide his reasoning with questions (no direct answer). 3 months later, his math went from 60 to 85. The key wasn't "AI explains well" but "he was forced to think" — retrieval and derivation trained the brain, not the illusion of "getting it".
2案例 2:英语对话的"陪练"用法。小王,英语听力口语差,没外教。每天 20 分钟让 AI 当"英语陪练"——用英语对话、纠正他的错误、给更地道的说法。90 天后,他能用英语进行 10 分钟日常对话。关键:AI 有无限耐心(真人外教会累),且能"换难度"(从简单到复杂),让他在"舒适区边缘"练习。Case 2: The "sparring" use for English conversation. Xiao Wang, weak in English listening/speaking, no tutor. 20 min daily with AI as "English sparring partner" — converse in English, correct his errors, give more natural phrasing. After 90 days, he can hold a 10-min daily English conversation. Key: AI has infinite patience (real tutors get tired) and can "adjust difficulty" (simple to complex), letting him practice at the "edge of comfort".
3案例 3:编程的"调试伙伴"用法。小陈,转行学编程,遇到 bug 就卡住。改为"让 AI 当调试伙伴"——先自己看 10 分钟,再让 AI 问 3 个问题帮他定位(而不是直接改)。半年后,他独立完成了一个小项目。关键:他"学会了调试思维"(定位→假设→验证),而不是"依赖 AI 改代码"。Case 3: The "debugging partner" use for programming. Xiao Chen, switching to programming, stuck on every bug. Switched to "AI as debugging partner" — look 10 min first, then have AI ask 3 questions to locate it (not fix directly). Half a year later, he independently completed a small project. Key: he "learned debugging thinking" (locate→hypothesis→verify), not "depended on AI to fix code".
4案例 4:历史的"故事 + 时间线"用法。小周,历史记不住。改为"让 AI 把历史讲成故事 + 画时间线",再追问"为什么会这样"。1 个月后,历史从 70 分提到 88 分。关键:AI 能"换叙事方式"(故事、时间线、因果链),比教材的"平铺直叙"更容易形成记忆网络。Case 4: The "story + timeline" use for history. Xiao Zhou, couldn't remember history. Switched to "have AI tell history as stories + draw timelines", then follow up "why did it happen this way". 1 month later, history went from 70 to 88. Key: AI can "change narrative style" (story, timeline, causal chain), easier to form memory networks than textbooks' "flat narration".
5案例 5:写作的"批改循环"用法。小刘,写作总被退回。改为"写→让 AI 批改(指出 3 个最弱处)→改→再批改"的循环,每周 3 篇。半年后,他的文章质量明显提升,被杂志采用。关键:AI 提供"即时 + 具体"的反馈(真人老师一周才批一次),高频反馈加速了改进。Case 5: The "grading loop" use for writing. Xiao Liu, writing always rejected. Switched to a "write → AI grades (points out 3 weakest) → revise → re-grade" loop, 3 pieces weekly. Half a year later, his article quality clearly improved, adopted by a magazine. Key: AI gives "instant + specific" feedback (real teachers grade once a week); high-frequency feedback accelerated improvement.
🎯共同点:5 个案例都不是"AI 多聪明",是"怎么用"——① 先想后问(不抄答案);② 让它出题/批改(动手);③ 高频反馈(每周多次);④ 换方式(故事/对话/调试)。AI 是"放大器"——它放大你"已有的努力",不创造你"没付出的能力"。你用得越"主动",它越"有效"。Common thread: none of the 5 cases is "AI is so smart", it's "how to use" — ① think before asking (don't copy answers); ② let it quiz/grade (hands-on); ③ high-frequency feedback (multiple per week); ④ change style (story/conversation/debugging). AI is an "amplifier" — it amplifies "effort you already put in", doesn't create "ability you didn't pay for". The more "proactively" you use it, the more "effective" it is.
11

30 条实战问答(把 AI 用对)

30 Practical Q&A (use AI right)

覆盖"怎么用、怎么问、怎么不依赖"。
Covers "how to use, how to ask, how to not depend".
QQAA
Q1:AI 说"我懂了",可信吗?Q1: AI says "I get it" — trustworthy?不可全信。它可能顺着你说。正确做法:让它"出更难的题"检验你,而不是问"我懂了吗"。Not fully. It may agree with you. Correct: have it "give harder questions" to test you, not ask "do I get it?".
Q2:让 AI 讲解,我该记笔记吗?Q2: When AI explains, should I take notes?必须。只"听"不"记",24 小时后忘 80%。用自己的话复述 + 写 3 个关键词。Must. Just "listening" without "noting" forgets 80% in 24h. Restate in your words + 3 keywords.
Q3:AI 的答案错了怎么办?Q3: AI's answer is wrong, what to do?先指出"哪里错了 + 正确答案是什么",让它"纠正 + 解释为什么错"。这样它"学会"你的标准,后续更准。First point out "where wrong + correct answer", have it "correct + explain why wrong". This makes it "learn" your standard, more accurate later.
Q4:AI 能帮我"作弊"吗?Q4: Can AI help me "cheat"?技术上能,但会毁了你。作业/考试用它,短期省事,长期废掉思考力。用 AI "学",不用它"代做"。Technically yes, but it'll ruin you. Using it for homework/exams saves short-term, destroys thinking long-term. Use AI to "learn", not to "do for you".
Q5:每次对话都要"重新介绍"吗?Q5: Do I "re-introduce" each conversation?不用。开启它的"记忆"功能,或开头一句话:"我是[身份],正在学[X],上次的进度是[Y]"。No. Enable its "memory" feature, or one line at start: "I'm [identity], learning [X], last progress was [Y]".
Q6:AI 讲得太深/太浅怎么办?Q6: AI explains too deep/too shallow?直接说"太深了,从[水平]重新讲"或"太浅了,再深一层"。它"能调",关键是你要"说"。Just say "too deep, re-explain from [level]" or "too shallow, go one level deeper". It "can adjust"; the key is you must "say".
Q7:AI 出的题和真考题一样吗?Q7: Are AI's questions like real exams?风格可以像,但"考点分布"未必准。考前 2 周,用 AI 出"历年真题风格"的题(给它真题当参考)。Style can be similar, but "point distribution" may not be accurate. 2 weeks before exam, have AI make "past-paper style" questions (give it past papers as reference).
Q8:AI 能"因材施教"吗?Q8: Can AI "teach to the student"?部分能——你告诉它水平/弱点,它调整。但"主动诊断"不行,需要你"反馈"(答错、卡住的地方)。Partially — tell it level/weakness, it adjusts. But "proactive diagnosis" no; you must "feed back" (wrong answers, stuck spots).
Q9:学习时 AI 打断了怎么办?Q9: AI interrupts while studying?把 AI 当"课后工具",不是"学习伙伴"。学习时"专注",结束后"问 AI"(复习 + 答疑)。Treat AI as "after-class tool", not "learning partner". "Focus" during study, "ask AI" after (review + Q&A).
Q10:AI 能帮"理解"但"不帮记忆"?Q10: AI helps "understand" but not "memorize"?对。理解靠"讲解",记忆靠"检索"(反复回忆)。让 AI "出题考你"才是记忆的关键(间隔重复)。Yes. Understanding needs "explanation", memory needs "retrieval" (repeated recall). Having AI "quiz you" is the key to memory (spaced repetition).
Q11:AI 的"语气"会影响我学习吗?Q11: AI's "tone" affects my learning?会。用"鼓励 + 具体"的语气("你答对了 3 道,第 4 道卡在这里"),不用"评判"("你不行")。让 AI 当"教练",不是"考官"。Yes. Use "encouraging + specific" tone ("you got 3 right, stuck on #4"), not "judging" ("you can't"). Have AI be a "coach", not an "examiner".
Q12:AI 能"监督"我学习吗?Q12: Can AI "supervise" my learning?部分能——你告诉它"我计划今天学 X",它问"你学完了吗"。但"主动监督"不行,需要你"主动汇报"。Partially — tell it "I plan to learn X today", it asks "did you finish?". But "proactive supervision" no; you must "report proactively".
Q13:AI 讲解"太长"了怎么办?Q13: AI's explanation "too long"?开头加"用 5 句话讲"或"用 200 字讲"。先要"精简版",需要再"展开"。别一上来就要"全文"。Add at start "in 5 sentences" or "in 200 words". Ask for "condensed version" first, "expand" if needed. Don't ask for "full text" upfront.
Q14:AI 能"教"我"怎么学"吗?Q14: Can AI "teach" me "how to learn"?能——它知道"学习科学"(间隔重复、主动回忆、检索练习)。让它"给你一份学习计划 + 检验方法",比"只讲知识"更有价值。Yes — it knows "learning science" (spaced repetition, active recall, retrieval practice). Have it "give you a study plan + verification method", more valuable than "just teaching knowledge".
Q15:AI 的"例子"能用吗?Q15: Can I use AI's "examples"?能,但要"查"。AI 的例子可能"编"(不存在的论文/数据)。重要的"自己验证"(搜原文/看教材)。Yes, but "check" them. AI's examples may be "invented" (nonexistent papers/data). Important ones "verify yourself" (find original/textbook).
Q16:AI 能"批改"作文吗?Q16: Can AI "grade" essays?能,但要"具体"。让它"指出 3 个最弱处 + 给出改法",而不是"打分"。"具体反馈"比"分数"有用。Yes, but "specifically". Have it "point out 3 weakest parts + give fixes", not "give a score". "Specific feedback" is more useful than "a grade".
Q17:AI 能"陪练"外语吗?Q17: Can AI "spar" with me on a foreign language?能——用"对话 + 纠正"模式。它"无限耐心 + 能调难度",比真人陪练方便。每天 20 分钟,3 个月见效。Yes — use "conversation + correction" mode. It's "infinitely patient + adjustable difficulty", more convenient than a real sparring partner. 20 min daily, visible in 3 months.
Q18:AI 能"帮我读论文"吗?Q18: Can AI "help me read papers"?能——先让它"摘要 + 关键论点",再"追问 + 质疑"。但"数据/结论"要"回原文验证"(AI 可能"概括错")。Yes — first have it "summarize + key arguments", then "follow up + question". But "data/conclusions" must "verify against the original" (AI may "summarize wrong").
Q19:AI 能"帮我做研究"吗?Q19: Can AI "help me research"?能当"助手"(找方向、列提纲、查文献),但当不了"研究者"。"判断 + 创新"要靠你自己。AI 给你"线索",你"验证 + 推进"。Can be an "assistant" (find direction, outline, check literature), but not a "researcher". "Judgment + innovation" are yours. AI gives you "leads", you "verify + advance".
Q20:AI 能"帮我写代码"吗?Q20: Can AI "help me write code"?能,但要"理解"。让它"解释每行",你"看懂再跑"。别"复制粘贴"——"不懂的代码"是"技术债"。Yes, but "understand" it. Have it "explain each line", you "run after understanding". Don't "copy-paste" — "code you don't understand" is "technical debt".
Q21:AI 能"帮我做决策"吗?Q21: Can AI "help me decide"?能给"选项 + 利弊",但"决定"是你的。AI 给"信息",你给"判断"。别让 AI "替你做决定"(尤其重大决策)。Can give "options + pros/cons", but the "decision" is yours. AI gives "information", you give "judgment". Don't let AI "decide for you" (especially major decisions).
Q22:AI 能"帮我学编程"吗?Q22: Can AI "help me learn programming"?能——"讲代码 + 找 bug + 出题"。但"动手写"不能省——让 AI 出题,你"手答",再让它"批改"。"只看不写"学不会。Yes — "explain code + debug + quiz". But "hands-on writing" can't be skipped — AI sets questions, you "hand-answer", it "grades". "Only reading" doesn't learn.
Q23:AI 能"帮我复习"吗?Q23: Can AI "help me review"?能——"出综合题 + 间隔重复"(第 1/3/7 天)。比"重新看笔记"有效 2–3 倍(检索练习)。Yes — "comprehensive questions + spaced repetition" (days 1/3/7). 2–3× more effective than "rereading notes" (retrieval practice).
Q24:AI 能"帮我备考"吗?Q24: Can AI "help me prep for exams"?能——"模拟考 + 弱点分析 + 针对性练习"。考前 2 周,每天 30 分钟"AI 模拟考",比"看 3 遍书"有效。Yes — "mock exams + weakness analysis + targeted practice". 2 weeks before, 30 min daily "AI mock exam" beats "reading the book 3 times".
Q25:AI 能"帮我学乐器"吗?Q25: Can AI "help me learn an instrument"?能当"理论老师"(乐理、指法、谱子),但"动手练"不能省。"理论 + 练习"结合,AI 给你"每天练什么 + 怎么检验"。Can be a "theory teacher" (music theory, fingerings, scores), but "hands-on practice" can't be skipped. "Theory + practice" combined — AI tells you "what to practice daily + how to verify".
Q26:AI 能"帮我学设计"吗?Q26: Can AI "help me learn design"?能当"审美老师"(风格分析、配色原理),但"动手做"不能省。让它"分析经典设计 + 给你 3 个练法"。Can be an "aesthetics teacher" (style analysis, color principles), but "hands-on making" can't be skipped. Have it "analyze classic designs + give 3 practice methods".
Q27:AI 能"帮我学统计"吗?Q27: Can AI "help me learn statistics"?能——"讲原理 + 出应用题 + 讲常见误区"。统计的"坑"多(相关≠因果),让 AI "专门讲误区"比"讲公式"更有用。Yes — "explain principles + application questions + common pitfalls". Statistics has many "traps" (correlation ≠ causation); having AI "focus on pitfalls" is more useful than "formulas".
Q28:AI 能"帮我学心理学"吗?Q28: Can AI "help me learn psychology"?能——"讲理论 + 给生活例子 + 自测"。但"自我分析"要"谨慎"——AI 不能"诊断","症状"要"找专业人士"。Yes — "teach theory + real-life examples + self-test". But "self-analysis" requires "caution" — AI can't "diagnose"; "symptoms" go to "professionals".
Q29:AI 能"帮我学历史"吗?Q29: Can AI "help me learn history"?能——"讲成故事 + 时间线 + 因果链"。但"史料"要"查"——AI 的"细节"可能"编"(不存在的引文/日期),"关键事实"自己验证。Yes — "tell as stories + timelines + causal chains". But "historical sources" must "check" — AI's "details" may be "invented" (nonexistent quotes/dates); "key facts" verify yourself.
Q30:AI 能"帮我终身学习"吗?Q30: Can AI "help me learn lifelong"?能——它是"免费的 + 全领域的 + 不打烊"老师。关键是"持续用"——每天 30 分钟"学→问→练→考"循环,30 年后你的"知识网络"会远超"只靠教材"的人。Yes — it's a "free + all-domain + never-closed" teacher. The key is "persistent use" — 30 min daily "learn→ask→practice→test" loop; 30 years later your "knowledge network" will far exceed someone "relying only on textbooks".
12

