# How to Build a Personal AI Learning System That Teaches You Any Skill in Half the Time (Full Course)
**作者**: NeilXbt
**日期**: 2026-05-07T05:08:55.000Z
**来源**: [https://x.com/neil_xbt/status/2052254328409231372](https://x.com/neil_xbt/status/2052254328409231372)
---
The way most people learn new skills is the most expensive method available.
They find a course. They work through it from start to finish, regardless of what they already know and what they do not. They absorb the same content at the same pace with the same examples as every other person who purchased the same course. They finish it, feel ready, and then discover that the real skill gap exists somewhere the course never touched.
This is the one-size-fits-all model of learning, and it fails systematically for anyone who does not happen to fit the size.
The alternative is not a better course. It is a personal learning system built around you specifically, your existing knowledge, your specific gaps, your schedule, your learning style, and the exact version of the skill you actually need. That system, built correctly with Claude at the center of it, produces results that the best course on the market cannot match.
Students in AI-powered learning environments achieve 54% higher test scores, show 30% better learning outcomes, and experience 10x more engagement compared to traditional methods.
A 2025 randomized controlled trial published in Scientific Reports found AI tutoring outperformed in-class active learning with an effect size between 0.73 and 1.3 standard deviations.
This article is the full system. Every component. Every prompt. Every practice method. In the exact order that produces results.
## Why Most Learning Systems Fail Before They Start
Before building a better system, you need to understand why the standard approach wastes so much time.
The core problem is that traditional learning treats your gaps as unknown. A course cannot know what you already understand, what you are close to understanding, and what you have no context for yet. So it covers everything, in order, at a pace set for the median learner. For the person who already understands half the material, this is slow and demotivating. For the person who has significant gaps in the foundational knowledge the course assumes, it is incomprehensible without additional help.
Either way, time is wasted.
Adaptive learning systems focus on knowledge gaps rather than repeating known material. Because adaptive learning systems continuously analyze learner behavior, skills, and performance to adjust learning paths in real time, they help close skill gaps faster, improve engagement, and deliver targeted learning at scale.
The personal AI learning system described here is adaptive by design. It starts by diagnosing exactly where you are, builds a path from that specific starting point to your specific goal, and adjusts as your understanding develops. Nothing is covered that does not need to be covered. No time is spent on concepts you already have. Every session is targeted at the actual gap.
## Module 1: The Diagnostic That Builds Your Learning Map
Every effective learning system starts not with content but with an honest assessment of current knowledge. Most people skip this because it feels like delay. It is the opposite. Time spent in accurate diagnosis is time saved everywhere else.
Open a conversation with Claude and run this prompt before you start learning anything:
```
I want to learn [skill]. Before we start, I want you to diagnose
my current knowledge accurately. Ask me 10 questions that test
different levels of understanding of this skill, from foundational
concepts to intermediate application to advanced nuance.
After I answer, give me:
1. An honest assessment of where my knowledge is strong
2. The specific gaps that will block my progress if not addressed first
3. The concepts I am close to understanding that will click
quickly with targeted explanation
4. An estimated learning timeline based on my starting point
Do not be encouraging. Be accurate.
```
The instruction "do not be encouraging, be accurate" is the most important part of this prompt. Claude's default mode when assessing someone's knowledge is to find what they got right and frame the gaps gently. For the purpose of building a learning system, you need the unvarnished version. You need to know exactly where you are, not a flattering approximation of it.
The output of this session is your learning map. It tells you what to cover, what to skip, what needs foundational work before more advanced concepts make sense, and a realistic timeline for the whole path. Every subsequent session is built on top of this map rather than starting from a generic curriculum.
Update this map every two to three weeks. As your understanding develops, the gaps change. The map should reflect where you are now, not where you were when you started.
## Module 2: The Socratic Learning Loop That Replaces Passive Reading
Most people learn by consuming content. They read, watch, or listen, and they build a feeling of familiarity with the material that they mistake for understanding.
