DEV Community: guanyi liu

ChatGPT 总说 "Great!"，但我什么都没记住——用 Claude Code 搭了一个真正的面试老师

guanyi liu — Sat, 30 May 2026 15:29:10 +0000

ChatGPT 总说 "Great!"，但我什么都没记住——用 Claude Code 搭了一个真正的面试老师

一个所有人的痛

你用 ChatGPT / Claude 准备面试的时候，有没有遇到过这种情况：

你答了一个明显有问题的答案，它回 "Great answer!" 然后直接下一题
每次开新对话，它不记得你昨天学了什么，你得从头介绍自己
你学了一堆八股，但不知道什么时候该复习，过两天全忘了
它给你讲概念，你看完觉得懂了，但面试时却张嘴说不出来

根本原因很简单：ChatGPT 是一个贴心的陪聊，不是一个严格的老师。

它的工作原则是"让用户满意"，而不是"让用户学会"。你说什么它都觉得好，因为它怕你不开心。

有没有办法让 AI 当严师？

有。我用 Claude Code（Anthropic 的 CLI 编程工具）做到了。

Claude Code 有一个独特能力：规则文件（CLAUDE.md）+ 持久化记忆系统。简单说，你可以通过写规则文件来"编程"AI 的行为，而且这些规则在每次对话中都会自动加载。

我把这个能力用在了面试准备上。

我做了什么

我写了一套配置文件，核心就两个东西：

1. 严格学习流程（rules/interview.md）

规则：
1. 混合出题（3-4个模块，不是只问一个方向）
2. 答对 → 简要确认 + 关联知识点 → 下一题
3. 答错 → 必须讲解（含参考资料URL）→ 出验证题 → 通过才出新题
4. 禁止跳步骤，不能答完直接出下一题

和普通ai 的区别：规则是写死的，不是 AI 自己决定的。它没法偷懒说 "Great"，因为规则文件告诉它必须走完讲解→验证的流程。

2. 遗忘曲线复习调度（knowledge/interview_tracker.md）

知识点	首次学习	D+1	D+2	D+4	D+7	D+15	状态
智能指针	5/29	✅	✅				已掌握
虚函数表	5/29	❌					薄弱

每次会话开始，Claude 自动检查哪些知识点到了复习时间，优先出复习题。你不需要自己记什么时候该复习。

3. 每日收尾验证

每天说"今天就到这"，触发收尾流程：

Claude 给出今日学习大纲
你根据大纲回忆，给出自己的理解
Claude 判断你是真懂了还是在背答案
更新复习调度

这是费曼学习法的 AI 版——不是你觉得自己懂了，而是你要能讲出来，AI 当裁判。

效果对比

能力	ChatGPT 面试	Anki	NotebookLM	SaaS 平台	本项目
跨会话记忆	无	仅卡组级	无	部分	有
间隔重复	无	有	无	无	有
答错→讲解→验证	无	无（只给答案）	无	部分	有
基于项目代码原生出题	需手动贴代码	不适用	需上传文件	不支持	原生支持
额外费用	无	无	无	\$29-300/次	无额外费用

核心差异：其他工具要么有记忆没理解（Anki），要么有理解没记忆（ChatGPT），要么两个都没有（SaaS）。 Claude Code 的规则文件 + 记忆系统是唯一把两者结合起来的方案。

最大的杀手锏：从你的项目代码出发

市面上所有面试工具都是"我出题你答"。

本项目的不同：从你自己的代码出发追问。

举个例子，我的项目里有一个 BoundedQueue（有界队列），用了两个 condition_variable。面试准备时，Claude 不是问你"请解释 condition_variable"，而是问：

"你的 BoundedQueue 用了两个 condition_variable（not_full_ 和 not_empty_），用一个行不行？"

