up Documentation

An advanced guide which might benefit you a lot 🎉 . 人生进阶指南 离谱的人生 离谱的英语学习指南/英语学习教程/英语学习/学英语

55,183 stars JavaScript JavaScript 1 file · ~21,181 tokens #chinese#english-learning#tutorial
RAW Doc

File: README.md

title: 人生进阶指南 updated: 2026-08-16

人生进阶指南

中文 | English

这是一份写给普通人的长期成长手册。它从 2017 年的《离谱的英语学习指南》出发，逐渐扩展到 AI 学习、工作能力、人生复盘、创业与恢复。目标不是提供更多值得收藏的工具，而是帮助你完成一个可检查的循环：诊断现状 → 选择任务 → 主动练习 → 获得反馈 → 留下证据 → 调整下一步。

这里不会承诺捷径，也不会把个人经验包装成普遍规律。文章会尽量区分三类信息：

今天就开始

  1. 用 英语能力诊断 或自己的真实任务确定起点。
  2. 把目标、证据和下一步写进 学习状态。
  3. 只选一个 25–45 分钟任务，完成可检查的输出，而不是继续收集资料。
  4. 一周后用 每周复盘 看证据，再调整难度与频率。

英语学习系统

先读 CEFR 目标与自测，用“我能完成什么任务”而不是模糊的词汇量或考试印象判断水平。

环节 章节 本轮应留下的证据
校准 认知篇 一个具体场景、基线样本和 12 周目标
词汇 词汇篇 20 个能听懂、读懂并在句子中使用的词块
听力 听力篇 一段听写、复述和错误分类
阅读 阅读篇 一页批注、五句摘要和一个迁移问题
口语 口语篇 录音、转写和第二版重录
写作 写作篇 初稿、反馈清单和修订稿
AI 用 AI 学英语 可复用提示词、错题记录和学习状态更新

词表只是查阅材料，不是学习目标。需要时可进入 通用词表 与各技术主题词表。

AI 学习与工作能力

使用 AI 学习一切采用按任务选择的工作流：

  • 需要循序提问、知识检查时，使用带引导学习能力的对话；
  • 需要长期材料、作品、错误和上下文时，使用项目空间并维护本地状态文件；
  • 需要有来源的研究时，优先选择能回到原始材料的检索或资料库；
  • 需要最终交付时，由人定义标准、检查事实并承担判断责任。

产品功能变化很快。涉及 ChatGPT、Gemini、Claude、NotebookLM、Perplexity、DeepL 等产品的说明都会标注核验日期、地区/套餐限制和官方来源。工具排名不会成为方法的中心。

人生复盘、创业与恢复

个人故事被保留，是因为成长从来不只发生在学习计划里。你可以阅读 我的故事、创业篇 和 旧文归档。这些内容是作者的经历，不是医疗、法律、投资或创业建议。

公开版本遵循最少必要原则：不再展示个人 QQ、二维码和不必要的第三方身份信息；涉及家人、儿童、前伴侣或其他第三方的照片，应在明确授权后才公开。发现隐私问题请按 安全政策 私下报告。

作者项目与透明说明

作者参与的产品、公司参访与现实项目统一放在 作者项目与现实实践。页面会明确作者关联、用途、更新时间和“非赞助”声明。正文不因商业关系改变推荐标准，站点默认不接入广告、分析脚本或追踪器。

如何判断自己在进步

不要只统计学习时长。每周至少观察四类证据：

  • 完成：是否完成计划中的真实任务；
  • 质量：正确率、可理解度、结构、流利度或反馈等级是否改善；
  • 保持：隔几天后能否主动回忆并再次使用；
  • 迁移：能否把所学用到新材料、新对话或真实工作中。

当表现稳定提高时增加间隔或难度；当错误持续重复时缩小任务、补足先修知识并增加反馈。计划服务于表现，而不是让表现服从日历。

项目边界

  • 本项目是“开放内容项目”，不是 OSI 意义上的开源软件：正文与作者内容采用 CC BY-NC 4.0，站点配置、检查脚本和构建代码采用 MIT。详见 许可证说明。
  • 引用、图片和第三方素材的来源及授权状态记录在 第三方素材与引用。
  • 贡献前请阅读 贡献指南 和 行为准则。
  • 页面中的产品信息核验于 2026-08-16；过期内容欢迎提交 issue。