AI 工具选型:不同任务用不同"老师"

AI Tool Selection: Different "Teachers" for Different Tasks

别一个工具打天下——按任务选。
Don't use one tool for everything — pick by task.
任务类型Task type推荐工具Recommended理由Why
讲解复杂概念Explain complex conceptsDeepSeek / 豆包 / KimiDeepSeek / Doubao / Kimi中文理解好,讲解清晰Good Chinese understanding, clear explanations
英文对话/陪练English conversation/sparringChatGPT / GeminiChatGPT / Gemini英文自然,纠正地道Natural English, idiomatic corrections
代码/编程Code/programmingDeepSeek / ChatGPT / ClaudeDeepSeek / ChatGPT / Claude代码能力强,能 debugStrong code ability, can debug
长文档/论文阅读Long documents/paper readingKimi / DeepSeek(长上下文)Kimi / DeepSeek (long context)能读 200 页文档,给摘要Can read 200-page docs, give summaries
写作/批改Writing/gradingChatGPT / DeepSeekChatGPT / DeepSeek给具体反馈,不空泛Specific feedback, not vague
数学/逻辑Math/logicDeepSeek / ChatGPTDeepSeek / ChatGPT分步推导清晰Clear step-by-step derivation
多语言翻译Multilingual translationDeepSeek / GeminiDeepSeek / Gemini小语种覆盖广Wide coverage of minor languages
创意/头脑风暴Creative/brainstormingChatGPT / ClaudeChatGPT / Claude发散思维强,点子多Strong divergent thinking, many ideas
教学/做课件Teaching/making coursewareKimi / DeepSeekKimi / DeepSeek能生成结构化课件Can generate structured courseware
本地/隐私敏感Local/privacy-sensitive本地模型(Llama/Qwen)Local models (Llama/Qwen)数据不出本地,最安全Data stays local, safest
💡选型原则:① 按"任务"选,不按"品牌"选——同一个工具,不同任务效果不同;② "免费够用"——大多数学习用免费档就够,不用充会员;③ "多备一个"——主力工具挂了,有个备份(网络/服务问题);④ "隐私优先"——敏感内容(医疗/法律/财务)用本地模型或最严格的隐私政策。记住:工具是"放大器",你的"提问方式"才是"质量"的决定因素。Selection principle: ① Pick by "task", not "brand" — same tool, different tasks, different results; ② "Free is enough" — most learning uses the free tier, no need to pay; ③ "Have a backup" — main tool down, have a backup (network/service issues); ④ "Privacy first" — sensitive content (medical/legal/financial) uses local models or strictest privacy policy. Remember: the tool is an "amplifier"; your "questioning style" is what determines "quality".
13

研究依据 + 12 条术语表

Research Basis + 12-Term Glossary

AI 学习的科学依据与关键术语。
Scientific basis of AI learning + key terms.

"用 AI 学习"不是跟风,背后有 4 条被反复验证的学习科学规律:① 检索练习效应(Karpicke & Roediger, 2008)——回忆比重复阅读有效 2–3 倍,AI 出题→你手答正是检索练习;② 生成效应(Slamecka & Soloman, 1978)——自己"生成"的内容记得更牢,"先想后问"比"直接要答案"有效;③ 间隔效应(Ebbinghaus 1885; Cepeda 2006 元分析)——分散复习比集中学习记忆保持率高 50%+,AI 的"第 1/3/7 天出题"实现间隔;④ 反馈效应(Shute, 2008)——即时、具体的反馈加速学习,AI 的"即时批改"提供了真人老师难以达到的高频反馈。

"Learning with AI" isn't following a trend; it rests on 4 repeatedly-verified learning-science laws: ① retrieval practice effect (Karpicke & Roediger, 2008) — recall is 2–3× more effective than rereading; AI-set questions → your hand-answers are exactly retrieval practice; ② generation effect (Slamecka & Soloman, 1978) — content you "generate" yourself is remembered better; "think before asking" beats "just demand the answer"; ③ spacing effect (Ebbinghaus 1885; Cepeda 2006 meta-analysis) — distributed review retains 50%+ better than massed; AI's "day 1/3/7 questions" implements spacing; ④ feedback effect (Shute, 2008) — instant, specific feedback accelerates learning; AI's "instant grading" provides the high-frequency feedback real teachers struggle to give.

术语Term定义Definition
大语言模型 (LLM)Large Language Model (LLM)基于海量文本训练的 AI,能理解并生成语言(ChatGPT/DeepSeek/豆包)AI trained on massive text; understands and generates language (ChatGPT/DeepSeek/Doubao)
幻觉 (Hallucination)HallucinationAI 自信地编造不存在的事实(论文/数据/引文)AI confidently invents nonexistent facts (papers/data/quotes)
提示词 (Prompt)Prompt你对 AI 的指令;写法决定输出质量Your instruction to the AI; its form determines output quality
上下文窗口Context windowAI 一次能"记住"的信息量(几千到几十万词)How much info the AI can "hold" at once (thousands to hundreds of thousands of words)
检索练习Retrieval practice主动回忆(做题、复述)比重复阅读更有效的学习方法Actively recalling (quizzing, restating) — more effective than rereading
间隔重复Spaced repetition分散在多个时间点复习(1/3/7 天),记忆保持率更高Reviewing at spread-out times (1/3/7 days); higher retention
苏格拉底式提问Socratic questioning通过连续提问引导对方自己思考,而非直接给答案Guiding the other to think via continuous questions, not giving answers
费曼学习法Feynman technique用最简单的语言把概念讲给"外行",暴露理解漏洞Explaining a concept in the simplest words to a "layperson", exposing gaps
生成效应Generation effect自己"产出"的内容(答案/解释)比"看到"的记得更牢Content you "produce" (answers/explanations) is remembered better than "seen" content
即时反馈Instant feedback答案后立即给出的具体纠正;加速学习的关键Specific correction given right after the answer; key to accelerating learning
知识截止 (Cutoff)Knowledge cutoffAI 训练数据的截止日期;之后的事件它可能不知道The cutoff date of the AI's training data; events after it may be unknown
依赖陷阱Dependency trap过度依赖 AI 导致独立思考能力退化的风险The risk that over-reliance on AI degrades independent thinking
📊66%的学生用 AI 时"先要答案"(2025 年调查),只有 1/3 用"引导式提问"——后者学习效果显著更高。of students "ask for the answer first" when using AI (2025 survey); only 1/3 use "guided questioning" — the latter shows significantly better learning.Common Sense Media 2025 / OECD Learning in the AI Era
14

10 个误区 vs 真相

10 Myths vs Truths

把常见误解一次说清。
Clear up common misconceptions in one pass.
误区Myth真相Truth
"AI 的答案都对""AI's answers are all right"会幻觉。重要事实必须自己查证。It hallucinates. Verify important facts yourself.
"用 AI 学习 = 偷懒""Learning with AI = laziness"用对了是"加速器"(高频反馈+无限耐心);用错了(抄答案)才是偷懒。Used right it's an "accelerator" (high-frequency feedback + infinite patience); used wrong (copying) is laziness.
"AI 会让我变笨""AI will make me dumb""先想后问"会让你更会思考;"直接要答案"才会退步。方法决定结果。"Think before asking" makes you a better thinker; "demanding answers" makes you decline. Method decides the outcome.
"AI 能替我学习""AI can study for me"不能。它能"讲解/出题/批改",但"理解+记忆"只能发生在你自己脑子里。No. It can "explain/quiz/grade", but "understanding + memory" only happen in your own head.
"免费的 AI 不够好""Free AI isn't good enough"对 90% 的学习场景,免费档足够。付费主要买"速度+更强模型",不是"能不能学会"。For 90% of learning scenarios, the free tier suffices. Paying mainly buys "speed + stronger models", not "whether you can learn".
"AI 比老师好""AI is better than teachers"各有所长:AI 强在"24 小时+无限耐心+即时反馈";真人老师强在"激励+情感+因材施教"。最佳是"AI + 真人"组合。Each has strengths: AI wins on "24/7 + infinite patience + instant feedback"; real teachers win on "motivation + emotion + teaching to the student". Best is the "AI + human" combo.
"一次问到位就行""One well-asked question is enough"第一轮回答只值 40 分,追问才值 60 分。理解来自"追问链"。The first answer is worth 40%; follow-ups are worth 60%. Understanding comes from the "follow-up chain".
"AI 出的题和真题一样""AI's questions are like real exams"风格可像,考点分布未必准。考前用真题"校准"(给 AI 真题当参考)。Style can be similar, but point distribution may be off. Before exams, "calibrate" with real papers (give AI real papers as reference).
"AI 能记住我的进度""AI remembers my progress"默认不记得。要么开"记忆"功能,要么每次开头复述进度。By default it doesn't. Either enable the "memory" feature or restate progress at the start of each chat.
"学会用 AI 就不用学基础""Mastering AI means no need to learn basics"相反——基础越扎实,越会用 AI(能判断它对错、能追问到深处)。AI 放大基础,不替代基础。The opposite — the stronger your foundation, the better you use AI (can judge right/wrong, can ask deep). AI amplifies the foundation; it doesn't replace it.
15