这种问题逼你从自己写过的代码里思考原理，知识有锚点，不是空背八股。

快速开始

1. Fork 仓库

gh repo fork happiness-cheng/ai-interview-engine --clone

2. 用 Claude Code 打开

cd ai-interview-engine
claude

3. 说第一句话

我要准备 C++ 后端开发面试，目标是大厂实习。开始复习。

就是这样。Claude 会读取规则文件，开始第一轮出题。

4. 换成你自己的技术栈

仓库自带 C++ 后端的示例题库。如果你是 Java / Go / 前端，用空白模板替换即可：

knowledge/TEMPLATE.md → 复制为 interview_tracker.md → 填入你的知识点

我的使用体验

我是一个大二学生，正在准备 2026 年秋招实习。用这套系统准备 C++ 后端面试，最明显的感受是：

不焦虑了 — 不用自己规划"今天复习什么"，系统按遗忘曲线自动调度
真的记住了 — 答错→讲解→验证的流程逼你真的理解，不是过眼云烟
项目不心虚了 — 从自己代码出发的追问，让你对项目里的每个设计决策都能说清楚"为什么"

这不只是面试工具

底层引擎是通用的：出题→回答→评判→讲解→验证→遗忘曲线调度。

面试只是一个场景。你可以用它准备：

技术面试（当前支持）
转行/跳槽复习
期末考试（原理相同，换知识点库）

开源地址

GitHub: https://github.com/happiness-cheng/ai-interview-engine

如果对你有帮助，给个 star。如果有改进建议，欢迎提 issue。

From "Great!" to Real Learning: Building an AI Interview Teacher with Claude Code

guanyi liu — Sat, 30 May 2026 14:38:06 +0000

ChatGPT Always Says "Great!" But I Remember Nothing — Built a Real Interview Teacher with Claude Code

A Universal Pain Point

Have you ever experienced this when using ChatGPT / Claude to prepare for interviews:

You give an obviously flawed answer, and it replies "Great answer!" then moves straight to the next question
Each time you start a new conversation, it doesn't remember what you learned yesterday, forcing you to reintroduce yourself
You memorize a bunch of canned answers, but don't know when to review them, and forget everything after two days
It explains concepts to you, you feel like you understand after reading, but when facing an interview, you can't articulate them

The root cause is simple: ChatGPT is an attentive chat companion, not a strict teacher.

Its operating principle is "make the user happy," not "make the user learn." It praises whatever you say because it fears you'll become unhappy.

Is There a Way to Make AI a Strict Teacher?

Yes. I achieved this using Claude Code (Anthropic's CLI programming tool).

Claude Code has a unique capability: Rules Files (CLAUDE.md) + Persistent Memory System. In simple terms, you can "program" the AI's behavior by writing rules files, and these rules are automatically loaded in every conversation.

I applied this capability to interview preparation.

What I Did

I created a set of configuration files, with two core components:

1. Strict Learning Process (rules/interview.md)

Rules:
1. Mix questions from different modules (3-4 areas, not just one direction)
2. Correct answer → Brief confirmation + Link to related knowledge → Next question
3. Wrong answer → Must explain (include reference URL) → Give verification question → Only proceed after passing
4. No skipping steps; cannot jump straight to the next question after answering

Compared to regular AI: Rules are hardcoded, not decided by the AI itself. It can't get lazy and say "Great" because the rules file forces it to complete the explanation → verification process.

2. Spaced Repetition Review Scheduling (knowledge/interview_tracker.md)

Knowledge Point	First Learned	D+1	D+2	D+4	D+7	D+15	Status
Smart Pointers	5/29	✅	✅				Mastered
Virtual Table	5/29	❌					Weak

At the start of each session, Claude automatically checks which knowledge points are due for review and prioritizes review questions. You don't need to remember when to review yourself.

3. Daily Wrap-up Verification

When you say "That's it for today," it triggers the wrap-up process:

Claude provides a summary outline of today's learning
You recall and provide your understanding based on the outline
Claude judges whether you truly understand or are just reciting answers
Updates the review schedule

This is the AI version of the Feynman Technique — not just feeling like you understand, but being able to explain it, with AI as the judge.