在线阅读

如果今天只做一件事：创建一份 学习状态，写下当前证据和下一步，然后完成其中最小的一项。

File: docs/en/threads/part-4/my-story.md

title: "My Story: Failure, Recovery, and Starting Again" updated: 2026-08-16

My Story: Failure, Recovery, and Starting Again

What Does It Feel Like to Spend 20 Million RMB on a Software Company and Watch It Collapse?

Hi, I’m Lipu. This is the story of one particularly absurd chapter of my life in the internet industry.

On April 1, 2017, I joined a tech company in Nanjing as a regular front-end developer. We were building software for lawyers and law firms. The first version had been outsourced. The code had no comments, no project management, and no real structure. To make things worse, there was no product manager. Most product ideas came directly from the boss’s intuition.

In my first month there, the timeline module I was responsible for was revised more than twenty times. That level of churn was exhausting.

Still, a month later, the first version finally went into testing and roughly matched what the boss wanted. After that, I led a front-end rewrite based on Vue. We spent more than half a year rebuilding it and eventually shipped. Sales started, and we thought the product would at least make some noise in the industry. It did not. For a long time, we could not close a single deal.

As the company struggled, salaries did not grow with time. A lot of people left. Watching old coworkers leave one after another, often with deep resentment toward the company, felt terrible.

From the beginning, I truly believed in the product. Even as the team turned over again and again, and even though my pay did not increase at all in those early years, I chose to stay.

Eventually, I became one of only two people who stayed throughout the whole journey. I went from an ordinary employee to department leader, then general manager, and step by step toward becoming a partner.

That was how the absurd story really began.

To devote myself fully to work, there was one part of life I cannot leave out: love.

Love

I was lucky enough to meet someone I truly loved at the right age. Her name was Ruohan (a pseudonym). In this guide, then called English-level-up-tips, I once used Gu Cheng’s words to describe how I felt when I first met her:

The grass is bearing its seeds.
The wind is shaking its leaves.
We stand together, saying nothing, and that is already beautiful.

On August 19, 2017, I went with her to Shanghai for something deeply significant. We took many photos on the Bund that day. Even now, whenever I see those photos, I still feel like crying. To this day, if they appear on my phone, I quickly scroll past them on purpose. I still do not have the courage to relive that pain.

That night was beautiful, but the air was heavy with pressure. Looking at the rows of bank buildings on the Bund and enjoying the night breeze, we both tried hard not to let the other person feel our emotions sinking, because we both knew that something was happening and there was nothing we could do about it.

At one point, while I was talking with great confidence about our bright future, she suddenly cried. Not loudly. Silently. I only realized it after she had been crying for a while. Tears and snot were running down her face, as if all the good things I was describing would never belong to her. She turned away from me, and in that moment I realized how fragile and thin she looked.

That was also the exact moment I realized I truly loved her, and I made up my mind that I had to give her happiness.

Later, we got married and had a lovely son.

Later, in order to expand my business world, I also tried hot spring resorts and private club projects. I lost heavily there too, but that is another story.

Food Business

One of my relatives opened a Huainan beef soup shop in Nanjing and did very well. Later my cousin opened a second one, and business was extremely good. After studying the category a bit, I opened one myself, and it performed far better than I had expected.

Then I started handling the design, renovation, and branding choices myself, and gradually expanded the product range. I opened a second store, then a third.

To make the products look better on food delivery platforms, I taught myself photography and Photoshop and created a full set of product images by hand.

Those small food businesses were genuinely exhausting, but the cash flow was strong. Although those stores gradually shut down during the pandemic years, they made me a decent amount of money before that. That money later became part of the capital that allowed me to become a partner in the software company. During that period, I also bought my Dream Car.

Running the Company

Once the company’s income and expenses were roughly balanced, we moved into a bigger office. Standing in front of an unfinished display wall, I imagined a very good future.