Familiarity is not understanding. The test of real understanding is whether you can use the knowledge to solve a problem you have not seen before, explain the concept to someone with no background in it, and identify when the concept applies and when it does not. Passive consumption produces familiarity. Active generation produces understanding.
The Socratic learning loop forces you into active generation from the first moment of every session.
Run this prompt at the start of every learning session:
```
I am learning [skill]. Today I want to work on [specific concept
from your learning map].
Do not explain this concept to me directly. Instead, ask me a
series of questions that lead me to understand it myself. Start
with what I already know and build from there. When I am wrong
or incomplete, ask a follow-up question that points me toward
the gap rather than filling it for me.
After I have worked through the concept through your questions,
give me a concise explanation that consolidates what I discovered
and fills any gaps I did not reach on my own.
```
This approach requires more effort per session than reading an explanation. It also produces significantly better retention and a qualitatively different kind of understanding. When you arrive at a conclusion through your own reasoning, guided by questions, you understand not just what is true but why it is true and how it connects to what you already know.
Generative AI demonstrates exceptional capabilities in providing adaptive learning experiences tailored to individual preferences and needs. Utilizing different forms of generative AI across various subjects yields superior learning outcomes.
The sessions feel harder. That difficulty is the learning. Ease during a study session is usually a sign that not much is being acquired.
## Module 3: The Spaced Repetition System That Prevents Forgetting
Learning something is not the same as retaining it. The science on memory is clear and has been replicated for over a century: without deliberate repetition at strategic intervals, most of what you learn disappears within days.
The challenge of manual spaced repetition is the logistics. Tracking what you learned when, calculating the optimal review intervals, and actually following through with review sessions is a significant overhead that most people abandon.
Claude handles the logistics. You handle the retrieval.
After each learning session, run this prompt:
```
I just finished learning [what you covered today]. Generate
five questions that test my understanding of the key concepts
from this session. Make them applied, not definitional — I
should have to use the knowledge, not just recall a term.
Save these as my review set for this concept. I will ask you
to generate a review session in 3 days, then 7 days, then
14 days.
```
On each review date, open a new session with:
```
Review session for [concept], [X] days after initial learning.
Generate the five questions you created after my initial session,
plus two new questions that connect this concept to what I have learned since then.
```
The connection questions are the most valuable part. They force you to integrate new knowledge with existing knowledge rather than keeping them in separate compartments, which is where deep understanding lives.
You do not need a flashcard app. You do not need a dedicated spaced repetition platform. You need a consistent habit of running review sessions at the right intervals. Claude tracks nothing automatically, so you track the review dates manually. A simple calendar reminder for each review date is enough.
## Module 4: The Application Sprint That Builds Real Capability
Knowledge without application is potential that never converts. Every skill has a gap between understanding the theory and being able to use it under real conditions. The only way to close that gap is repetitive application, not more theory.
Application Sprints are structured practice sessions designed to take you from understanding to capability as efficiently as possible.
After the foundational concepts of a skill are in place, run this prompt:
I understand the theory of [skill/concept]. Now I need to
build actual capability. Design a 20-session practice
curriculum for me.
Each session should:
- Be complete in 20-30 minutes
- Build on the previous session
- Introduce one new constraint or challenge
- Include a specific success criterion so I know when I am done
Start with the simplest possible application and progress
to the most complex version of the real-world situation I
will face.
The key principle in this prompt is the progression from simple to complex and the explicit success criterion for each session. Both matter.
Starting simple reduces the cognitive load of early practice sessions, which means you can focus entirely on the skill being developed rather than on managing complexity. Success criteria tell you when you have actually completed the session rather than when you have stopped, which is a different thing.
Run one session per day during an Application Sprint. Twenty sessions at 25 minutes each is eight hours of deliberate practice. Eight hours of this kind of targeted, progressive application produces more capability than most people develop from fifty hours of passive study.
## Module 5: The Expert Interview Technique That Compresses Years
Every skill has knowledge that is not in any book or course. It exists in the practical experience of people who have spent years applying the skill in real conditions. This knowledge is usually tacit and difficult to transfer, which is why most courses do not contain it.