Feature Comparison

Capability	ChatGPT Interview Prep	Anki	NotebookLM	SaaS Platforms	This Project
Cross-session Memory	No	Deck-level only	No	Partial	Yes
Spaced Repetition	No	Yes	No	No	Yes
Wrong→Explain→Verify	No	No (just gives answer)	No	Partial	Yes
Native Question Generation from Project Code	Manual code pasting required	N/A	File upload required	Not supported	Native Support
Additional Costs	None	None	None	$29-300/session	No additional costs

Core difference: Other tools either have memory without understanding (Anki), or understanding without memory (ChatGPT), or lack both (SaaS). Claude Code's rules files + memory system is the only solution that combines both.

The Biggest Killer Feature: Starting from Your Project Code

All interview tools on the market follow the "I ask, you answer" model.

This project is different: Asks follow-up questions based on your own code.

For example, my project has a BoundedQueue (bounded queue) using two condition_variable. During interview preparation, Claude won't ask you "Please explain condition_variable," but rather:

"Your BoundedQueue uses two condition_variables (not_full_ and not_empty_). Could you use just one?"

Such questions force you to think about principles from code you've actually written, giving knowledge an anchor point rather than empty memorization of canned answers.

Quick Start

1. Fork the Repository

gh repo fork happiness-cheng/ai-interview-engine --clone

2. Open with Claude Code

cd ai-interview-engine
claude

3. Say Your First Words

I'm preparing for C++ backend development interviews, aiming for internship positions at top companies. Start reviewing.

That's it. Claude will read the rules files and begin the first round of questions.

4. Switch to Your Own Tech Stack

The repository includes sample question banks for C++ backend. If you're using Java / Go / Frontend, replace with the blank template:

knowledge/TEMPLATE.md → Copy as interview_tracker.md → Fill in your knowledge points

My User Experience

I'm a sophomore student preparing for the 2026 fall recruitment internship. Using this system for C++ backend interview preparation, the most noticeable benefits are:

No more anxiety — No need to plan "what to review today" myself; the system automatically schedules based on the forgetting curve
Actually remember — The wrong→explain→verify process forces genuine understanding, not just fleeting impressions
Confident about projects — Follow-up questions based on your own code let you clearly explain the "why" behind every design decision in your project

This Is Not Just an Interview Tool

The underlying engine is generic: Question generation → Answer → Evaluation → Explanation → Verification → Forgetting curve scheduling.

Interviews are just one scenario. You can use it to prepare for:

Technical interviews (currently supported)
Career transition/job change review
Final exams (same principle, different knowledge base)

Open Source Repository

GitHub: https://github.com/happiness-cheng/ai-interview-engine

If this helps you, give it a star. If you have suggestions for improvement, feel free to open an issue.

Stop Saying 'Great!'—Build a Real AI Interview Coach with Claude Code

guanyi liu — Sat, 30 May 2026 14:05:58 +0000

ChatGPT always says "Great!", but I remember nothing — I built a real interview tutor with Claude Code

A universal pain point

When using ChatGPT / Claude to prepare for interviews, have you ever encountered these situations:

You gave a clearly flawed answer, it replied "Great answer!" and moved directly to the next question
Every time you start a new chat, it doesn't remember what you studied yesterday, you have to introduce yourself from scratch
You memorized a bunch of standard answers, but don't know when to review, so you forget everything after a couple of days
It explains a concept, you think you understand after reading, but when it's interview time, you can't articulate it

The root cause is simple: ChatGPT is an accommodating companion, not a strict teacher.

Its operational principle is "keep the user satisfied," not "help the user learn." It agrees with everything you say because it fears upsetting you.

Is there a way to make AI a strict tutor?

Yes. I achieved this using Claude Code (Anthropic's CLI programming tool).

Claude Code has a unique capability: Rules files (CLAUDE.md) + persistent memory system. In simple terms, you can "program" the AI's behavior by writing rules files, and these rules are automatically loaded in every conversation.

I applied this capability to interview preparation.