To create a better working environment, I hired more young staff, improved employee benefits step by step, and organized dinners and hangouts every so often.

The photo above is a precious picture of me with our front-end leader, Nas.

But the good period did not last long. Starting in early 2022, software sales fell off a cliff, and we could not find a solution. I had never been truly satisfied with the product. It was huge, bloated, and full of bugs. Frankly speaking, it was a mess. Customers said there was a bug every three steps and called it garbage. If the price had not been so low, some of them might have sued us.

To solve some core problems, I brought in my high school deskmate, who was then doing container security at ByteDance. He had always been someone special in my mind. After he joined us, he discovered some extremely serious issues.

First, the project structure was outdated, and many parts had serious performance and security problems. Second, some of the “AI” features were not generated through any actual model training at all. They were basically powered by Elasticsearch search behavior. We had spent years talking about big data, but in reality a lot of it was just preset outputs wrapped around business requirements.

That meant what we had been treating as core assets were, in truth, mostly junk. Meanwhile, we had spent most of our time on multi-platform adaptation, UI polish, and stuffing the product with miscellaneous features.

We kept iterating on UI and experience details, but the real problem was that there was no usable dataset. The so-called big data extraction had no meaningful training samples behind it, which meant nothing real could be trained.

In other words, the team that had been “doing big data” for years had never even built the dataset properly.

When I realized how serious that was, I felt a chill down my back.

At that stage, we were trapped. Continuing to iterate on the old foundation meant staying in a total garbage pile of code that was almost impossible to move. Rebuilding it would require far more money than the company still had.

One night I stayed late in the office discussing this with the other two partners. Mr. Gong, our largest shareholder, asked me, “Just tell me whether this thing can still be done. If you tell me it can’t, I’ll jump downstairs right now.”

I understood his emotions. He had already put more than ten million RMB into the company, all of it his own money. Not a single cent came from outside investors. I had also put in several million myself. It is hard to describe how everyone felt at that point.

Later, he slammed his forty-page employee management handbook onto the meeting table like a madman. The sound was so loud it felt like the office might collapse. In truth, some things had already become impossible to continue. We were just still holding onto illusions.

As the company’s losses got worse, we could no longer pay salaries properly. We tried everything to find money, including borrowing. Under extreme pressure, Mr. Gong started tightening attendance rules, cutting benefits, pushing meaningless overtime, and even monitoring employees’ work. Weekly reports became daily reports, then hourly reports. The team was being disgusted by one new control tactic after another.

I knew the team breaking apart was only a matter of time. I strongly opposed these performative management methods, but one sentence shut me down:

“I put in the most money. I make the decisions.”

By September, the atmosphere was so bad that nobody wanted to come to work anymore. Then people resigned together, and that was the end of the game.

The company collapsed. Leave aside the money for a moment. I had given five or six years of my life to it. How could I forget all those days and nights spent rushing projects? On top of that, I had dragged my deskmate into this mess. If he had stayed at ByteDance, maybe he would have had a much better future.

I used to imagine all of us celebrating success together one day, drinking and laughing. Instead, everything went the other way.

After the company shut down, I did not even have the courage to apologize to him. I felt awful inside, and I had no way to explain it to anyone. When I think about all those teammates who fought alongside me, I remember the times we cheered over difficult breakthroughs, the nights we were so excited we could not sleep, the promises I made, and the grand words I spoke in meetings.

I still feel deeply guilty.

To them, I just want to say:

I’m so sorry.

Even though they may never see this.

When My Body Broke Down

Starting in June, because of severe conflicts with Mr. Gong over how the company should be managed, and because my own management decisions were being overridden harshly, my body began to react.

My stomach started feeling bad. Later I developed erosive gastritis that never really healed. Then came insomnia. Then the stomach problems got worse. Then I got COVID. After that, it felt like every possible illness was landing on me at once. My vision became blurry. I lost my appetite. Long-term insomnia plus wandering pain throughout my body made me feel like I was falling apart.

One night, I suddenly told Ruohan that I thought I might not make it. Then I started talking nonsense and even began trying to explain what should happen after I died. She cried loudly enough to wake our sleeping child. I still remember the two of them crying for a long time, impossible to calm down.