Claude has processed an enormous amount of text written by practitioners at every level of every skill. It cannot tell you things that nobody has ever written, but it can surface patterns across thousands of practitioners' experiences in a way that no single mentor or course can.
The Expert Interview technique uses Claude as a synthesizer of that accumulated practitioner knowledge.
```
I am learning [skill] and I want to learn from the experience
of expert practitioners. Interview me about where I am in my
learning, then give me the answers to these questions from
the perspective of someone with ten years of real-world
experience:
1. What do beginners almost always get wrong that they cannot
see from where they are?
2. What does the path from competent to genuinely skilled
actually look like in practice, not in theory?
3. What do you know now that you wish you had known when
you were at my current level?
4. What is the most common reason people plateau at
intermediate level and never reach advanced?
5. If you were starting from where I am right now, what
would you do differently?
```
The practitioner perspective question series surfaces the kind of insight that saves months of trial and error. The gap between what a skill looks like from the outside and what it looks like from the inside of expert practice is almost always significant, and it almost never shows up in beginner-level content.
Run this session once per skill, after you have enough foundational knowledge to understand the answers, but before you have spent months developing the wrong habits.
## Module 6: The Error Analysis Loop That Accelerates Mastery
Most people respond to mistakes in one of two ways. They either feel bad and try to avoid the feeling by moving on quickly, or they fix the specific mistake without understanding the category of error it represents.
Neither produces fast improvement. Fast improvement comes from understanding not just that you made a mistake but what type of mistake it was, why your reasoning led to it, and what change in thinking would prevent that entire category of error in the future.
After any practice session where you made significant mistakes, run this:
```
I just practiced [skill/task] and made the following errors:
[Describe the specific mistakes.
For each error, do not just correct it. Analyze it:
1. What category of error is this? (conceptual gap,
procedural error, application failure, judgment error)
2. What does this error reveal about my current mental
model that is incomplete or incorrect?
3. What is the correct mental model that would have
prevented this mistake?
4. What practice would cement the correct mental model?
```
The category analysis is what makes this prompt more valuable than a simple correction. Understanding that you made a conceptual gap error versus a judgment error tells you something different about what to do next. A conceptual gap requires going back to the theory. A judgment error means you understand the theory but have not yet built the pattern recognition that good judgment requires. These need different responses.
Run the Error Analysis Loop after any session where mistakes appeared. Over time, patterns in your error categories will emerge that tell you something important about how you learn this particular skill.
## Module 7: The Progress Milestone System That Keeps You Moving
The biggest risk in any learning system is stalling. Not quitting, not deciding the skill is not worth learning, but the gradual drift away from practice that happens when progress becomes invisible.
Progress in complex skills is often not linear. There are extended plateaus where your performance seems stuck regardless of practice, followed by sudden jumps in capability as underlying understanding reorganizes itself. These plateaus are the moments when most people abandon their learning, convinced they have hit a ceiling, just before the breakthrough that was forming underneath.
The Progress Milestone System keeps you moving through plateaus by making them legible.
Every two weeks, run this session:
```
I have been learning [skill] for [duration]. Here is a summary
of what I have covered and practiced: [brief summary].
Evaluate my progress honestly:
1. Where am I relative to where I was two weeks ago?
2. What specific capabilities do I have now that I did not
have then, even if the improvement feels invisible to me?
3. Am I on a plateau? If so, what typically causes this plateau
in learners at my stage and how do people get through it?
4. What is the most important thing I should focus on in the
next two weeks to maintain forward momentum?
5. What does the next visible milestone look like and how far
am I from it?
```
The third question is the most valuable. Claude can tell you what typically happens to learners at your specific stage of a specific skill, which provides context for the experience you are having that prevents the plateau from feeling like failure. Plateaus are a normal feature of skill acquisition.
Knowing that does not make them shorter but it makes them survivable.
## The Full System in Practice
These seven modules are not independent techniques. They form a system in which each component reinforces the others.