30 条资源(工具/课程/研究)

30 Resources (Tools / Courses / Research)

全部真实可查,附用途。
All verifiable, with use-cases.
ChatGPT免费综合最强,英文对话/写作/代码All-round best: English conversation, writing, code DeepSeek免费中文强,数学/代码好,性价比高Strong Chinese, good math/code, high value 豆包 (Doubao)免费字节出品,中文口语化好By ByteDance, natural colloquial Chinese Kimi免费超长上下文(200 页文档阅读)Ultra-long context (reads 200-page docs) Gemini免费谷歌出品,多语言/搜索整合By Google, multilingual, search-integrated Claude免费长文写作/创意/代码质量高High-quality long-form writing, creativity, code Hugging Face免费开源模型库(本地部署 Qwen/Llama)Open-source model hub (local Qwen/Llama deployment) Khan Academy免费K 12 全科 + 练习(配合 AI 讲解)K–12 all subjects + practice (pairs with AI explanations) Coursera部分免费大学课程(可旁听免费)University courses (free audit available) edX部分免费哈佛/MIT 课程(可旁听免费)Harvard/MIT courses (free audit available) MIT OCW免费MIT 全课程公开(讲义+视频+习题)All MIT courses open (notes + videos + exercises) freeCodeCamp免费编程零基础到项目(配合 AI 批改)Coding from zero to projects (pairs with AI grading) Duolingo免费语言入门(AI 陪练进阶)Language basics (AI sparring for advanced) BBC Learning English免费英语新闻+听力(真实语料)English news + listening (real corpus) 网易公开课免费名校课程中文字幕Top-university courses with Chinese subtitles 哔哩哔哩免费中文知识区(配合 AI 消化)Chinese knowledge zone (digest with AI) Anki免费间隔重复记忆卡(AI 帮你生成卡)Spaced-repetition flashcards (AI generates them) Notion免费AI 对话记录/学习卡片库Store AI chats / study cards Obsidian免费知识库 + AI 插件(第二大脑)Knowledge base + AI plugins (second brain) Overleaf免费LaTeX 论文写作(AI 辅助)LaTeX paper writing (AI-assisted) Wolfram Alpha部分免费数学/科学计算(AI 结果的验证器)Math/science compute (verifier of AI results) arXiv免费论文库(验证 AI 引用的论文是否存在)Paper archive (verify AI-cited papers exist) DOI.org免费学术文献验证(防 AI 幻觉)Academic reference verification (guards vs AI hallucination) OECD AI 教育专题免费政策与研究(AI 与教育的关系)Policies & research (AI × education) UNESCO 数字教育免费全球数字教育框架Global digital-education framework Common Sense Media免费青少年 AI 使用调查Surveys on teens' AI use Prompting Guide免费提示词技巧教程Prompt-technique tutorials 微软提示工程指南免费系统性提示工程方法Systematic prompt-engineering method ResearchGate免费论文全文(AI 引用的第二验证源)Full papers (second verification for AI citations) Google Scholar免费学术搜索(验证 AI 的"研究"是否存在)Academic search (verify AI's "research" exists)
16

10 个真实学习场景:完整 AI 方案(直接套用)

10 Real Learning Scenarios: Complete AI Plans (copy and apply)

每个场景给"目标→提示词序列→检验标准"。
Each scenario gets "goal → prompt sequence → verification standard".
场景Scenario目标Goal提示词序列(按顺序问)Prompt sequence (ask in order)检验标准Verify
备考一门新科目Prep for a new subject4 周从 0 到能考试0 to exam-ready in 4 weeks①"给我 4 周学习计划,每天 1h,每周检验点" ②每天"学→问→练→考" ③第 7/14/28 天"出综合卷" ④考前 3 天"按真题风格出 3 套"①"Give a 4-week plan, 1h/day, weekly checkpoints" ②daily "learn→ask→practice→test" ③day 7/14/28 "comprehensive paper" ④3 days before "3 past-paper-style sets"模拟卷 ≥ 目标分 80%Mock papers ≥ 80% of target
啃一本厚书Chew through a thick book2 周精读 + 内化2-week intensive read + internalize①"这本书的核心论点 3 条" ②每章读完"我复述一遍,你找漏洞" ③"出 10 题测我这章" ④全书"费曼式:我讲给你听" ⑤"出综合题"①"3 core claims of this book" ②after each chapter "I restate, you find gaps" ③"quiz me 10 questions on this chapter" ④whole book "Feynman: I teach you" ⑤"comprehensive questions"能向别人讲清全书框架Can explain the whole book's framework
学一门外语Learn a foreign language3 个月日常对话Daily conversation in 3 months①"每天 20 分钟对话课,从 [水平] 开始" ②"纠正我的每个错误 + 给地道说法" ③每周"出 10 题测语法" ④"用 100 个高频词造对话" ⑤每月"模拟 10 分钟对话"①"Daily 20-min conversation class from [level]" ②"correct every error + idiomatic phrasing" ③weekly "10 grammar questions" ④"dialogues with 100 high-freq words" ⑤monthly "10-min mock conversation"能进行 10 分钟无准备对话10-min unscripted conversation
从零学编程Learn programming from zero3 个月做出小项目Small project in 3 months①"给我 3 个月 Python 计划,每周 1 项目" ②每段代码"解释每行 + 我重写成自己的" ③"出 5 道动手题 + 批改" ④"帮我 debug:先问我 3 个问题" ⑤最后"带我做一个完整项目"①"3-month Python plan, 1 project/week" ②each code "explain every line + I rewrite" ③"5 hands-on questions + grade" ④"help me debug: ask me 3 questions first" ⑤finally "guide a full project"独立完成一个小项目Independently finish a small project
读一批论文Read a batch of papers1 周消化 10 篇Digest 10 papers in 1 week①每篇"摘要 3 句 + 方法 + 结论 + 局限" ②"这 10 篇的关系图" ③"互相矛盾的点" ④"出 10 题测我" ⑤"帮我写综述提纲"①each "3-sentence summary + method + conclusion + limits" ②"relationship map of the 10" ③"contradictions between them" ④"quiz me 10" ⑤"outline a review"能写出 500 字综述Can write a 500-word review
提升写作水平Improve writing2 个月明显提升Visible improvement in 2 months①"分析 3 篇范文的结构" ②每篇作文"指出 3 个最弱处 + 改法" ③"给我 10 个提升技巧" ④"批改 + 对比修改前后" ⑤每周"出一篇限时写作"①"analyze structure of 3 model essays" ②each "3 weakest + fixes" ③"10 improvement techniques" ④"grade + compare before/after" ⑤weekly "timed writing"同类文章评分提升 2 档Same-type essays score 2 levels higher
学一门乐器Learn an instrument3 个月弹 5 首曲子Play 5 songs in 3 months①"每天 30 分钟练习计划(乐理+指法+曲子)" ②"解释这个谱子的手法" ③"我弹的描述,你判断哪里错" ④"给我 5 首从易到难的曲单" ⑤每月"评估进度"①"Daily 30-min plan (theory + fingerings + songs)" ②"explain this score's technique" ③"from my description, judge what's wrong" ④"5 songs easy to hard" ⑤monthly "assess progress"能完整弹 5 首(含乐理理解)Play 5 songs fully (with theory)
学统计学Learn statistics1 个月能用 SPSSUse SPSS in 1 month①"给我 4 周统计计划" ②每个概念"公式 + 例子 + 常见误区" ③"出应用题 + 批改" ④"用真实数据带我跑一遍" ⑤"出 20 题综合卷"①"4-week statistics plan" ②each concept "formula + example + pitfalls" ③"application questions + grade" ④"walk me through real data" ⑤"20-question paper"能独立分析一组数据Independently analyze a dataset
准备面试(求职)Prepare for job interviews2 周通过面试Pass interviews in 2 weeks①"分析这个 JD 的核心要求" ②"出 20 道行为面试题 + 参考思路" ③"模拟面试:你问,我答,你评价" ④"我答得差的地方,给改进版" ⑤"帮我改简历 + 模拟 3 轮"①"analyze this JD's core requirements" ②"20 behavioral questions + reference approach" ③"mock interview: you ask, I answer, you evaluate" ④"improve my weak answers" ⑤"fix my resume + mock 3 rounds"模拟面试评分 ≥ 80%Mock interview score ≥ 80%
自学一个跨学科领域Self-study a cross-disciplinary field4 周建立知识框架Knowledge framework in 4 weeks①"给我这个领域的知识地图(核心概念+关系)" ②每周"深入 3 个概念(学→问→练)" ③"和我已知领域的连接点" ④第 4 周"出综合题 + 我费曼式讲给你" ⑤"指出我框架的漏洞"①"knowledge map of this field (core concepts + relations)" ②weekly "deep-dive 3 concepts (learn→ask→practice)" ③"connections to my known fields" ④week 4 "comprehensive questions + I Feynman-teach you" ⑤"point out gaps in my framework"能画出该领域的知识地图Can draw the field's knowledge map
🎯用法:选一个你正在做的场景,按"提示词序列"逐条问(不要跳步)。每一步都"动手"(写答案/写复述/写提纲),不要只"看"。检验标准不达标就重做那一步。10 个场景覆盖了 90% 的学习需求——备考、读书、外语、编程、论文、写作、乐器、统计、面试、跨学科。把它们存成你的"场景手册",换主题直接套。How to use: pick a scenario you're actually doing, follow the "prompt sequence" one by one (don't skip). Each step is "hands-on" (write answers/restatements/outlines), not just "reading". If the verification standard isn't met, redo that step. The 10 scenarios cover 90% of learning needs — exam prep, books, language, programming, papers, writing, instrument, statistics, interviews, cross-disciplinary. Save them as your "scenario manual"; swap the topic and apply directly.
17

50 条分学科提示词库(拿来就用)

50 Discipline-Specific Prompts (ready to use)

每学科 5 条,覆盖"学/问/练/考"。
5 per discipline, covering "learn/ask/practice/test".