In the following months, I went from one hospital in Nanjing to another. Waiting in lines, doing scans, MRIs, blood tests. I felt like a shell of myself. My whole world was wrapped in a layer of black fog.

The tests found a few minor problems, but doctors kept telling me with certainty that I was basically fine. I could not believe them. I was convinced that something serious had not been found yet. So every day became the same cycle: register, wait, get checked, wait for results, ask another doctor, rule out one disease, then look for the next one. I kept searching websites for symptoms like mine, telling myself I was okay and then doubting myself again.

Years of programming had also damaged my lower back. Please take that seriously. It is not reversible. The despair and pain of that kind of problem are hard to understand unless you have lived through it.

When people get desperate, they try everything. During that period, I visited famous traditional Chinese doctors, and my family even took me to pray and ask folk spiritual healers for help.

I knew this was not a real solution, but I still could not sleep night after night. I had no energy left at all.

The losses from our group alone were more than 20 million RMB. Once the resort and club projects were included, the pressure was enormous. At the hardest point, I even considered selling Ruohan’s family home. I never borrowed a single cent from relatives. I simply could not bring myself to ask.

I know that once certain words are spoken, relationships can change forever. If you borrow money, you should only ask when you are fully certain the other person will lend it. Otherwise, do not ask at all.

I even considered selling the car I had bought just over a year earlier, but my family stopped me. The house was already gone. If I sold the car too, it would directly affect daily life, especially with a child going to school, and it would not even recover that much money anyway.

After Going Back Home

At the end of March 2023, I transferred away my last small food business. By the end of June, I packed everything up and returned to my hometown.

Leaving a city I had lived in for more than ten years, with my three-year-old son beside me, was painful in a quiet way.

As my physical discomfort kept getting worse, I became quieter and quieter. First I stopped wanting to go outside. Then I stopped wanting to talk. Eventually I did not even want to eat.

If my mom had not messaged me every day telling me to come eat, I probably would have skipped meals entirely.

It was ridiculous. I was already nearly thirty, and I was still making my mother worry about whether I was eating.

I hated that version of myself too. But I felt powerless, like I was trapped in the center of a whirlpool. No matter how hard I struggled, I could not climb out.

Doing nothing for a long time is also unbearable. One day I reassembled my desktop computer and started playing League of Legends. I played day and night, until even I felt bored by it. But since I could not sleep anyway, I just let myself keep spiraling.

A Life That Felt Hopeless

I became addicted to games. My days and nights flipped upside down. I earned nothing. The contrast with my old modest but decent life was brutal.

Ruohan was never a very materialistic person, but when you are living off money from selling your house, sharing daily life with your parents, and showing no real sign of moving forward, disappointment naturally builds up. She was deeply unhappy with how I was living and asked me how long I was going to continue like this.

I shouted back at her, “What do you want me to do?”

The truth is, I really was lost. I had no idea what to do.

The software company was gone, and with it, my life as a programmer seemed to end too. I silently left many chat groups, closed GitHub, muted some contacts, and stopped visiting websites I used to browse every single day.

During the startup years, I even fell out with my best friend, Lele, over money. So many years of friendship collapsed because of my stubbornness. We barely spoke after that. I lost a lot.

More than once, I asked myself:

Why the hell did I fight so hard to keep this going?

Would things have ended differently if I had made other choices? If I had worked even harder at certain turning points, would it have helped? If I had started earlier or later, would it have gone differently? If only I had...

Looking back, building industry software really cannot be done through passion and a few seemingly clever ideas alone. Every founder believes he understands the user, believes his product solves a real pain point, and believes growth will eventually explode.

But success depends on many things outside your control: timing, conditions, luck.

For example, we spent huge effort building an “intelligent Q&A bot.” If we had done that after ChatGPT arrived, it would have been an entirely different story. But in real life, you do not get to wait for a world-changing product to arrive exactly when your own company needs it.

Reflection

If you ask me what it feels like to watch a company fail after building it with that much money and effort, I can only say this:

Falling hard while you are still young is not necessarily the worst thing. It may be only one episode in life.