What I Did

I wrote a set of configuration files, with two core components:

1. Strict Learning Process (rules/interview.md)

Rules:
1. Mixed questioning (3-4 modules, not just one topic)
2. Correct answer → Brief confirmation + Related knowledge points → Next question
3. Wrong answer → Must explain (including reference URL) → Verification question → Proceed only after passing
4. No skipping steps, cannot jump directly to the next question after answering

Compared to standard AI: Rules are fixed, not AI's own decision. It cannot be lazy and say "Great" because the rules file dictates it must complete the explanation → verification process.

2. Forgetting Curve Review Scheduling (knowledge/interview_tracker.md)

Knowledge Point	First Learned	D+1	D+2	D+4	D+7	D+15	Status
Smart Pointers	5/29	✅	✅				Mastered
Virtual Table	5/29	❌					Weak

At the start of each session, Claude automatically checks which knowledge points are due for review and prioritizes review questions. You don't need to remember when to review yourself.

3. Daily Closing Verification

Saying "That's it for today" triggers the closing process:

Claude provides a summary of today's study outline
Based on the outline, you recall and give your own understanding
Claude judges whether you truly understand or are just reciting answers
Update the review schedule

This is the AI version of the Feynman learning technique — it's not about you thinking you understand, but about you being able to explain it, with AI as the judge.

Feature Comparison

Capability	ChatGPT Interview	Anki	NotebookLM	SaaS Platforms	This Project
Cross-session memory	None	Deck-level only	None	Partial	Yes
Spaced repetition	None	Yes	None	None	Yes
Wrong → Explain → Verify	None	None (Answers only)	None	Partial	Yes
Project-based native questions	Manual code pasting	N/A	File upload required	Not supported	Native support
Additional cost	None	None	None	\$29-300/session	No extra cost

Core difference: Other tools either have memory without understanding (Anki), or have understanding without memory (ChatGPT), or lack both (SaaS). Claude Code's rules file + memory system is the only solution that combines the two.

The Biggest Killer Feature: Starting from Your Project Code

All interview tools on the market follow a "I ask, you answer" model.

What this project does differently: Starts questioning from your own code.

For example, my project has a BoundedQueue (bounded queue), using two condition_variable. During interview preparation, Claude doesn't ask "Please explain condition_variable," but instead asks:

"Your BoundedQueue uses two condition_variables (not_full_ and not_empty_), would one be sufficient?"

Such questions force you to think about principles from code you've actually written, giving knowledge an anchor point instead of rote memorization of standard answers.

Quick Start

1. Fork the Repository

gh repo fork happiness-cheng/ai-interview-engine --clone

2. Open with Claude Code

cd ai-interview-engine
claude

3. Say the First Sentence

I want to prepare for C++ backend developer interviews, targeting internships at major companies. Let's start reviewing.

That's it. Claude will read the rules files and begin the first round of questions.

4. Switch to Your Own Tech Stack

The repository comes with sample question banks for C++ backend. If you're using Java / Go / Frontend, simply replace with the blank template:

knowledge/TEMPLATE.md → Copy to interview_tracker.md → Fill in your knowledge points

My User Experience

I'm a sophomore student preparing for the 2026 fall internship recruitment. Using this system for C++ backend interview preparation, the most noticeable effects are:

Less anxiety — No need to plan "what to review today," the system automatically schedules based on the forgetting curve
Actually remembering — The wrong → explain → verify process forces genuine understanding, not fleeting impressions
Confidence in projects — Questions starting from your own code make you able to clearly explain the "why" behind every design decision in your project

This is Not Just an Interview Tool

The underlying engine is universal: Question → Answer → Judge → Explain → Verify → Forgetting Curve Scheduling.

Interview preparation is just one scenario. You can use it for:

Technical interviews (currently supported)
Career change/job switch review
Final exams (same principle, just change the knowledge base)

Open Source Address

GitHub: https://github.com/happiness-cheng/ai-interview-engine

If this helps you, give it a star. If you have suggestions for improvement, feel free to open an issue.