数学(5 条)

Math (5)

##提示词Prompt
11"这道题分步讲,每步说明为什么这样做(别跳步)""Explain this problem step by step, why each step (no skipping)"
22"给我这道题的 3 种解法,比较哪种最快""Give 3 solutions to this problem, compare which is fastest"
33"出 5 道同类型的题(由易到难),我答完你批改""Give 5 same-type questions (easy to hard); I answer, you grade"
44"这个公式的推导过程是什么?用文字讲清每一步""What's the derivation of this formula? Explain each step in words"
55"我卡在这一步:[步骤]。别给答案,问我 2 个问题帮我定位""I'm stuck here: [step]. Don't answer; ask me 2 questions to locate it"

物理(5 条)

Physics (5)

##提示词Prompt
66"用生活里的例子解释这个物理现象(越具体越好)""Explain this physics phenomenon with a real-life example (as concrete as possible)"
77"这个实验的原理是什么?如果改变 [条件] 会发生什么?""What's the principle of this experiment? If I change [condition], what happens?"
88"画出这个过程的受力分析(用文字描述每个力)""Describe the force analysis of this process (each force in words)"
99"出 5 道应用题(用这个定律解决实际问题)+ 批改""Give 5 application questions (use this law for real problems) + grade"
1010"这个概念和 [另一个概念] 的区别是什么?各举 1 个例子""Difference between this concept and [another]? One example each"

外语(5 条)

Languages (5)

##提示词Prompt
1111"和我用英语对话 10 分钟,纠正我的每个错误 + 给更地道的说法""Converse with me in English for 10 min, correct every error + more natural phrasing"
1212"用这 20 个单词各造 1 句对话(主题:点餐)""Make 1 dialogue sentence each with these 20 words (topic: ordering food)"
1313"这段英文我翻译得对吗?[你的翻译]。指出问题 + 给参考译法""Is my translation right? [your translation]. Point out issues + reference translation"
1414"给我 5 个 [语法点] 的例句(含 1 个易错句)+ 解释为什么错""Give 5 example sentences of [grammar point] (incl. 1 error-prone) + why it's wrong"
1515"模拟一段 [场景] 的英文对话,我演 A 你演 B,结束后给评价""Role-play a [scene] English dialogue, I'm A you're B, evaluate after"

编程(5 条)

Programming (5)

##提示词Prompt
1616"这段代码哪里错了?别直接改,先问我 3 个问题帮我定位""Where's the bug? Don't fix directly; ask me 3 questions to locate it"
1717"解释这段代码每一行在做什么(用我能懂的话)""Explain what each line of this code does (in words I can understand)"
1818"出 5 道 [语言] 的动手题(我手答,你批改 + 给优化版)""Give 5 hands-on [language] questions (I hand-answer, you grade + optimized version)"
1919"这段代码怎么优化?先讲思路,再给代码""How to optimize this code? Explain the approach first, then the code"
2020"带我一步步实现这个功能(每步我写,你检查)""Guide me to implement this feature step by step (I write each step, you check)"

写作(5 条)

Writing (5)

##提示词Prompt
2121"批改我的作文:指出 3 个最弱处 + 每个给改法(别打分)""Grade my essay: 3 weakest parts + a fix each (no score)"
2222"分析这篇范文的结构(开头/论证/结尾各用了什么手法)""Analyze this model essay's structure (opening/argument/ending techniques)"
2323"我的论点是 [X]。帮我找 3 个反论 + 怎么反驳""My thesis is [X]. Help me find 3 counterarguments + how to rebut"
2424"这段太啰嗦,帮我精简到一半(保留核心)""This is too wordy; cut it to half (keep the core)"
2525"给我 5 个提升 [文风] 的技巧 + 各举 1 个例子""Give 5 tips to improve [style] + one example each"

历史(5 条)

History (5)

##提示词Prompt
2626"把这段历史讲成一个故事(有冲突、有转折)""Tell this history as a story (with conflict, with turns)"
2727"这个事件的因果链是什么?从原因到结果一步步""What's the causal chain of this event? Step by step from cause to effect"
2828"对比 [A 事件] 和 [B 事件]:背景/过程/影响各什么""Compare [Event A] and [Event B]: background/process/impact each"
2929"出 10 道历史选择题(含 3 道易混淆的)+ 解释""Give 10 history MCQs (incl. 3 easily-confused) + explanations"
3030"这个历史人物为什么重要?3 个贡献 + 1 个争议""Why is this historical figure important? 3 contributions + 1 controversy"

经济/商科(5 条)

Economics/Business (5)

##提示词Prompt
3131"用 [模型] 分析这个案例(结论 + 局限)""Analyze this case with [model] (conclusion + limits)"
3232"这个经济学概念用 1 个生活例子讲清""Explain this economics concept with 1 real-life example"
3333"给出 3 种 [商业问题] 的解决方案 + 各自利弊""3 solutions for [business problem] + pros/cons each"
3434"出 5 道案例分析题(用这个理论)+ 批改""Give 5 case-analysis questions (using this theory) + grade"
3535"这个数据的含义是什么?可能的解读有 3 种""What does this data mean? 3 possible interpretations"

法律(5 条)

Law (5)

##提示词Prompt
3636"这条法律在什么情况下适用?给 3 个例子(含 1 个不适用)""When does this law apply? 3 examples (incl. 1 where it doesn't)"
3737"用 3 句话讲清这个法律条款的核心""Explain this statute's core in 3 sentences"
3838"出 5 道法律案例分析题(我答,你批改 + 指出漏洞)""Give 5 legal case questions (I answer, you grade + point out gaps)"
3939"这个判例的法律原则是什么?适用边界在哪?""What's the legal principle of this precedent? Its boundary?"
4040"对比 [A 法] 和 [B 法]:保护对象/构成要件/惩罚各什么""Compare [Law A] and [Law B]: protected object/elements/punishment"

医学(5 条)

Medicine (5)

##提示词Prompt
4141"这个症状的鉴别诊断有哪些?按可能性排序 + 依据""Differential diagnoses for this symptom? Ranked by likelihood + basis"
4242"这个药的机制是什么?副作用 3 个 + 禁忌 2 个""This drug's mechanism? 3 side effects + 2 contraindications"
4343"出 10 道临床推理题(我给患者信息,你出题我答)""Give 10 clinical-reasoning questions (I give patient info, you ask, I answer)"
4444"这个生理过程的步骤 + 每一步的作用""Steps of this physiological process + what each does"
4545"用 1 个生活类比讲清这个医学概念""Explain this medical concept with 1 real-life analogy"

统计/数据(5 条)

Statistics/Data (5)

##提示词Prompt
4646"这个统计方法适用什么条件?不满足会怎样?""What conditions does this statistical method require? What if not met?"
4747"相关≠因果,举 3 个'相关但非因果'的例子""Correlation ≠ causation; give 3 'correlated but not causal' examples"
4848"用这组数据做 [分析],告诉我每一步在做什么""Do [analysis] on this data, tell me what each step does"
4949"出 5 道'读图表'题(我解读,你批改)""Give 5 'read-the-chart' questions (I interpret, you grade)"
5050"这个研究的设计有什么漏洞?结论可靠吗?""What design flaws does this study have? Are its conclusions reliable?"
18

30 天逐日日历(AI 学习从 0 到习惯)

30-Day Day-by-Day Calendar (AI learning from 0 to habit)

每天 30 分钟,逐日安排,照着做就行。
30 min daily, day-by-day, just follow.
天Day主题Theme30 分钟做什么30-min task
11认识 AI 老师Meet your AI tutor选一个工具 + 用"费曼讲解"让它讲一个你在学的主题Pick a tool + "Feynman explain" a topic you're studying
22第一次追问First follow-up对昨天主题追问 3 轮(为什么→换说法→反例)Follow up 3 rounds on yesterday (why→rephrase→counterexample)
33第一次练First practice让它出 5 题,你手写答案,再让它批改It gives 5 questions; hand-answer; it grades
44第一次考First test让 AI 出 3 道综合题测你(不看笔记)AI gives 3 comprehensive questions (no notes)
55完整循环Full loop新主题,完整走"学→问→练→考"New topic, full "learn→ask→practice→test"
66针对弱点Target weakness把前 5 天错题整理,让 AI 出 3 道针对性练习Compile wrong answers; AI gives 3 targeted exercises
77周复盘Week review写 3 句:哪 3 个提示词最有用?存成模板Write 3: which 3 prompts most useful? Save as templates
88学科定制Discipline-custom从 C5 表选你的学科,用对应"推荐循环"学 1 个主题Pick your discipline from C5; use its "loop" on 1 topic
99间隔复习Spaced review让 AI 测你第 1 天的主题(间隔 7 天)AI tests your day-1 topic (7-day gap)
1010苏格拉底日Socratic day全天只问"为什么",不抄答案(用句式 19)Only ask "why" all day, no copying (pattern 19)
1111出题人日Question-setter day让 AI 出 10 题(5 选择 3 简答 2 应用),限时 20 分钟AI sets 10 questions, timed 20 min
1212批改循环Grading loop写 1 篇小作文(200 字),让 AI 批改 3 处最弱Write a 200-word piece; AI grades 3 weakest
1313费曼反向Reverse Feynman让 AI 当学生,你教它第 1–7 天的内容(它追问你)AI as student; you teach it days 1–7 (it questions you)
1414周复盘 2Week review 2测你第 1 周全部内容(AI 出综合卷);写 3 句收获AI comprehensive paper on week 1; write 3 takeaways
1515跨学科日Cross-discipline day用 C16"跨学科场景"建一个知识地图Build a knowledge map using C16 cross-discipline scenario
1616间隔复习 2Spaced review 2AI 测你第 8 天的学科主题(间隔 8 天)AI tests your day-8 topic (8-day gap)
1717对话陪练Conversation sparring让 AI 当陪练(外语/辩论/面试任选),20 分钟AI as sparring partner (language/debate/interview), 20 min
1818读长文Long-reading用 Kimi 读一篇长文/论文,按 C11 Q18 流程Read a long article/paper with Kimi, C11 Q18 flow
1919验证日Verification day把 AI 给的事实/论文/数据,逐条去查证(防幻觉训练)Verify AI's facts/papers/data one by one (anti-hallucination drill)
2020中点大考Mid-point big examAI 出 20 题综合卷(覆盖 1–19 天),限时 30 分钟AI 20-question paper covering days 1–19, timed 30 min
2121弱点歼灭Weakness elimination针对中点大考的错题,让 AI 出 5 道歼灭题AI 5 "elimination" questions on your wrong answers
2222教别人Teach others把 1 个主题讲给 AI(它当学生追问),录下来自评Teach 1 topic to AI (it questions you), record + self-rate
2323场景实战Scenario practice选 C16 的一个场景,完整走一遍"提示词序列"Pick a C16 scenario, run its full "prompt sequence"
2424间隔复习 3Spaced review 3AI 测你第 8、15 天的主题(间隔 16/9 天)AI tests your day-8 & day-15 topics (16/9-day gap)
2525自主命题Self-set questions你自己出 5 题(参考 AI 的出题方式),再让它批改你的题You set 5 questions (mimic AI's style); AI grades your questions
2626模板固化Template solidify整理你的"提示词模板库"(学/问/练/考各 3 条)Build your "prompt template library" (3 each of learn/ask/practice/test)
2727深度追问Deep follow-up对 1 个主题追问 5 轮以上(用 C4 的 10 句式)Follow up 5+ rounds on 1 topic (C4's 10 patterns)
2828周复盘 3Week review 3AI 测你第 15–27 天内容;写 3 句"我能教别人的"AI tests days 15–27; write 3 "things I can teach"
2929模拟大考Mock finalAI 出 30 题终考(覆盖 30 天),限时 40 分钟AI 30-question final covering 30 days, timed 40 min
3030总结+规划Summary + plan写 5 句:30 天学会的 3 个方法 + 下 30 天的 2 个目标Write 5: 3 methods learned + 2 goals for next 30 days
🏁30 天后的检查点(5 条):① 有 12 条以上自己的"提示词模板";② 能独立走完"学→问→练→考"循环;③ 会间隔复习(1/7/16 天);④ 会"反向费曼"(教 AI);⑤ 会"验证"(查证 AI 的事实)。5 条都达成,AI 学习就成了你的"第二本能"——以后学任何新东西,第一个反应就是"打开 AI,走一遍循环"。Checkpoints after 30 days (5): ① 12+ personal "prompt templates"; ② can run "learn→ask→practice→test" independently; ③ does spaced review (1/7/16 days); ④ does "reverse Feynman" (teach AI); ⑤ does "verification" (check AI's facts). Hit all 5 and AI learning becomes your "second nature" — for anything new, your first reaction is "open AI, run the loop".
19

自测量表:你的"AI 学习力"处于哪一级

Self-Assessment: What Level Is Your "AI Learning Ability"