That is the tough version of the answer, anyway. The honest version is that it hurt like hell.

I still clearly remember my son’s birthday. Early that morning, Ruohan said, “Let’s have hotpot tonight.” I hesitated for a long time and finally said, “There’s no need. Let’s just buy a small cake and celebrate at home.”

That answer was obviously not what she expected. She suddenly went quiet, washed up, and tried hard to hide her disappointment. At some point, even going out for a “real meal” had become something I saw as a luxury. And this was not just any day. It was supposed to be a day worth celebrating.

I had never imagined I would end up in such a state. That kind of sadness is probably understood only by people who have truly fallen low.

That night we still went to Haidilao, but we barely ordered anything. When the staff sang the birthday song and got to the line about “saying goodbye to all your troubles,” I completely broke down inside.

If only it were that easy.

Maybe many people are the same. You can endure your own suffering, but the moment your wife or child suffers even a little because of you, your heart turns upside down. A tiny, ordinary moment can shatter all the armor you built for yourself. All the sadness and frustration rushes back at once.

2026-04-22 Update

A lot of people have asked how I am doing now. I’m still here, and I’m doing well.

One more thing: this article contains a lot of criticism toward my partner Mr. Gong. In reality, he is a very capable and impressive person. At the critical stage when the company was fighting for survival, our principles and management philosophies clashed sharply. And because this article is written entirely from my personal point of view, it naturally makes him look like my opposite.

That is not the full truth. He has many admirable qualities.

Even after failure, the important thing is to keep learning, keep improving, and keep yourself in a state where you are always ready to start again.

May we both keep going.

2026-06-16 Update

On June 16, 2026, as Chairman of China Ciyuan Cloud Computing Co., Ltd., I was invited by Alibaba in Hangzhou to visit and study at Alibaba Cloud's Hangzhou headquarters.

This visit made one thing even clearer to me: AI should not remain only as a concept, and it should not be a tool held only by a small group of people. Truly valuable AI must enter real industries, real land, and the everyday lives of ordinary families.

Going forward, I hope to keep exploring practical ways for AI to empower agriculture, forestry, animal husbandry, fisheries, and other parts of the real economy. In particular, I plan to provide free AI training for rural cooperatives, so that more ordinary people can understand AI, use AI, improve production efficiency, and discover new ways to grow their businesses.

Seeking benefits for ordinary people and doing practical work should not be only a slogan. After going through the lowest points of my life, I increasingly believe that turning the ability to stand up again into something useful for others is one of the most grounded reasons to keep moving forward.

File: docs/en/threads/part-4/week-1.md

title: "Week 1: Real-life English Practice" updated: 2026-08-16

WHY

The reason I wanted to write this article is because I get one question very often from GitHub. Actually, I get a lot of questions, but there is one that stands out:

How can I speak fluently about many different topics?

LOL~

To be honest, I couldn’t speak English fluently even though I had been learning it for many years.

I know you want to speak like a native English speaker. You want to sound natural. You really want to speak fluently.

Well, in order to do that, you must use the words and expressions that we actually use in real life. And that’s why I created this series for you: Real life English.

Now I’m going to teach you words and expressions we really use. Are you ready?

Well then, let’s jump right in.

FIRST TOPIC

One of the most common problems I hear about nowadays is: I am so tired.

And you are not alone.

I experienced the biggest failure in my life in the past year. The startup company I ran went bankrupt. I lost my wealth, a precious friendship, and the person I loved. For a whole year, I couldn’t sleep well and felt very tired every day.

Trust me, I know that terrible feeling.

So, our first topic will be: why are you always tired?

Whether you are experiencing unemployment, or just broke up, or are confused about the future, let’s calm down and talk like old friends.

Before I created this course, I searched online for many articles on this topic. Some YouTube videos and even some ChatGPT answers inspired me a lot. I summarized the ideas of several popular videos and shared them with you here.

Apart from medical health issues, I’d put most reasons into three buckets, and we’ll talk about them one by one:

  • sleep anxiety
  • unhealthy lifestyle
  • stress

Sleep Anxiety

The first one is Sleep Anxiety.