Why ChatGPT Fails as an Interview Tutor (And How I Built a Better One with Claude Code)

guanyi liu — Sat, 30 May 2026 13:16:19 +0000

ChatGPT Always Says "Great!", But I Remember Nothing — I Built a Real Interview Teacher with Claude Code

A Universal Pain Point

When you use ChatGPT / Claude to prepare for interviews, have you ever encountered this situation:

You gave an obviously flawed answer, and it replied "Great answer!" then moved straight to the next question.
Every time you start a new conversation, it doesn't remember what you learned yesterday, so you have to introduce yourself from scratch.
You've memorized a bunch of standard answers, but you don't know when to review them, and forget everything after a couple of days.
It explains a concept to you, and you feel like you understand it after reading, but during an interview, you can't articulate it.

The root cause is simple: ChatGPT is an attentive chat companion, not a strict teacher.

Its working principle is "make the user satisfied," not "make the user learn." It thinks everything you say is great because it's afraid you'll be unhappy.

Is There a Way to Make AI a Strict Teacher?

Yes. I did it using Claude Code (Anthropic's CLI coding tool).

Claude Code has a unique capability: Rule files (CLAUDE.md) + Persistent memory system. In simple terms, you can "program" the AI's behavior by writing rule files, and these rules are automatically loaded in every conversation.

I applied this capability to interview preparation.

What I Did

I wrote a set of configuration files, with two core components:

1. Strict Learning Process (rules/interview.md)

Rules:
1. Mixed question sets (3-4 modules, not just one direction)
2. Correct answer → Brief confirmation + Related knowledge points → Next question
3. Wrong answer → Must explain (including reference URLs) → Verification question → New question only after passing
4. No skipping steps. Cannot ask the next question immediately after an answer.

The difference from a regular AI: The rules are hard-coded, not decided by the AI itself. It can't be lazy and say "Great," because the rule file tells it it must complete the explanation → verification process.

2. Forgetting Curve Review Scheduling (knowledge/interview_tracker.md)

Knowledge Point	First Learned	D+1	D+2	D+4	D+7	D+15	Status
Smart Pointers	5/29	✅	✅				Mastered
Virtual Function Table	5/29	❌					Weak

At the start of each session, Claude automatically checks which knowledge points are due for review and prioritizes review questions. You don't need to remember when to review yourself.

3. Daily Wrap-up Verification

Say "That's it for today" to trigger the wrap-up process:

Claude provides a summary of today's learning outline.
You recall and give your own understanding based on the outline.
Claude judges whether you truly understand or are just reciting answers.
The review schedule is updated.

This is the AI version of the Feynman Technique — not just feeling like you understand, but being able to explain it, with AI as the judge.

Effectiveness Comparison

Capability	ChatGPT Interview	Anki	NotebookLM	SaaS Platforms	This Project
Cross-session Memory	None	Deck-level only	None	Partial	Yes
Spaced Repetition	None	Yes	None	None	Yes
Wrong → Explain → Verify	None	No (Only gives answer)	None	Partial	Yes
Native Question Generation from Project Code	Need to manually paste code	N/A	Requires file upload	Not supported	Native Support
Additional Cost	None	None	None	\$29-300/time	No Extra Cost

The core difference: Other tools either have memory without understanding (Anki), or understanding without memory (ChatGPT), or neither (SaaS). Claude Code's rule file + memory system is the only solution that combines both.

The Biggest Killer Feature: Starting from Your Project Code

All interview tools on the market are "I ask, you answer."

This project is different: Asking follow-up questions starting from your own code.

For example, my project has a BoundedQueue (Bounded Queue) using two condition_variable. During interview prep, Claude doesn't ask you "Please explain condition_variable," but rather:

"Your BoundedQueue uses two condition_variables (not_full_ and not_empty_). Can you use just one?"

This kind of question forces you to think about the principles from code you've written yourself, giving the knowledge an anchor point, not just memorizing standard answers.

Quick Start

1. Fork the Repository

gh repo fork happiness-cheng/ai-interview-engine --clone

2. Open with Claude Code

cd ai-interview-engine
claude

3. Say the First Sentence

I want to prepare for a C++ backend developer interview, aiming for a big tech internship. Let's start reviewing.