5 级量表,对照打分,找到自己的位置。
5-level scale; score yourself, find your position.
级别Level特征Traits典型行为Typical behavior下一步Next step
L1 答案机L1 Answer machine只问答案,抄下来Only asks answers, copies them"这道题答案是什么?""What's the answer to this?"改成"为什么 + 引导"Switch to "why + guide"
L2 提问者L2 Questioner会追问 1–2 轮Follows up 1–2 rounds"能再讲讲为什么吗?""Can you explain why more?"加"练"(让它出题)Add "practice" (have it quiz)
L3 练习者L3 Practicer会手答 + 批改Hand-answers + gets graded"出 5 题,我答完你批改""Give 5 questions; I answer, you grade"加"考"(间隔复习)Add "test" (spaced review)
L4 循环者L4 Looper完整"学→问→练→考"Full "learn→ask→practice→test"每个主题走完循环Every topic runs the full loop加"教"(反向费曼)Add "teach" (reverse Feynman)
L5 自主者L5 Autonomous会验证 + 会教学 + 有模板库Verifies + teaches + has template library"AI 给的事实我去查证;我教 AI""I verify AI's facts; I teach AI"带别人一起学Bring others to learn
📏怎么打分:对照"典型行为",你平时最常用的是哪一句?那就是你的级别。L1–L2 的人"用了 AI 但没学会";L3–L5 的人"用 AI 真正在学"。每 7 天自评一次,看是否升了一级。从 L1 到 L5 平均需要 30–60 天(每天 30 分钟)——不是天赋,是"方法 + 重复"。How to score: match "typical behavior" — which line do you use most often? That's your level. L1–L2 "use AI but don't learn"; L3–L5 "actually learn with AI". Self-score every 7 days, check if you leveled up. L1 to L5 takes 30–60 days on average (30 min daily) — not talent, but "method + repetition".
🧩

自测:10 道题检验你"会不会用 AI 学习"

Self-Test: 10 Questions on "Can You Learn with AI"

每题选一个答案,最后点"检查答案"。
Pick one answer each, then click "Check answers".
1. 用 AI 学习时,最重要的第一步是什么?
1. The most important first step when learning with AI?
"先想后问"是核心——生成效应证明自己"生成"的内容记得更牢。
"Think before asking" is core — the generation effect shows self-generated content is remembered better.
2. "AI 说我有篇 2023 年的研究",正确做法是?
2. AI says "there's a 2023 study" — correct move?
AI 会幻觉——编造不存在的论文。重要事实必须自己查证。
AI hallucinates — invents nonexistent papers. Important facts must be verified yourself.
3. "检索练习"为什么比重复阅读有效?
3. Why is retrieval practice more effective than rereading?
Karpicke & Roediger (2008):检索(回忆)激活提取路径,记忆保持率显著更高(2–3 倍)。
Karpicke & Roediger (2008): retrieval (recall) activates extraction pathways, retention significantly higher (2–3×).
4. 第一轮 AI 回答之后,最有效的动作是?
4. After the first AI answer, the most effective action?
第一轮回答只值 40 分,追问才值 60 分。理解来自"追问链"。
The first answer is worth 40%; follow-ups are worth 60%. Understanding comes from the "follow-up chain".
5. 让 AI 出 5 道题后,正确的做法是?
5. After AI gives 5 questions, the correct move?
"手答 + 批改"是检索练习 + 即时反馈,学习真正发生的地方。
"Hand-answer + grading" is retrieval practice + instant feedback, where learning actually happens.
6. "间隔重复"的复习节奏应该是?
6. The spacing rhythm for review should be?
Cepeda (2006) 元分析:分散复习比集中学习记忆保持率高 50%+。
Cepeda (2006) meta-analysis: distributed review retains 50%+ better than massed.
7. AI 说"你懂了",可信吗?
7. AI says "you got it" — trustworthy?
AI 可能顺着你说("讨好")。真正的检验是"出更难的题",而不是"问懂没懂"。
AI may agree with you (flattery). True test is "give harder questions", not "ask if you get it".
8. 学编程时,AI 帮你写代码,正确姿势是?
8. Learning programming, AI writes code — correct posture?
"不懂的代码"是技术债。理解每行 + 自己重写,才是"学会"。
"Code you don't understand" is technical debt. Understanding each line + rewriting is "learning".
9. 医学/法律/财务问题,AI 的角色应该是?
9. For medical/legal/financial questions, AI's role should be?
敏感领域 AI"只参考不决策"——它给方向,专业人士给结论。
In sensitive fields AI is "reference only, no decisions" — it gives direction, pros give conclusions.
10. 30 天后,"会用 AI 学习"的标志是?
10. After 30 days, the mark of "can learn with AI"?
标志是"能力"(独立走循环 + 验证 + 模板),不是"数量"(问了多少题)。
The mark is "ability" (independent loop + verify + templates), not "quantity" (how many questions).
8 分以上 = L4 循环者;5–7 分 = L3 练习者;5 分以下 = 重读 C1–C4。8+ = L4 Looper; 5–7 = L3 Practicer; below 5 = reread C1–C4.
✅

20 项"AI 学习"自查清单

20-Item "AI Learning" Self-Check List

逐项打勾,看你的"AI 学习力"。
Check each item, see your "AI learning ability".
0 / 20
📚

来源(真实可查)

Sources (verifiable)

本页引用的研究/数据/工具。
Research/data/tools cited on this page.
  1. Karpicke, J.D. & Roediger, H.L. (2008). The testing effect: does memory monitoring provide metacognitive control? Memory & Cognition. — 检索练习效应
  2. Slamecka, N.J. & Soloman, D. (1978). Effect of generating information on retention. Journal of Experimental Psychology: Human Learning and Memory. — 生成效应
  3. Ebbinghaus, H. (1885). Memory: A Contribution to Experimental Psychology. — 遗忘曲线/间隔效应
  4. Cepeda, N.J. et al. (2006). Distributed practice in verbal recall tasks. Psychological Bulletin. — 间隔效应元分析
  5. Shute, V. (2008). Feedback. In International Handbook of Educational Achievement Testing. — 反馈效应
  6. Feynman, R. (1985). The Feynman Lectures on Physics. — 费曼学习法
  7. Common Sense Media (2025). Teens and AI survey. — 96% 青少年每周用 AI 学习
  8. OECD (2025). Learning in the Age of Artificial Intelligence. — AI 与教育政策
  9. UNESCO (2024). Guidance for generative AI in education and research. — 生成式 AI 教育指南
  10. Wei, J. et al. (2022). Chain-of-thought prompting elicits reasoning. NeurIPS. — 提示工程基础
  11. Microsoft (2024). Prompt Engineering Guide. — 提示工程系统方法
  12. Pew Research Center (2025). AI and education use survey. — AI 教育使用调查
  13. Brynjolfsson, E. & McAfee, A. (2014). The Second Machine Age. — AI 与人类技能互补
  14. Doshi, S. et al. (2025). Generative AI in education: a systematic review. — 生成式 AI 教育应用综述
20

一页速查表 + 10 条金句

One-Page Quick Reference + 10 Golden Lines

整页的浓缩。记住这 5 句话,就能用上 80% 的效果。
The whole page condensed. Remember these 5 lines and you get 80% of the benefit.
环节Step一句话One-liner核心动作Core action
学Learn让 AI"讲给 12 岁孩子听"Have AI "explain to a 12-year-old"听讲解 + 写 3 个关键词Listen + write 3 keywords
问Ask追问链:为什么→换说法→反例Follow-up chain: why→rephrase→counterexample追问 3–5 轮Follow up 3–5 rounds
练PracticeAI 出题,我手答,它批改AI sets questions, I hand-answer, it grades必须动手写Must write by hand
考Test第 1/3/7 天综合题(间隔重复)Day 1/3/7 comprehensive questions (spacing)不看笔记回忆Recall without notes
验证Verify重要事实"自己查证""Verify" important facts yourself搜原文/看教材Find original/textbook
1AI 老师的价值 = 你追问的深度。第一轮回答只值 40 分,追问才值 60 分。问得越深,学得越多。An AI tutor's value = the depth of your follow-ups. The first answer is worth 40%; the follow-ups are worth 60%. The deeper you ask, the more you learn.
2先想后问。自己先想 10 分钟,带着思路问,而不是直接要答案。生成效应让你"想出来的"记得更牢。Think before asking. Think 10 min yourself first, then ask with your thinking, not just demand the answer. The generation effect makes "thought-out" content stick better.
3必须动手。光"听讲解"只学到 30%。让 AI 出题、你手答、它批改——检索练习才是学习真正发生的地方。Must do it by hand. Just "listening" gets only 30%. AI sets questions, you hand-answer, it grades — retrieval practice is where learning happens.
4重要事实"自己查证"。AI 会幻觉(编造论文/数据)。重要的,你去搜原文、看教材、查 DOI。别让它"自信地错"。"Verify" important facts yourself. AI hallucinates (invents papers/data). For important ones, find the original, check the textbook, look up the DOI. Don't let it be "confidently wrong".
5AI 是放大器,不是替代。它放大你"已有的努力",不创造你"没付出的能力"。基础越扎实,越会用 AI。AI is an amplifier, not a substitute. It amplifies "effort you already put in"; it doesn't create "ability you didn't pay for". The stronger your foundation, the better you use AI.
🎯从明天开始的 3 件小事:① 选一个你在学的主题,用"费曼讲解"句式让 AI 讲一遍;② 追问 3 轮(为什么→换说法→反例);③ 让它出 5 题,你手答,再让它批改。不需要意志力,只需要一个决定——从明天开始,用"学→问→练→考"循环。30 天后,你会发现自己"会的东西"明显变多——不是学得更多,是"学进去"了更多。3 small things starting tomorrow: ① pick a topic you're studying, use the "Feynman" pattern to have AI explain it; ② follow up 3 rounds (why→rephrase→counterexample); ③ have it set 5 questions, hand-answer, then have it grade. No willpower needed, just one decision — from tomorrow, use the "learn→ask→practice→test" loop. After 30 days, you'll find you "know" noticeably more — not studying more, but "absorbing" more.
21