Sleep anxiety is the feeling of being scared, worried, or concerned about sleep, which paradoxically makes it harder to fall asleep. That creates a vicious cycle.

I’ve been reading about the science of sleep for a while, but even with a medical background, it can get confusing. And when you listen to different people, it’s hard to tell what advice is good and what’s just noise.

For sleep anxiety, here are a few myths to watch out for:

Myth one: `It doesn’t matter when you sleep as long as you sleep enough`

In the 1990s, Russell and his research group discovered that there are some cells in our eyes that detect light rather than helping us see images. This light-detection mechanism tells our brain what time it is and helps regulate our internal clock.

Since our internal clocks are sensitive to light, generally:

  • we want to be awake when it’s bright outside
  • we want to be asleep when it’s dark outside

Russell and his team also studied what happens when the circadian rhythm is disrupted by sleeping at odd times (jet lag, night shifts). They found higher stress hormones, higher risk of heart disease, and people getting sick more often. They also became more prone to emotional and cognitive problems.

Reference video: https://www.youtube.com/watch?v=qlf9-573MhI 00:02:28,070 - 00:02:38,210

It’s also worth remembering: our internal clocks aren’t identical. That’s why some people are morning people and some are night owls. It’s related to your chronotype.

If you’re a morning person, it often makes sense to do high-focus work earlier. If you’re a night owl, that can work too—if you can control your schedule.

Myth two: Everyone needs eight hours of sleep

The truth is that sleep needs vary a lot. Instead of panicking over “only seven hours”, focus on habits that improve sleep quality.

Myth three: We should wake up at the same time every day

It’s generally good to wake up around the same time, but you don’t need to be insanely strict. Also, chronotypes change across life stages (teenagers often skew later).

Myth four: You should avoid blue light before sleep

Blue light has a short wavelength and higher energy. Some researchers believe it affects the eye/brain in ways that make sleep harder.

UNHEALTHY LIFESTYLE

The second bucket is unhealthy lifestyle.

My short advice: move more. Be physically active during the day. Exercise can boost mood, metabolism, and energy.

I’m not a fitness expert, so I’ll keep this part simple.

DEAL WITH STRESS

The most important thing I want to share is how to deal with stress.

Manage your stress and learn to relax properly.

Here is a video that might help you: https://www.youtube.com/watch?v=9QiE-M1LrZk

I mentioned in an article on Zhihu that I struggled with compulsive gaming for a period. I once described the change as a “dopamine detox”, but that popular label is not a clinical mechanism or established treatment. What helped me was changing cues and routines, reducing access, restoring sleep and daily structure, and seeking support. Persistent loss of control deserves qualified professional help.

Even when you logically know that studying, exercising, building a business, or doing something productive will benefit you in the long run, you still prefer watching TV, playing video games, and scrolling social media.

One might argue it’s obvious: one activity is easy and effortless, while the other is difficult and demands focus.

But some people seem to have no problem studying, exercising, or working on their side projects regularly.

At the low point of my life, one thing kept me moving forward: programming.

I needed a high-performance rich text editor for my previous company’s product. I couldn’t find an open-source project that met our needs, so I built one and open sourced it on GitHub.

Whenever I felt depressed or distracted, I quietly worked on it. Challenges I faced and overcame:

  • designing and coding the editor
  • developing the core and plugin system
  • designing a data structure for complex tables and submitting a PR to slate.js
  • refactoring the rendering engine to move away from React

If you are interested in this project, you can join and contribute.

I was very happy that I could finish it after more than a year of continuous development instead of giving up halfway.

Today’s lesson was very long, but fortunately it is OVER now.

Thank you for your attention, and I hope you learned something useful.

Prev: Misc

File: docs/en/threads/part-3/1-ai-learning.md

title: "Learning Anything with AI: From Chat to Verifiable Work" updated: 2026-08-16

Learning Anything with AI: From Chat to Verifiable Work

AI makes it easy to confuse a complete answer with personal learning. A real result survives after the answer is closed: you can explain, judge, perform, transfer, and leave work another person can inspect.