That's it. Claude will read the rule files and start the first round of questions.

4. Switch to Your Own Tech Stack

The repository comes with an example question bank for C++ backend. If you use Java / Go / Frontend, replace it with the blank template:

knowledge/TEMPLATE.md → Copy to interview_tracker.md → Fill in your knowledge points

My User Experience

I'm a sophomore student preparing for the 2026 fall internship recruitment. Using this system for C++ backend interview prep, the most noticeable feelings are:

No more anxiety — No need to plan "what to review today" myself; the system automatically schedules based on the forgetting curve.
Actually remembered — The wrong answer → explanation → verification process forces genuine understanding, not just fleeting recognition.
No more fear about my project — Follow-up questions from my own code let me clearly explain the "why" behind every design decision in my project.

This is Not Just an Interview Tool

The underlying engine is general-purpose: Question → Answer → Judge → Explain → Verify → Forgetting Curve Scheduling.

Interviews are just one scenario. You can use it to prepare for:

Technical Interviews (Currently supported)
Career change/job hop review
Final Exams (Same principle, swap the knowledge base)

Open Source Address

GitHub: https://github.com/happiness-cheng/ai-interview-engine

If it helps you, give it a star. If you have suggestions for improvement, feel free to open an issue.

Built an AI Interview Coach Because ChatGPT Just Says 'Great!

guanyi liu — Sat, 30 May 2026 13:08:35 +0000

ChatGPT Always Says "Great!" But I Remember Nothing — Building a Real Interview Teacher with Claude Code

A Universal Pain Point

When using ChatGPT / Claude to prepare for interviews, have you ever encountered these situations:

You give an obviously flawed answer, and it replies "Great answer!" then immediately moves to the next question.
Every time you start a new conversation, it doesn’t remember what you studied yesterday, forcing you to re-introduce yourself.
You’ve memorized a bunch of rote answers, but don’t know when to review them, forgetting everything after a couple of days.
It explains a concept, you feel like you understand after reading, but when it’s interview time, you can’t articulate it.

The root cause is simple: ChatGPT is a considerate chat companion, not a strict teacher.

Its operating principle is "keep the user satisfied," not "ensure the user learns." It agrees with everything you say because it fears upsetting you.

Is There a Way to Make AI a Strict Teacher?

Yes. I achieved this using Claude Code (Anthropic’s CLI coding tool).

Claude Code has a unique capability: rule files (CLAUDE.md) + a persistent memory system. In short, you can "program" the AI’s behavior by writing rule files, and these rules are automatically loaded in every conversation.

I applied this capability to interview preparation.

What I Did

I wrote a set of configuration files, with two core components:

1. Strict Learning Process (rules/interview.md)

Rules:
1. Mixed question selection (3-4 modules, not just one direction)
2. Correct answer → Brief confirmation + link to related knowledge points → Next question
3. Wrong answer → Must explain (including reference URL) → Provide a verification question → Only proceed after passing
4. No skipping steps; cannot immediately go to the next question after answering

The difference from regular AI: The rules are hard-coded, not decided by the AI itself. It can’t be lazy and say "Great," because the rule file mandates it must complete the explanation → verification process.

2. Forgetting Curve Review Scheduling (knowledge/interview_tracker.md)

Knowledge Point	First Learned	D+1	D+2	D+4	D+7	D+15	Status
Smart Pointers	5/29	✅	✅				Mastered
Vtable	5/29	❌					Weak

At the start of each session, Claude automatically checks which knowledge points are due for review and prioritizes those questions. You don’t need to remember when to review yourself.

3. Daily Wrap-up Verification

Say "that’s it for today" to trigger the wrap-up process:

Claude provides a summary of today’s learning outline.
Based on the outline, you recall and give your own understanding.
Claude judges whether you truly understand or are just reciting answers.
The review schedule is updated.

This is the AI version of the Feynman Technique—it’s not about you feeling like you understand, but about being able to explain it, with AI as the judge.