5 个加餐场景:AI 老师还能这样用

5 Bonus Scenarios: More Ways to Use the AI Tutor

除了"学知识",AI 老师还能帮你做这些。
Beyond "learning knowledge", the AI tutor can also help you do these.
1场景 A:AI 当"阅读陪读"。读厚书/论文时,每读完一章让 AI 做 3 件事:① 摘要 3 句;② 提出 3 个你"应该问"的问题;③ 出 5 题测你这章。这样"读"变成"读+问+测",吸收率翻倍。用法:"我刚读完第 X 章,[贴内容]。摘要 3 句 + 3 个我该问的问题 + 5 题测我。"Scenario A: AI as "reading companion". While reading thick books/papers, after each chapter have AI do 3 things: ① summarize in 3 sentences; ② pose 3 questions "you should ask"; ③ set 5 questions to test this chapter. This turns "reading" into "read+ask+test", doubling absorption. Usage: "I just finished chapter X, [paste content]. 3-sentence summary + 3 questions I should ask + 5 questions to test me."
2场景 B:AI 当"辩论对手"。学到一个观点,让 AI 扮"反方"跟你对辩——你正方,它反方,各 3 轮。用法:"我主张 [观点]。你当反方,提 3 个最强反驳,我回应,你再反驳。3 轮后总结谁更有力。"这训练"批判性思维 + 多角度",比"单方面听"有效得多。Scenario B: AI as "debate opponent". Learn a viewpoint, have AI play "opponent" and debate you — you're pro, it's con, 3 rounds each. Usage: "I hold [viewpoint]. You play con; raise 3 strongest rebuttals, I respond, you rebut again. After 3 rounds, summarize who was stronger." This trains "critical thinking + multiple perspectives", far more effective than "one-sided listening".
3场景 C:AI 当"学习教练"。每周让 AI 做 2 件事:① 根据你上周的进度/错题,定本周 3 个目标;② 周五测你本周 3 个目标是否达成。用法:"我上周学了 [X],错题在 [Y]。定本周 3 个目标 + 周五测我。"这把"自学"变成"有教练的自学",有目标、有检验、有节奏。Scenario C: AI as "study coach". Weekly have AI do 2 things: ① set 3 goals this week based on last week's progress/wrong answers; ② Friday test whether you hit the 3 goals. Usage: "Last week I learned [X], wrong answers in [Y]. Set 3 goals this week + test me Friday." This turns "self-study" into "coached self-study" — with goals, verification, rhythm.
4场景 D:AI 当"知识连接器"。学到新知识点,让 AI 帮你"连"进已有网络:① 它和你已知的 3 个知识点有什么关系?② 它能解决什么实际场景?③ 它推翻了什么旧认知?用法:"我新学了 [X]。它和我已知的 [A][B][C] 有什么关系?能解决什么实际场景?推翻了什么旧认知?"这训练"知识网络",而不是"孤立的点"。Scenario D: AI as "knowledge connector". Learn a new concept, have AI help "connect" it into your existing network: ① how does it relate to 3 things you know? ② what real scenarios does it solve? ③ what old beliefs does it overturn? Usage: "I newly learned [X]. How does it relate to [A][B][C] I know? What real scenarios does it solve? What old beliefs does it overturn?" This trains a "knowledge network", not "isolated points".
5场景 E:AI 当"输出检验器"。你写了一篇文章/做了一个方案/答了一道题,让 AI 做"挑刺者":① 最弱的 3 处是什么?② 读者/考官会怎么质疑?③ 怎么改进?用法:"这是我写的 [内容]。指出 3 个最弱处 + 读者会怎么质疑 + 怎么改进。"这训练"输出质量",比"自己看 3 遍"有效——因为 AI 是"没情感包袱的批评者"。Scenario E: AI as "output checker". You wrote an article / made a plan / answered a question, have AI be the "nagger": ① what are the 3 weakest parts? ② how would readers/examiners challenge it? ③ how to improve? Usage: "This is my [content]. Point out 3 weakest parts + how readers would challenge + how to improve." This trains "output quality", more effective than "reading it 3 times yourself" — because AI is a "critic without emotional baggage".
💡5 个加餐的共同逻辑:AI 不只是"输入"(给你知识),更是"输出"(逼你产出 + 检验产出)。输入让你"知道",输出让你"会"。所以最好的用法是"输入 30% + 输出 70%"——让 AI 出题、批改、挑刺、对辩,把"听"变成"做"。记住:学不会的东西,90% 是因为"只输入没输出"。Common logic of the 5 bonuses: AI isn't just "input" (giving you knowledge) but also "output" (forcing you to produce + checking your output). Input makes you "know"; output makes you "can do". So the best usage is "input 30% + output 70%" — have AI set questions, grade, nag, debate, turning "listening" into "doing". Remember: 90% of what you can't learn is because "only input, no output".
22

12 学科 × 3 个真实案例:AI 老师怎么改变学习

12 Disciplines × 3 Real Cases: How the AI Tutor Changes Learning

36 个"人 + 场景 + 方法 + 结果",对号入座找你的同款。
36 "person + scenario + method + result" — find your twin.
学科Discipline案例 1Case 1案例 2Case 2案例 3Case 3
数学Math考研弱生:苏格拉底引导 3 个月 60→85Grad-exam weak: Socratic guidance 3mo 60→85高中生:错题"3 种解法"比较 2 周提升一档High schooler: wrong answers "3 solutions" compare, 2 weeks up a level初中生:公式推导"文字讲清"1 月不再死记Middle schooler: formula derivation "explained in words", 1 month no more rote
物理Physics实验课:AI 解释原理 + 变条件预测,实验分 90+Lab: AI explains principles + predicts conditions, lab score 90+竞赛生:受力分析"文字描述"3 个月进省赛Contest student: force analysis "described in words", 3 months to provincial大学物理:生活例子类比 1 周开窍University physics: real-life analogy, 1 week clicked
化学Chemistry高考生:记忆口诀 + 规律对比 1 月 70→85Gaokao student: mnemonics + rule compare, 1 month 70→85高中生:方程式"分步 + 每步为什么"2 周不再卡High schooler: equations "step + why each step", 2 weeks unstuck大学化学:实验"变条件预测"实验课优秀University chem: experiment "condition-prediction", lab excellent
外语Language口语差:AI 陪练 90 天能 10 分钟对话Weak speaking: AI sparring 90 days → 10-min conversation写作差:批改循环半年被杂志采用Weak writing: grading loop half year, magazine adopted备考雅思:模拟对话 + 纠错 8 周 6→7IELTS prep: mock conversation + correction, 8 weeks 6→7
编程Programming转行者:调试伙伴 6 月独立做小项目Career switcher: debugging partner, 6mo independent small project学生:每行解释 + 重写 3 月能读开源代码Student: explain-each-line + rewrite, 3mo can read open-source工程师:AI 出题练新语言 4 周上手Engineer: AI quizzes on new language, 4 weeks proficient
写作Writing大学生:批改循环 2 月论文分提升 2 档Undergrad: grading loop 2mo, paper score up 2 levels自媒体:挑刺者 1 月阅读量翻倍Blogger: nagger 1 month, views doubled高中生:范文结构分析 1 周作文开窍High schooler: model-essay structure analysis, 1 week clicked
历史History高考生:故事 + 时间线 1 月 70→88Gaokao student: story + timeline, 1 month 70→88大学生:因果链分析 2 周开卷考试 90+Undergrad: causal-chain analysis, 2 weeks 90+ open-book兴趣者:对比事件 3 月能讲清朝代脉络Hobbyist: event comparison, 3 months can narrate dynasty flow
经济Economics大学生:模型分析案例 2 月课程优秀Undergrad: model case analysis, 2 months excellent职场人:数据解读练习 1 月能看财报Professional: data-interpretation practice, 1 month can read financial reports学生:生活例子类比 1 周微观开窍Student: real-life analogy, 1 week micro clicked
法律Law法考:条文"适用边界 + 例子"3 月过线Bar exam: statute "boundaries + examples", 3 months passed大学生:判例原则 + 边界 2 周案例分析高分Undergrad: precedent principles + boundaries, 2 weeks high case score职场人:3 句讲清条款 1 月能起草合同Professional: 3-sentence statute, 1 month can draft contracts
医学Medicine医学生:鉴别诊断"排序 + 依据"3 月诊断课优秀Med student: differential "ranked + basis", 3 months excellent临床:临床推理题 2 周定位能力明显提升Clinical: reasoning questions, 2 weeks clearer localization护生:生理"步骤 + 作用"1 月不再死记Nursing: physiology "steps + function", 1 month no more rote
统计Statistics商科生:误区专练 1 月 SPSS 上手Business student: pitfall drill, 1 month SPSS proficient研究者:读图表题 2 周能审论文图表Researcher: chart-reading questions, 2 weeks can audit paper charts学生:相关≠因果例子 1 周不再混淆Student: correlation≠causation examples, 1 week no more confusion
音乐Music自学者:乐理"文字讲清 + 谱子分析"3 月弹 5 首Self-taught: theory "in words + score analysis", 3 months 5 songs学生:指法"描述判断"2 周错误率减半Student: fingering "description-judgment", 2 weeks error halved兴趣者:曲单"由易到难"6 月能合奏Hobbyist: "easy-to-hard" song list, 6 months can play ensemble
🎯36 个案例的共性(5 条):① 都不是"AI 多神奇",是"用法对"(先想后问 + 动手 + 验证);② 时间都是"周/月"级,不是"天"级——方法对,时间会说话;③ 提升幅度真实(60→85、70→88、6→7),不是"从 0 到 100"的奇迹;④ 都"高频反馈"(每周多次),不是"一次性";⑤ 都"输出"(手答/写/弹/讲),不是"只输入"。记住:AI 老师的回报 = 你的方法 × 你的时间。方法对,时间复利。Common traits of the 36 cases (5): ① none is "AI is so magical", it's "right usage" (think-first + hands-on + verify); ② time is "week/month" scale, not "day" scale — right method, time speaks; ③ gains are real (60→85, 70→88, 6→7), not "0 to 100" miracles; ④ all "high-frequency feedback" (multiple per week), not "one-shot"; ⑤ all "output" (hand-answer/write/play/explain), not "input-only". Remember: AI tutor's payoff = your method × your time. Right method, time compounds.
23

25 个常见难题:AI 老师怎么解(对照表)