The diagnostic builds the map. The Socratic loop builds genuine understanding rather than familiarity. Spaced repetition prevents that understanding from disappearing. The Application Sprint converts understanding into capability. The Expert Interview compresses practitioner wisdom. The Error Analysis Loop accelerates improvement within practice sessions. The Progress Milestone System prevents the stalls that derail long-term learning.
Run the diagnostic once at the start and update it every few weeks. Run the Socratic loop in every conceptual learning session. Run spaced repetition reviews at three, seven, and fourteen day intervals. Run Application Sprints after each major conceptual section is complete. Run the Expert Interview once per skill. Run Error Analysis after any session where significant mistakes appeared. Run the Progress Milestone every two weeks.
AI improves completion rates by 70% and reduces dropout rates by 15% while simultaneously increasing student motivation. Employees complete training programs faster while demonstrating superior mastery and better retention when tested weeks or months later.
The system requires genuine effort. There is no version of learning any complex skill that does not require substantial practice and sustained attention. What the system eliminates is wasted effort: time spent on material you already know, passive consumption that produces familiarity without understanding, practice without diagnosis, and the long plateaus that happen when learners have no visibility into why they are stuck.
The same hours that a traditional approach turns into modest progress become genuinely productive learning. That is what half the time means: not that the skill requires half the effort, but that every unit of effort you put in produces twice the result.
## One Place to Start
If you take nothing else from this article, take this.
Open a conversation with Claude right now. Paste the diagnostic prompt from Module 1. Fill in the skill you have been meaning to learn. Answer the ten questions it asks honestly.