The Seven-step Loop

  1. Define a real outcome: exam, report, program, presentation, decision, or lesson.
  2. Save a baseline: attempt it independently to expose prerequisites and errors.
  3. Select sources: prefer primary documentation, textbooks, papers, data, and cases.
  4. Use guided practice: ask for questions, hints, and layers rather than substitution.
  5. Produce actively: close sources and explain, code, write, calculate, or demonstrate.
  6. Obtain plural feedback: AI is one vote; tests, sources, experts, or users validate important judgments.
  7. Update state: save evidence, errors, and the next action for another session.

Tool-selection Order

  1. Is the data sensitive and permitted to upload?
  2. Does the task need sources, long material, voice, code execution, or collaboration?
  3. Can output be exported, reproduced, and verified independently?
  4. Is it available for the region, language, age, account, and plan?
  5. What does it cost in money and attention?

See Learning English with AI for the boundary between guided learning and projects and for dated official product notes. The same principles apply to programming, writing, exams, and research.

Cross-session State

Use Learning State instead of asking a model to remember you. At the end of every session save:

  • completed work and evidence location;
  • claims verified by source or test;
  • errors that still recur;
  • the smallest next task and acceptance criteria.

The file should make sense to any model, another person, or your future self. Platform memory is a convenience layer, not the only record.

Apply by Domain

Programming

  • write acceptance tests, interfaces, or failing examples first;
  • ask for assumptions and risks before minimal changes;
  • run tests, static checks, and security review yourself;
  • separate “runs” from “correct requirement, safe boundary, maintainable design”.

Writing

  • keep the draft and review audience, purpose, facts, and structure first;
  • require quoted passages rather than wholesale rewriting;
  • return facts to primary sources and inspect quotation context;
  • the named author remains responsible for the final position.

Exams and Certifications

  • build a capability matrix from the official syllabus and samples;
  • answer before hints and create parallel tasks from error causes;
  • complete regular timed, unaided simulations;
  • do not treat a model-predicted score as an assessment.

Industry Research

  • separate facts, inference, forecasts, and recommendations;
  • record publication date, measurement definition, and primary link for numbers;
  • seek counterexamples, interests, and missing data;
  • use qualified professionals for high-stakes decisions.

Work Delivery

  • state audience, decision, deadline, and acceptance criteria;
  • split research, draft, validation, review, and publication;
  • use only tools and data allowed by the organisation;
  • record human reviewers, tests, and changes.

Feedback Protocol

text
Restate the task and acceptance criteria, then let me attempt it. Separate feedback into facts/sources, task completion, structure/reasoning, skill errors, and expression detail. Quote evidence and choose only one to three issues that most affect the result. Mark uncertainty and do not invent sources. Give one minimal hint and one parallel task rather than completing the final deliverable.

Prevent Skill Outsourcing

  • complete one unaided baseline each week;
  • explain important concepts without the answer;
  • explain every critical decision in important work;
  • test generated code, cite research from primary sources, and revise prose as the author;
  • reduce assistance when performance collapses without it.

Follow organisational policy and applicable law first. Do not upload secrets, unpublished data, customer/student records, health or identity data, unauthorised work, or third-party private information. Use the minimum necessary excerpt. When deleting, inspect projects, shared links, history, and local copies. AI can produce infringing, biased, or dangerous material; the publisher remains responsible.

Seven Days, Thirty Days, Twelve Weeks

Seven days: one micro-deliverable

  • define the real outcome and acceptance criteria;
  • save an unaided baseline;
  • use the loop for a one-to-three-hour deliverable;
  • record where AI helped, misled, and what remains without it.

Thirty days: repeatable workflow

  • complete four loops for a recurring task;
  • build prompt, source, test, feedback, and state templates;
  • compare time, quality, rework, and independent performance;
  • remove tool steps that do not improve the result.

Twelve weeks: real delivery

  • choose a project with value in work, study, or life;
  • obtain real-audience feedback every two to four weeks;
  • complete fact, test, security, privacy, and copyright review;
  • export material and state, then write a one-page retrospective.