Feature Comparison

Capability	ChatGPT Interview	Anki	NotebookLM	SaaS Platforms	This Project
Cross-session Memory	None	Deck-level only	None	Partial	Yes
Spaced Repetition	None	Yes	None	None	Yes
Wrong Answer → Explain → Verify	None	No (only provides answer)	None	Partial	Yes
Native Question Generation from Project Code	Need to manually paste code	Not applicable	Requires file upload	Not supported	Native Support
Additional Costs	None	None	None	\$29-300/session	No extra costs

Core difference: Other tools either have memory without understanding (Anki), understanding without memory (ChatGPT), or neither (SaaS). Claude Code’s rule file + memory system is the only solution that combines both.

The Biggest Killer Feature: Starting from Your Project Code

All interview tools on the market operate on a "I ask, you answer" basis.

This project’s difference: It probes based on your own code.

For example, my project has a BoundedQueue (bounded queue) using two condition_variable. During interview preparation, Claude doesn’t ask "Please explain condition_variable," but instead asks:

"Your BoundedQueue uses two condition_variables (not_full_ and not_empty_). Can you use just one?"

Such questions force you to think about principles from code you’ve written, giving knowledge an anchor point, not just empty rote memorization.

Quick Start

1. Fork the Repository

gh repo fork happiness-cheng/ai-interview-engine --clone

2. Open with Claude Code

cd ai-interview-engine
claude

3. Say the First Sentence

I am preparing for a C++ backend development interview, targeting an internship at a top-tier company. Let’s start reviewing.

That’s it. Claude will read the rule files and start the first round of questions.

4. Switch to Your Own Tech Stack

The repository comes with a sample question bank for C++ backend. If you are using Java / Go / Frontend, replace it with a blank template:

knowledge/TEMPLATE.md → Copy to interview_tracker.md → Fill in your knowledge points

My User Experience

I am a sophomore student preparing for the 2026 fall internship recruitment. Using this system to prepare for C++ backend interviews, the most noticeable feelings are:

Less anxiety — No need to self-plan "what to review today"; the system automatically schedules based on the forgetting curve.
Actually remembered — The wrong answer → explanation → verification process forces genuine understanding, not fleeting recognition.
More confident about projects — Probing from your own code lets you clearly explain the "why" behind every design decision in your project.

This Is Not Just an Interview Tool

The underlying engine is general-purpose: Question → Answer → Judge → Explain → Verify → Forgetting Curve Scheduling.

Interviews are just one scenario. You can use it to prepare for:

Technical Interviews (currently supported)
Career Transition / Job Change Review
Final Exams (Same principle, just swap the knowledge base)

Open Source Address

GitHub: https://github.com/happiness-cheng/ai-interview-engine

If this helps you, give it a star. If you have suggestions for improvement, feel free to open an issue.

When ChatGPT Says 'Great!' But Teaches Nothing—Building a Real Interview Coach with Claude Code

guanyi liu — Sat, 30 May 2026 11:54:50 +0000

ChatGPT always says "Great!" but I remember nothing—Built a real interview teacher with Claude Code

A pain point for everyone

When you use ChatGPT / Claude to prepare for interviews, have you ever encountered this:

You gave an obviously flawed answer, it replied "Great answer!" and moved straight to the next question
Every time you start a new conversation, it doesn't remember what you learned yesterday, you have to introduce yourself from scratch
You learned a bunch of rote answers, but don't know when to review them, and forget everything in two days
It explains concepts to you, you feel like you understand after reading, but during interviews you can't articulate it

The fundamental reason is simple: ChatGPT is a considerate chat companion, not a strict teacher.

Its working principle is "make the user satisfied," not "help the user learn." It agrees with everything you say because it's afraid you might get upset.

Is there a way to make AI act as a strict teacher?

Yes. I did it using Claude Code (Anthropic's CLI programming tool).

Claude Code has a unique capability: rules files (CLAUDE.md) + persistent memory system. In simple terms, you can "program" the AI's behavior by writing rules files, and these rules are automatically loaded in every conversation.