The learning map it produces in the next fifteen minutes will tell you more about what you actually need to learn than most courses tell you across their entire content.
Everything else in the system builds from that map.
The skill is not further away than you think. The gap between where you are and where you want to be is just less visible than it needs to be. The diagnostic makes it visible.
Start there. Build from what you find.
## 相关链接
- [NeilXbt](https://x.com/neil_xbt)
- [@neil_xbt](https://x.com/neil_xbt)
- [4.7K](https://x.com/neil_xbt/status/2052254328409231372/analytics)
- [Upgrade to Premium](https://x.com/i/premium_sign_up)
- [1:08 PM · May 7, 2026](https://x.com/neil_xbt/status/2052254328409231372)
- [4,748 Views](https://x.com/neil_xbt/status/2052254328409231372/analytics)
- [View quotes](https://x.com/neil_xbt/status/2052254328409231372/quotes)
---
*导出时间: 2026/5/7 21:42:24*
---
## 中文翻译
# 如何构建一个个人 AI 学习系统，让你用一半的时间掌握任何技能（完整教程）
**作者**: NeilXbt
**日期**: 2026-05-07T05:08:55.000Z
**来源**: [https://x.com/neil_xbt/status/2052254328409231372](https://x.com/neil_xbt/status/2052254328409231372)
---
大多数人学习新技能的方式，是现有方法中成本最高的一种。
他们找一门课程。他们从头到尾地学，完全不管自己已经懂了什么、还有什么不懂。他们和其他购买了同一门课程的每个人一样，以相同的节奏、通过相同的案例吸收相同的内容。他们学完了，感觉准备就绪，然后发现真正的技能缺口存在于课程从未触及的地方。
这就是“一刀切”的学习模式，对于任何碰巧不合尺寸的人来说，它都在系统地失效。
替代方案不是更好的课程。而是一个专门为你量身定制的个人学习系统，围绕你的既有知识、你的具体缺口、你的日程安排、你的学习风格以及你实际所需的技能版本而构建。如果这个系统构建得当并以 Claude 为核心，它能产生的效果是市场上最好的课程无法比拟的。
与传统方法相比，处于 AI 驱动学习环境中的学生，其考试成绩高出 54%，学习效果提升 30%，且参与度高出 10 倍。
2025 年发表在《科学报告》上的一项随机对照试验发现，AI 辅导的效果优于课堂主动学习，效应量在 0.73 到 1.3 个标准差之间。
本文就是完整的系统。每一个组件。每一个提示词。每一种练习方法。按照能产生结果的确切顺序排列。
## 为什么大多数学习系统在开始之前就注定失败
在构建一个更好的系统之前，你需要理解为什么标准方法会浪费这么多时间。
核心问题在于，传统学习将你的知识缺口视为未知数。一门课程无法知道你已经理解了什么、你即将理解什么以及你尚无任何语境的内容。因此，它按照为中位数学习者设定的节奏，按部就班地覆盖所有内容。对于已经理解了一半内容的人来说，这种方式既缓慢又令人丧失动力。对于在课程假定的基础知识方面存在重大缺口的人来说，如果没有额外的帮助，这是无法理解的。
不管怎样，时间都被浪费了。
自适应学习系统专注于知识缺口，而不是重复已知的内容。因为自适应学习系统会持续分析学习者的行为、技能和表现，以实时调整学习路径，它们能帮助更快地缩小技能差距，提高参与度，并大规模地提供定向学习。
这里描述的个人 AI 学习系统在设计上是自适应的。它首先准确诊断你所处的位置，构建从该特定起点到你特定目标的路径，并随着你的理解发展进行调整。不需要覆盖任何不需要覆盖的内容。不在你已经掌握的概念上花费任何时间。每一个环节都针对实际的缺口。