25 Common Problems: How the AI Tutor Solves Them (Reference Table)

把"学习卡点"列出来,每个给"AI 怎么帮 + 提示词"。
Lays out "learning sticking points", each with "how AI helps + prompt".
##难题ProblemAI 怎么帮How AI helps提示词Prompt
11概念听了一知半解Concept half-understood换 3 种说法 + 反例3 rephrasings + counterexample"换 3 种方式讲 [X] + 1 个反例""Explain [X] 3 ways + 1 counterexample"
22公式记不住Can't remember formula讲推导 + 编口诀Explain derivation + mnemonic"讲清这个公式怎么推出来的 + 编口诀""Explain how this formula is derived + make a mnemonic"
33应用题不会列式Can't set up equations先问 3 个问题定位Ask 3 questions to locate"我列不出式子,问我 3 个问题帮我定位""I can't set the equation; ask me 3 questions to locate"
44背单词总忘Forgets vocabulary间隔出题 + 造句Spaced questions + sentences"这 20 个词,第 1/3/7 天各出 5 题 + 造句""These 20 words, 5 questions each on day 1/3/7 + sentences"
55口语不敢说Afraid to speak低压力对话 + 只纠错Low-pressure conversation + error-only"和我轻松对话,只纠大错,别打断我""Casual conversation, only correct big errors, don't interrupt"
66阅读长文读不进去Can't get into long reading先摘要 + 问题引导Summary first + question-guided"先 3 句摘要 + 3 个我应该问的问题,再读""3-sentence summary + 3 questions I should ask, then read"
77论文看不懂Can't understand papers分层讲:摘要→方法→结论Layered: abstract→method→conclusion"这篇论文的摘要 3 句 + 方法 + 结论 + 局限""This paper's 3-sentence abstract + method + conclusion + limits"
88代码报错不会查Can't debug errors先问 3 个问题定位Ask 3 questions to locate"报错 [错误],别直接改,问我 3 个问题""Error [error], don't fix; ask me 3 questions"
99写东西没思路No ideas for writing头脑风暴 + 挑刺Brainstorm + nag"给我 5 个 [主题] 的切入角度 + 各自优劣""5 angles for [topic] + pros/cons each"
1010历史时间线乱History timeline messy时间线 + 因果链Timeline + causal chain"把 [时期] 的事件按时间 + 因果画出来""Draw [era] events by time + causality"
1111经济学概念抽象Econ concepts abstract生活例子类比Real-life analogy"用 [生活场景] 解释 [概念]""Explain [concept] with [life scene]"
1212法律条文记不住Can't remember statutes3 句核心 + 适用例子3-sentence core + application examples"这条法律 3 句核心 + 3 个适用例子""This statute's 3-sentence core + 3 application examples"
1313医学术语太多Too many medical terms术语"生活类比"讲Terms "in real-life analogy""用生活类比讲 [术语],别用术语解释术语""Explain [term] in real-life analogy, no term-explains-term"
1414统计方法选错Wrong statistical method讲适用条件 + 反例Explain conditions + counterexample"这个数据该用什么方法?什么情况下不能用?""What method for this data? When can't you use it?"
1515乐器谱子看不懂Can't read instrument scores谱子"逐句分析"Score "phrase-by-phrase""这段谱子的手法 + 难点在哪 + 怎么练""This score's technique + where's the difficulty + how to practice"
1616学完就忘Forgets after learning间隔重复出题Spaced-repetition questions"第 1/3/7 天各出 5 题测我 [X]""5 questions each on day 1/3/7 testing me on [X]"
1717考试紧张Exam anxiety模拟考 + 限时Mock exam + timed"出 20 题,限时 30 分钟,模拟真实考试""20 questions, 30 min, simulate a real exam"
1818知识点孤立Isolated knowledge points知识连接器Knowledge connector" [X] 和我已知的 [A][B] 有什么关系?""How does [X] relate to [A][B] I know?"
1919不知道自己懂没懂Don't know if I got it反向费曼(教 AI)Reverse Feynman (teach AI)"我讲 [X] 给你听,你当学生追问""I teach you [X]; you be the student and question"
2020学习没动力No motivation周教练 + 小目标Weekly coach + small goals"给我本周 3 个小目标 + 周五测我""3 small goals this week + test me Friday"
2121资料太多不知看哪个Too much material, don't know what to read知识地图 + 优先级Knowledge map + priority" [X] 领域,我先看哪 3 个资料?为什么?""In [X] field, which 3 resources first? Why?"
2222跨领域没背景No background in cross-field知识地图 + 连接点Knowledge map + connection points" [X] 领域知识地图 + 和我 [Y] 的连接点""[X] knowledge map + connections to my [Y]"
2323输出质量差Poor output quality挑刺者 + 对比Nagger + compare"这是我写的,3 个最弱处 + 改进版""This is my writing, 3 weakest + improved version"
2424依赖 AI 不想动Dependent, don't want to think苏格拉底强制思考Socratic forced thinking"别给答案,只提问,引我自己想""No answers, only questions, guide me to think"
2525AI 回答太长抓不住重点AI answers too long, miss the point限制长度 + 3 关键词Length limit + 3 keywords"用 5 句话 + 3 个关键词总结 [X]""Summarize [X] in 5 sentences + 3 keywords"
🎯用法:把这张表存成你的"AI 学习急救包"。学习卡住了,别硬扛——查表找到对应行,把"提示词"里的 [X] 换成你的主题,直接问。25 个难题覆盖了 90% 的学习卡点:概念、公式、应用、记忆、口语、阅读、论文、代码、写作、历史、经济、法律、医学、统计、乐器、遗忘、考试、孤立、动力、资料、跨域、输出、依赖、太长。记住:卡住不是"笨",是"还没找到对的提示词"。How to use: save this table as your "AI learning first-aid kit". Stuck? Don't grind — find the matching row, swap [X] in the "prompt" with your topic, ask directly. The 25 problems cover 90% of learning sticking points: concepts, formulas, application, memory, speaking, reading, papers, code, writing, history, econ, law, medicine, statistics, instrument, forgetting, exams, isolation, motivation, materials, cross-field, output, dependency, too-long. Remember: being stuck isn't "stupid", it's "haven't found the right prompt yet".
24

18 个判断场景:该不该用 AI 学 + 10 条速记

18 Judgment Scenarios: Should You Learn with AI? + 10 Quick Lines

不是所有事都适合用 AI。判断清楚,才用得对。
Not everything suits AI. Judge clearly, use it right.
##场景Scenario该用 AI 吗Use AI?怎么定位Position
11理解抽象概念Understand abstract concepts✅ 强烈推荐✅ Strongly换 3 种说法 + 类比 + 反例3 rephrasings + analogy + counterexample
22刷选择题练手感Drill multiple-choice✅ 推荐✅ YesAI 出 50 题 + 批改 + 弱点分析AI sets 50 + grades + weakness analysis
33背单词/公式Memorize words/formulas✅ 推荐(间隔出题)✅ Yes (spaced)AI 第 1/3/7 天出题 + 你手答AI questions day 1/3/7 + you hand-answer
44练口语/对话Practice speaking✅ 强烈推荐✅ StronglyAI 陪练 + 纠错 + 调难度AI sparring + correction + difficulty
55读长文/论文Read long text/papers✅ 推荐(配合查证)✅ Yes (with verification)AI 摘要 + 引导问题,数据回原文AI summary + guiding questions; data back to original
66写代码/调试Write code/debug✅ 推荐(必须理解)✅ Yes (must understand)AI 解释每行 + 你重写 + 它批改AI explains each line + you rewrite + it grades
77写文章/练表达Write/express✅ 推荐(自己写+AI 批改)✅ Yes (you write + AI grades)你先写,AI 挑 3 个最弱处You write first; AI nags 3 weakest
88练乐器/运动Instrument/sport⚠️ 理论用,动手不用⚠️ Theory yes, hands-on noAI 讲乐理/要领,练必须动手AI teaches theory; practice is hands-on
99做实验/动手操作Experiments/hands-on⚠️ 只辅助不替代⚠️ Assist onlyAI 讲原理 + 预测,操作自己做AI explains + predicts; you operate
1010重大决策(投资/医疗/法律)Major decisions❌ 只参考不决策❌ Reference onlyAI 给选项+利弊,专业人士决策AI gives options; pros decide
1111应付作业/考试作弊Do homework/cheat❌ 别用(毁思考力)❌ Don't (destroys thinking)用 AI"学",不用它"代做"Use AI to "learn", not "do for you"
1212获取"权威事实"(法律条文/医学数据)Authoritative facts (statutes/medical data)⚠️ 用 AI 找线索,自己查证⚠️ AI for leads, you verifyAI 指出去哪查,原文为准AI points where to check; original is truth
1313培养创造力(写故事/设计)Creativity (stories/design)⚠️ 辅助不替代⚠️ Assist onlyAI 给灵感 + 挑刺,作品必须自己出AI gives inspiration + nags; work is yours
1414建立学科知识框架Build discipline framework✅ 推荐✅ YesAI 画知识地图 + 你填内容 + 它查漏AI maps + you fill + it checks gaps
1515复盘总结(学完一个阶段)Review (after a phase)✅ 推荐✅ YesAI 出综合卷 + 你费曼式讲 + 它挑漏洞AI comprehensive paper + you Feynman + it nags
1616情感陪伴/心理疏导Emotional support/counseling❌ 不能替代专业❌ Can't replace prosAI 可以聊,但"症状"找专业人士AI can chat; "symptoms" go to pros
1717了解"时事/最新进展"News/latest progress⚠️ 用 AI 整理,自己核源⚠️ AI organizes, you check source问清 AI 知识截止 + 用搜索验证Ask AI's cutoff + verify with search
1818学"元技能"(学习方法/思维方法)Meta-skills (learning/thinking methods)✅ 强烈推荐✅ StronglyAI 讲学习科学 + 给你可检验的计划AI teaches learning science + verifiable plan
##一句话速记One-line quick note
11AI 老师的价值 = 你追问的深度。An AI tutor's value = the depth of your follow-ups.
22先想 10 分钟,再问 AI。Think 10 minutes, then ask AI.
33手答 + 批改,学习才发生。Hand-answer + grading — that's where learning happens.
44第 1/3/7 天,让它测你。Days 1/3/7 — let it quiz you.
55重要事实,自己查证。Verify important facts yourself.
66AI 说"你懂了",让它出更难的题。AI says "you got it" — have it give harder questions.
77输入 30%,输出 70%。Input 30%, output 70%.
88敏感领域只参考,不决策。Sensitive fields: reference only, no decisions.
99用 AI"学",不用它"代做"。Use AI to "learn", not to "do for you".
1010AI 是放大器,不是替代。AI is an amplifier, not a substitute.
🎯判断口诀(三问):遇到一件事,问自己 3 个问题——① 这事需要"理解/练习/反馈"吗?(是→用 AI)② 这事需要"动手/做实验/面对真人"吗?(是→AI 只辅助)③ 这事涉及"重大决策/敏感领域"吗?(是→AI 只参考)。3 问走完,你就知道 AI 该在哪、不该在哪。把"用对"比"多用"更重要。Decision mantra (3 questions): for any task, ask 3 questions — ① Does it need "understanding/practice/feedback"? (yes → use AI) ② Does it need "hands-on/experiments/real people"? (yes → AI assists only) ③ Does it involve "major decisions/sensitive fields"? (yes → AI references only). After the 3 questions, you know where AI belongs and where it doesn't. "Using right" beats "using more".
25