I applied this capability to interview preparation.

What I did

I wrote a set of configuration files, with two core components:

1. Strict learning process (rules/interview.md)

Rules:
1. Mixed questions (3-4 modules, not just one direction)
2. Correct answer → Brief confirmation + related knowledge points → Next question
3. Wrong answer → Must explain (including reference URL) → Give verification question → New question only after passing
4. No skipping steps, cannot jump to next question immediately after answering

Difference from regular AI: The rules are hardcoded, not decided by the AI itself. It can't be lazy and say "Great" because the rules file dictates it must complete the explanation → verification process.

2. Forgetting curve review scheduling (knowledge/interview_tracker.md)

Knowledge Point	First Learned	D+1	D+2	D+4	D+7	D+15	Status
Smart Pointers	5/29	✅	✅				Mastered
Virtual Function Table	5/29	❌					Weak

At the start of each session, Claude automatically checks which knowledge points are due for review and prioritizes review questions. You don't need to remember when to review yourself.

3. Daily wrap-up verification

When you say "let's end here for today," it triggers the wrap-up process:

Claude provides today's learning outline
You recall based on the outline and give your own understanding
Claude judges whether you truly understand or just memorized answers
Update the review schedule

This is the AI version of the Feynman Technique—not just thinking you understand, but being able to explain it, with AI as the judge.

Effect comparison

Capability	ChatGPT Interview	Anki	NotebookLM	SaaS Platforms	This Project
Cross-session memory	None	Deck level only	None	Partial	Yes
Spaced repetition	None	Yes	None	None	Yes
Wrong→Explain→Verify	None	None (answer only)	None	Partial	Yes
Question generation from project code	Manual code pasting	Not applicable	File upload required	Not supported	Native support
Additional cost	None	None	None	\$29-300/time	No additional cost

Core difference: Other tools either have memory without understanding (Anki), or understanding without memory (ChatGPT), or neither (SaaS). Claude Code's rules file + memory system is the only solution that combines both.

The biggest killer feature: Starting from your project code

All interview tools on the market are "I ask, you answer."

This project is different: Asking follow-up questions based on your own code.

For example, my project has a BoundedQueue (bounded queue) using two condition_variable. During interview preparation, Claude doesn't ask "Please explain condition_variable," but rather:

"Your BoundedQueue uses two condition_variables (not_full_ and not_empty_), would one be enough?"

Such questions force you to think about principles from code you've written yourself, giving knowledge an anchor point rather than memorizing abstract answers.

Quick Start

1. Fork the repository

gh repo fork happiness-cheng/ai-interview-engine --clone

2. Open with Claude Code

cd ai-interview-engine
claude

3. Say the first sentence

I'm preparing for C++ backend development interviews, targeting internships at major companies. Let's start reviewing.

That's it. Claude will read the rules files and start the first round of questions.

4. Switch to your own tech stack

The repository comes with example question banks for C++ backend. If you're using Java / Go / frontend, replace with the blank template:

knowledge/TEMPLATE.md → Copy to interview_tracker.md → Fill in your knowledge points

My user experience

I'm a sophomore student preparing for 2026 fall internship recruitment. Using this system for C++ backend interview preparation, the most obvious feelings are:

No more anxiety — No need to plan "what to review today," the system automatically schedules based on the forgetting curve
Actually remember — The wrong→explain→verify process forces real understanding, not just fleeting impressions
Project confidence — Follow-up questions from your own code let you clearly explain the "why" behind every design decision

This isn't just an interview tool

The underlying engine is universal: Question→Answer→Judge→Explain→Verify→Forgetting curve scheduling.

Interview preparation is just one scenario. You can use it to prepare for:

Technical interviews (currently supported)
Career change/job transition review
Final exams (same principle, different knowledge base)

Open source link

GitHub: https://github.com/happiness-cheng/ai-interview-engine

If this helps you, give it a star. If you have improvement suggestions, feel free to open an issue.