## 模块 1：构建你学习路径图的诊断
每一个有效的学习系统都不是从内容开始，而是从对当前知识的诚实评估开始。大多数人跳过这一步，因为感觉像是在拖延。事实恰恰相反。花在准确诊断上的时间，是在其他任何地方节省下来的时间。
打开与 Claude 的对话，在你开始学习任何东西之前运行这个提示词：
```
我想学习 [技能]。在我们开始之前，我想让你准确诊断我当前的知识水平。问我 10 个问题，测试我对该技能不同层次的理解，从基础概念到中级应用，再到高级的细微差别。
在我回答后，给我：
1. 对我知识扎实部分的诚实评估
2. 如果不首先解决将阻碍我进步的具体缺口
3. 我即将理解的概念，这些概念通过针对性的解释会很快豁然开朗
4. 根据我的起点预估的学习时间表
不要只是鼓励我。要准确。
```
指令“不要只是鼓励我，要准确”是这个提示词中最重要的部分。Claude 在评估某人知识时的默认模式是找出对方答对的部分，并委婉地表达缺口。为了构建学习系统，你需要未经修饰的版本。你需要确切知道你处于什么位置，而不是一个恭维性的近似值。
这个环节的输出就是你的学习路径图。它告诉你该覆盖什么、该跳过什么、在理解更高级的概念之前哪些需要基础工作，以及整个路径的现实时间表。随后的每一个环节都是建立在这张路径图之上的，而不是从一个通用的课程大纲开始。
每两到三周更新一次这张路径图。随着你的理解发展，缺口也会发生变化。路径图应该反映你现在的位置，而不是你开始时的位置。
## 模块 2：取代被动阅读的苏格拉底式学习循环
大多数人通过消费内容来学习。他们阅读、观看或聆听，并对材料建立了一种熟悉感，而他们误把这种熟悉感当成了理解。
熟悉不是理解。真正理解的考验在于，你能否利用这些知识解决一个以前从未见过的问题，能否向没有相关背景的人解释这个概念，以及能否识别出该概念在何时适用、何时不适用。被动消费产生熟悉感。主动生成产生理解。
苏格拉底式学习循环迫使你在每个环节的第一刻就进入主动生成状态。
在每个学习环节开始时运行这个提示词：
```
我正在学习 [技能]。今天我想攻克 [来自你学习路径图的具体概念]。
不要直接向我解释这个概念。相反，问我一系列问题，引导我自己去理解它。从我已经知道的内容开始，并以此为基础构建。当我出错或不完整时，提出一个追问，指引我走向缺口，而不是替我填补它。
在我通过你的问题弄清楚这个概念后，给我一个简洁的解释，总结我发现的内容，并填补我自己未达到的任何缺口。
```
这种方法每次环节比阅读解释需要更多的努力。它也能产生显著更好的记忆保持力和一种在质量上截然不同的理解。当你通过自己的推理（由问题引导）得出结论时，你不仅理解了什么是正确的，还理解了为什么它是正确的，以及它如何与你已知的知识相联系。
生成式 AI 在提供针对个人偏好和需求定制的自适应学习体验方面展现出卓越的能力。在各学科中利用不同形式的生成式 AI，能产生更优的学习成果。
这些环节感觉起来更难。这种难度就是学习本身。学习过程中的轻松感通常意味着没有学到多少东西。
## 模块 3：防止遗忘的间隔重复系统
学到某样东西并不等于记住它。关于记忆的科学结论很明确，并且已经被复现了一个多世纪：如果在战略间隔内没有进行刻意的重复，你学到的大部分内容几天内就会消失。
手动进行间隔重复的挑战在于后勤工作。追踪你什么时候学了什么、计算最佳复习间隔，以及实际执行复习环节，这是一笔巨大的开销，大多数人都会因此放弃。
Claude 处理后勤工作。你负责提取记忆。
在每个学习环节结束后，运行这个提示词：
```
我刚刚学完了 [今天涵盖的内容]。生成五个问题，测试我对本环节核心概念的理解。让它们是应用题，而不是定义题 —— 我应该必须运用这些知识，而不仅仅是回忆一个术语。
将这些保存为该概念的复习集。我会让你在 3 天后、然后 7 天后、然后 14 天后生成一个复习环节。
```
在每个复习日期，打开一个新环节：
```
[概念] 的复习环节，初次学习后 [X] 天。
生成你在我初次环节后创建的五个问题，外加两个新问题，将这个概念与我此后学到的内容联系起来。
```
关联问题是最有价值的部分。它们迫使你将新知识与现有知识整合，而不是将它们分隔在不同的隔间里，而深度理解正存在于这种整合之中。
你不需要闪卡应用。你不需要专门的间隔重复平台。你需要的是在正确的间隔运行复习环节的一致习惯。Claude 不会自动追踪任何东西，所以你需要手动追踪复习日期。为每个复习日期设置一个简单的日历提醒就足够了。
## 模块 4：构建真实能力的应用冲刺
没有应用的知识是永远不会转化的潜能。每一项技能在理解理论和能在真实条件下使用它之间都存在差距。缩小这一差距的唯一途径是重复应用，而不是更多的理论。
应用冲刺是结构化的练习环节，旨在以最有效率的方式将你从理解带向能力。
在掌握了一项技能的基础概念后，运行这个提示词：
我理解了 [技能/概念] 的理论。现在我需要建立实际能力。为我设计一个包含 20 个环节的练习课程。