30 类人 × 30 种"AI 学习"起步方案

30 Types of People × 30 "AI Learning" Starting Plans

不管你是谁、学什么、几岁,都能找到你的"第一步"。
No matter who you are, what you study, your age — find your "first step".
##你是谁Who you are在学什么Studying30 天起步方案(每天 30 分钟)30-day start (30 min/day)
11初中生Middle schooler数理化Math/Physics/Chem每天 1 个概念"费曼讲解"+ 3 道手答题 + 周末 1 次模拟考Daily 1 concept "Feynman" + 3 hand-answered Qs + weekly mock exam
22高中生High schooler高考全科Gaokao all subjects每天 2 个薄弱概念"追问 3 轮"+ 5 道真题风格题 + 间隔复习Daily 2 weak concepts "3-round follow-up" + 5 past-paper Qs + spaced review
33考研党Grad-exam candidate数学/专业课Math/professional courses苏格拉底式解题 + 错题 3 种解法 + 第 1/3/7 天综合卷Socratic problem-solving + wrong answers 3 ways + day 1/3/7 papers
44大学生Undergrad专业课/论文Courses/papers每章"摘要+3 问+5 题"+ 论文"分层讲"+ 考前模拟考Each chapter "summary+3Q+5Qs" + paper "layered" + mock exam
55研究生Grad student论文/文献Papers/literature每周 10 篇"摘要+关系图+矛盾点"+ 自己写综述 + AI 挑刺10 papers/week "summary+map+contradictions" + own review + AI nags
66职场新人Junior professional职业技能Job skills每天 1 个岗位技能"案例+练习"+ 每周 1 次模拟面试/汇报Daily 1 job skill "case+practice" + weekly mock interview/report
77转行者Career switcher新领域New field4 周"知识地图+每 3 概念深学+连接点+费曼讲给 AI"4 weeks "map + 3 concepts deep + connections + Feynman to AI"
88程序员Programmer新技术栈New tech stack"解释每行+重写+5 道动手题"+ 周末 1 个小项目 + AI debug"Explain each line + rewrite + 5 hands-on Qs" + weekly project + AI debug
99设计师Designer审美/理论Aesthetics/theory每天 1 幅经典作品"手法分析"+ 自己仿作 + AI 挑 3 个最弱Daily 1 classic work "technique analysis" + own imitation + AI nags 3
1010教师Teacher教学法/课程Pedagogy/courses用 AI"出课件+出题+批改"+ 反思教学 + 学生案例Use AI "courseware+questions+grading" + teaching reflection + student cases
1111家长Parent陪孩子学Learning with kids每天用 AI"给孩子讲 1 个概念"+ 亲子"反向费曼"(孩子讲给 AI)Daily AI "explains 1 concept to child" + parent-child "reverse Feynman"
1212银发族Senior兴趣/数字技能Hobbies/digital skills每天 1 个数字技能"分步讲"+ 动手做 + AI 耐心纠错Daily 1 digital skill "step-by-step" + hands-on + AI patient correction
1313自由职业者Freelancer多技能Multiple skills每周 1 个新技能"知识地图+3 概念+1 项目"+ 输出作品集1 skill/week "map+3 concepts+1 project" + portfolio output
1414创业者Entrepreneur商业/市场Business/market每天 1 个商业案例"模型分析"+ 自己决策 + AI 当"红队"挑刺Daily 1 business case "model analysis" + own decision + AI "red team"
1515外贸人Foreign trade英语/跨文化English/cross-cultural每天 10 分钟"邮件/对话模拟"+ AI 纠错 + 地道表达Daily 10-min "email/conversation mock" + AI correction + idiomatic
1616医学生Med student临床/解剖Clinical/anatomy鉴别诊断"排序+依据"+ 临床推理题 + 生活类比讲术语Differential "ranked+basis" + reasoning Qs + real-life analogies
1717法学生Law student法律/判例Law/precedents条文"3 句核心+适用例子"+ 案例分析 + 判例原则Statute "3-sentence core+examples" + case analysis + precedent
1818金融从业者Finance professional数据/模型Data/models"相关≠因果"例子 + 读图表题 + 模型适用条件"Correlation≠causation" + chart-reading + model conditions
1919语言学习者Language learner外语Foreign language每天 20 分钟"对话陪练+纠错"+ 间隔背词 + 每月模拟对话Daily 20-min "conversation+correction" + spaced vocab + monthly mock
2020写作爱好者Writing enthusiast文学/表达Literature/expression每天 200 字 + AI 挑 3 个最弱 + 每周 1 篇限时写作Daily 200 words + AI nags 3 + weekly timed writing
2121音乐学习者Music learner乐理/乐器Theory/instrument乐理"文字讲清"+ 谱子分析 + 每天 30 分钟练 + AI 判断Theory "in words" + score analysis + 30 min practice + AI judges
2222体育爱好者Sports enthusiast战术/科学Tactics/science战术"文字描述+预测"+ 运动科学"原理+应用"Tactics "describe+predict" + sport science "principle+application"
2323心理学习者Psychology learner理论/自我Theory/self理论"生活例子+自测"+ 谨慎自我分析(症状找专业人士)Theory "real-life+self-test" + cautious self-analysis (symptoms to pros)
2424历史爱好者History enthusiast史实/脉络Facts/flow"故事+时间线+因果链"+ 对比事件 + 出题自测"Story+timeline+causal chain" + compare events + self-quiz
2525编程初学者Coding beginnerPython/WebPython/Web"解释每行+重写+5 题动手"+ 每周 1 个小项目 + AI debug"Explain+rewrite+5 hands-on" + weekly project + AI debug
2626数据分析师Data analyst统计/工具Statistics/tools"方法适用条件+反例"+ 真实数据跑一遍 + 读图表题"Method conditions+counterexample" + real data run + chart Qs
2727内容创作者Content creator选题/文案Topics/copy"5 个切入角度+各自优劣"+ 自己写 + AI 挑刺 + 对比改进"5 angles+pros/cons" + own writing + AI nags + compare
2828求职者Job seeker面试/简历Interview/resumeJD 分析 + 20 道行为题 + 模拟面试 + 简历修改 + 3 轮JD analysis + 20 behavioral Qs + mock interview + resume + 3 rounds
2929终身学习者Lifelong learner跨领域Cross-field每周 1 个跨领域"知识地图+3 概念+1 输出"+ 30 天 1 个框架1 cross-field/week "map+3 concepts+1 output" + 30 days 1 framework
3030AI 学习者(学 AI 本身)AI learner (learning AI itself)AI 原理/应用AI principles/apps"原理+例子+局限"+ 动手试 + 验证 AI 的幻觉 + 写心得"Principle+example+limits" + hands-on + verify hallucinations + write notes
🎯怎么用这张表:找到"你是谁"那一行,看"30 天起步方案",从明天开始照着做。不需要想"我该学什么"——你的职业/身份已经决定了。30 类人覆盖 95% 的学习者:学生(初中/高中/考研/大学/研究生)、职场(新人/转行/程序员/设计师/教师/家长/银发/自由职业/创业/外贸/医/法/金融)、爱好者(语言/写作/音乐/体育/心理/历史/编程/数据/创作/求职/终身/AI 本身)。记住:没有"太老/太忙/太基础"的人,只有"还没找到对的 30 天方案"的人。你的方案就在上面。How to use this table: find "who you are" row, read the "30-day start plan", start tomorrow. No need to wonder "what should I learn" — your job/identity already decided. 30 types cover 95% of learners: students (middle/high/grad-exam/undergrad/grad), professionals (junior/switcher/programmer/designer/teacher/parent/senior/freelancer/entrepreneur/foreign-trade/medical/legal/finance), enthusiasts (language/writing/music/sports/psychology/history/coding/data/creation/job-seeking/lifelong/AI itself). Remember: no one is "too old/busy/basic" — only "haven't found the right 30-day plan yet". Your plan is above.
26

10 条金句 + 最终总结:把 AI 学习变成你的第二本能

10 Golden Lines + Final Summary: Turn AI Learning into Your Second Nature

整页的最后一句话:AI 老师的价值,取决于你怎么用它。
The page's final line: an AI tutor's value depends on how you use it.
##金句Golden line
11"问答案"的人得到答案,"问为什么"的人得到理解。"Answer-askers" get answers; "why-askers" get understanding.
22AI 不会让你变聪明,但"用对方法的 AI"会。AI won't make you smarter; "AI used with the right method" will.
33听懂了 30%,做出来 70%。Listening gets 30%; doing gets 70%.
44AI 的耐心是无限的,你的时间不是——所以把时间花在"追问"上。AI's patience is infinite; your time isn't — spend it on follow-ups.
55先想 10 分钟,比直接要答案值 100 倍。Thinking 10 minutes is worth 100× demanding the answer.
66AI 是"放大器"——放大你已有的努力,不创造你没付出的能力。AI is an "amplifier" — amplifies effort you already put in, doesn't create ability you didn't pay for.
77学不会的东西,90% 是因为"只输入没输出"。90% of what you can't learn is because "only input, no output".
88重要的事实,永远自己查证——AI 会"自信地错"。Always verify important facts yourself — AI can be "confidently wrong".
9930 天后,你的"第一个反应"应该是"打开 AI,走一遍循环"。After 30 days, your "first reaction" should be "open AI, run the loop".
1010AI 老师不打烊,但"学"只发生在你自己脑子里。The AI tutor never closes, but "learning" only happens in your own head.
🎯最终总结(5 句话):① AI 老师是"免费的、24 小时的、无限耐心的"老师——但它不是"答案机",是"苏格拉底";② 学习 = "学(听讲解)→ 问(追问链)→ 练(手答+批改)→ 考(间隔重复)"的循环,缺任何一步都打折;③ 三条铁律:先想后问、必须动手、重要事实自己查证;④ 30 天 5 个标志:有模板库、能独立走循环、会间隔复习、会反向费曼、会验证;⑤ 从今天开始的 3 件小事:选 1 个主题用"费曼讲解"问 AI + 追问 3 轮 + 让它出 5 题你手答。不需要意志力,只需要一个决定。30 天后,"AI 学习"就从"工具"变成了你的"第二本能"——这是这个时代,普通人和普通人之间最大的差距之一。Final summary (5 lines): ① The AI tutor is "free, 24/7, infinitely patient" — but it's not an "answer machine", it's a "Socrates"; ② Learning = the loop "learn (listen) → ask (follow-up chain) → practice (hand-answer + grade) → test (spaced repetition)", skip any and it drops; ③ Three iron rules: think before asking, must do by hand, verify important facts yourself; ④ 30-day 5 marks: template library, independent loop, spaced review, reverse Feynman, verification; ⑤ 3 small things starting today: pick 1 topic, ask AI with "Feynman explain", follow up 3 rounds, have it set 5 questions you hand-answer. No willpower needed, just one decision. After 30 days, "AI learning" goes from "tool" to your "second nature" — one of the biggest gaps between ordinary people in this era.
27

附录:5 条补充问答 + 你的"第一句提示词"

Appendix: 5 Bonus Q&A + Your "First Prompt"

最后的 5 个常见问题,和你明天就能用的"第一句"。
The last 5 FAQs, and your "first sentence" to use tomorrow.
Q1我基础很差,AI 会嫌弃我吗?不会。AI 没有情绪、不会嫌弃、不会觉得你"问得太简单"。它最大的价值就是"可以同一个问题问 100 遍"。基础差反而要用 AI——它可以从"最底层"讲起,真人老师不一定有这个耐心。My foundation is weak — will AI look down on me? No. AI has no emotions, doesn't look down, doesn't think "too simple". Its biggest value is "asking the same thing 100 times". Weak foundation is exactly when to use AI — it can start "from the bottom"; real teachers may not have that patience.
Q2我要花多少钱?0 元起步。DeepSeek、豆包、Kimi 免费档就够 90% 的学习场景。ChatGPT/Gemini 免费版也够用。付费主要买"速度+更强模型",不是"能不能学会"。先用免费,确定要长期用再考虑付费。How much money do I need? Start at 0. Free tiers of DeepSeek, Doubao, Kimi cover 90% of learning. ChatGPT/Gemini free tiers also suffice. Paying mainly buys "speed + stronger models", not "whether you can learn". Start free; consider paying once you confirm long-term use.
Q3用 AI 学习会被老师/学校发现吗?"用 AI 学习"(自己思考+手答+复习)完全正当,鼓励。"用 AI 代做作业/考试"(复制粘贴答案)才是问题。区别在"你在思考吗"。只要"先想后问 + 手答 + 验证",就是"用工具学习",和用词典/计算器一样正常。Will teachers/school find out I use AI to learn? "Learning with AI" (own thinking + hand-answer + review) is completely legitimate, encouraged. "Using AI to do homework/exams" (copy-paste) is the problem. The difference is "are you thinking?". As long as you "think first + hand-answer + verify", it's "using a tool to learn", as normal as using a dictionary/calculator.
Q4我每天只有 15 分钟,够用吗?够。把 30 分钟方案压缩成 15 分钟:1 个主题"学 + 问"(8 分钟)+ 2 道题手答(7 分钟)。频率比时长重要——每天 15 分钟 × 30 天 > 每周 2 小时。AI 学习的复利来自"每天",不是"一次很多"。I only have 15 min/day, is it enough? Yes. Compress the 30-min plan to 15: 1 topic "learn + ask" (8 min) + 2 hand-answered questions (7 min). Frequency beats duration — 15 min daily × 30 days > 2 hours weekly. AI learning's compound interest comes from "daily", not "a lot at once".
Q5我用 AI 一段时间后,还会"离开 AI 不会学"吗?如果方法对,不会。"先想后问"训练的是你自己的思考力,不是依赖。检验标准:关掉 AI,你能不能独立讲清、独立做题、独立复习?能,说明"AI 学习"内化成了你的能力;不能,说明你"只输入没输出"。每 30 天做一次"无 AI 测试",确认能力在自己身上。After using AI a while, will I "can't learn without AI"? If the method is right, no. "Think before asking" trains your own thinking, not dependency. Test: turn off AI — can you explain, solve, review independently? Yes → "AI learning" has internalized into your ability; no → you were "input-only". Do a "no-AI test" every 30 days, confirm the ability is yours.
🚀你的"第一句提示词"(明天就能用):打开任意一个免费 AI(DeepSeek/豆包/Kimi),输入——"你是我的老师。我是[你的身份],正在学[主题],水平是[水平]。请像给 12 岁孩子讲一样,用 1 个生活类比给我讲清这个概念,然后出 3 道题考我,我答完你批改。"把 3 个括号填好,回车。这就是你的"AI 学习"的第一天。30 天后,它会变成你的习惯。Your "first prompt" (use tomorrow): open any free AI (DeepSeek/Doubao/Kimi), type — "You're my teacher. I'm [your identity], learning [topic], level [level]. Explain this concept like to a 12-year-old with one real-life analogy, then quiz me with 3 questions; after I answer, grade me." Fill the 3 brackets, hit enter. That's day one of your "AI learning". After 30 days, it becomes your habit.