每个环节应该：
- 在 20-30 分钟内完成
- 建立在上一个环节之上
- 引入一个新的约束或挑战
- 包含一个具体的成功标准，以便我知道何时完成
从最简单的可能应用开始，逐步推进到我将面对的现实世界情况的最复杂版本。
这个提示词的关键原则是从简单到复杂的递进，以及每个环节明确的标准。两者都很重要。
从简单的开始减少了早期练习环节的认知负荷，这意味着你可以完全专注于正在发展的技能，而不是管理复杂性。成功标准会告诉你何时真正完成了环节，而不是何时停止了，这是两回事。
在应用冲刺期间，每天进行一个环节。20 个环节，每个 25 分钟，就是 8 小时的刻意练习。8 小时这种针对性的、递进的应用，所产生的能力比大多数人通过 50 小时的被动学习所发展的能力还要强。
## 模块 5：浓缩数年经验的专家访谈法
每一项技能都有任何书籍或课程中不包含的知识。它存在于那些多年在真实条件下应用该技能的人的实践经验中。这种知识通常是隐性的，难以转移，这就是为什么大多数课程不包含它的原因。
Claude 处理了大量由各级别各技能的从业者撰写的文本。它无法告诉你那些从未有人写下来的事情，但它可以综合数千名从业者的经验模式，这是任何单一导师或课程都无法做到的。
专家访谈法利用 Claude 作为这些积累的从业者知识的综合器。
```
我正在学习 [技能]，我想向专家从业者的经验学习。面试我关于我学习所处的位置，然后从一个拥有 10 年现实世界经验的人的视角，给我这些问题的答案：
1. 初学者几乎总是弄错什么，而这是他们从自己的位置看不到的？
2. 从胜任到真正熟练，在实践中（而非理论上）的路径实际上是什么样的？
3. 你现在知道哪些是你处于我目前水平时希望知道的？
4. 人们在中级水平停滞不前且从未达到高级水平，最常见的原因是什么？
5. 如果你从我现在的地方开始，你会做哪些不同的选择？
```
从业者视角的问题系列能揭示出那种可以节省数月试错时间的见解。一项技能从外部看是什么样，与处于专家实践内部看是什么样，这两者之间的差距几乎总是巨大的，而且这种差距几乎从未出现在初学者内容中。
每个技能运行一次这个环节，在你拥有足够的基础知识来理解答案之后，但在你花费数月养成错误的习惯之前。
## 模块 6：加速精通的错误分析循环
大多数人以两种方式之一回应错误。他们要么感觉糟糕，试图通过快速转移话题来避免这种感觉；要么只是修复具体的错误，而不理解它所代表的错误类别。
这两种方式都不会带来快速的进步。快速进步源于理解不仅是你犯了错误，还有它是哪种类型的错误、你的推理为何导致了它，以及思维上的什么改变能在未来防止整个类别的错误。
在任何你犯下重大错误的练习环节后，运行这个：
```
我刚刚练习了 [技能/任务] 并犯了以下错误：[描述具体的错误]。
对于每个错误，不要只是纠正它。分析它：
1. 这是什么类型的错误？（概念缺口、程序性错误、应用失败、判断错误）
2. 这个错误揭示了我当前的心智模型中有哪些不完整或不正确的地方？
3. 能预防这个错误的正确心智模型是什么？
4. 什么练习能巩固正确的心智模型？
```
类别分析是使这个提示词比简单纠正更有价值的关键。理解你犯了概念缺口错误与判断错误，会告诉你关于下一步该做些什么的完全不同的信息。概念缺口需要回到理论。判断错误意味着你理解理论，但尚未建立良好判断所需的模式识别。这些需要不同的应对。
在任何出现错误的环节后运行错误分析循环。随着时间的推移，你的错误类别中的模式会浮现出来，告诉你关于你如何学习这项特定技能的重要信息。
## 模块 7：保持你前进的进度里程碑系统
任何学习系统中最大的风险是停滞。不是放弃，不是决定该技能不值得学，而是当进度变得不可见时，那种逐渐远离练习的 drift。
复杂技能的进步往往不是线性的。会有漫长的平台期，无论你如何练习，表现似乎都停滞不前，随后随着底层理解的重组，能力会出现突然的跳跃。这些平台期是大多数人放弃学习的时刻，他们确信自己遇到了天花板，而这恰恰发生在原本正在形成的突破之前。
进度里程碑系统通过使平台期变得清晰，让你能够穿越它们。
每两周运行一次这个环节：
```
我已经学习 [技能] [时长]。以下是我学习和练习内容的简要总结：[简要总结]。
诚实评估我的进度：
1. 相比两周前，我处于什么位置？
2. 我现在拥有哪些当时没有的具体能力，即使这种改进对我来说感觉并不明显？
3. 我处于平台期吗？如果是，对于我阶段的学习者来说，通常是什么导致了这种平台期，人们是如何度过它的？
4. 为了保持前进的动力，接下来两周我应该专注于的最重要的事情是什么？
5. 下一个可见的里程碑是什么样的，我距离它还有多远？
```
第三个问题是最有价值的。Claude 可以告诉你，对于特定技能的特定阶段，学习者通常会发生什么，这为你当下的体验提供了语境，从而防止平台期感觉像是失败。平台期是正常现