Acceptance Questions

  1. Can I explain key conclusions without the chat?
  2. Can facts, numbers, and citations return to primary sources?
  3. Has the output passed a real test or audience review?
  4. What is inference and what remains uncertain?
  5. Did I expose unnecessary data or third-party content?
  6. Can I complete the next similar task with less assistance?

Previous: Learning English with AI | Next: Learning State

File: docs/en/threads/part-2/x-misc.md

title: Miscellaneous Notes updated: 2026-08-16

Miscellaneous Notes

Source (中文): 扯淡篇

This chapter is intentionally a bit off-topic: personal opinions, stories, and some “learned the hard way” moments.

A Few Words About IT Bootcamps in China

In recent years, the internet industry in China grew fast. It became both “the hottest field” and also one of the most misunderstood.

Because the pay can be good, a lot of people from other industries try to jump in.

If the only reason you want in is “the salary is higher than what I have now”, but you don’t like the work at all… then the more urgently you push yourself into the industry, the more painfully you’ll eventually push yourself out.

Try to optimize for work that you enjoy and can do well. I hope you get to live the life you want—not trade life away just to survive.

I’ll quote a passage from Front-end Developer Handbook 2017 (with the named institutions removed):

Lately a lot of non-accredited, expensive, front-end code schools/bootcamps have emerged.
These avenues of becoming a front-end developer are typically teacher directed courses…
Keep in mind… this is the web! Everything you need to learn is on the web for the taking…
However, if you need someone to tell you how to take and learn what is actually free, and hold you accountable… you might consider an organized course.
Otherwise… I am not aware of any other profession that is practically free for the taking with an internet connection… and a burning desire for knowledge.
— Front-end Developer Handbook 2017

So if you want to enter the industry through training, my advice is:

You’re trying to join the internet industry. You should have “use the internet” as your first instinct. Spend that expensive tuition on something fun for yourself.

If you’re still a student and a bootcamp tries to push you into “student loans / installment plans”, be extra careful. Some of them are genuinely shady.

Are Bootcamps Worth It?

I trained for 4 months and spent 20,000 RMB on web front-end development.

Two weeks after finishing, I got an offer from Meituan.

I admit I wasn’t the best student in class.

But I was the first one to find a job.

And it was a big company.

I’ve always believed hard work can make up for talent. If you’re determined, a lot is possible.

It’s been one week since I joined. Everyone is nice. They even gave me an electric scooter, a helmet, and a coat.

Anyway, gotta go—new order just came in.

“5 Minutes a Day to Quit English Forever”

A lot of training/education businesses sell the idea of “N minutes per day and your English will magically improve”. It’s basically candy: lure you in first, talk later.

English is a language. You can’t truly acquire it in a ridiculously short time.

“Fragmented learning” often ends in:

  • giving up
  • or getting distracted into something else (like discovering you really love American TV shows)

If you’re working full-time, you need a real plan: combine methods and build a system.

So yeah—I don’t recommend paying for that kind of empty promise.

Kindle and iPad

I know a lot of people’s Kindle and iPad are collecting dust… or being used as instant noodle lids.

That might be funny, but is that really why we bought them?

A Pretty Painful Learning Story (Personal)

In middle school, I ranked No.1 in my grade. I got arrogant.

In 8th grade, a new math teacher arrived. I couldn’t stand that he sometimes explained easy problems incorrectly. I looked down on him, ignored homework, and basically did my own thing in class.

One day he messed up another problem. I laughed and shook my head. He must have been holding it in for a while—he snapped and beat me. Later the principal found out and fired him.

In 9th grade, I repeated the same mistake with another math teacher. This time it wasn’t because he was wrong, but because his method wasn’t “optimal”. I got beaten again. My deskmate even cried. I somehow didn’t feel much pain.

Later that teacher apologized. He slapped himself, grabbed my hand, and tried to make me hit him. I said, “The last teacher who hit me got fired.”

He told me his father was seriously ill, his brother was unemployed, and he needed this job. We made a deal: I didn’t have to attend his math class; I could go play.

And I really didn’t attend his math class after that.

Afterwards, I completely lost interest in math. I even ripped up my certificates from math competitions.

Youth. Ego. Sigh.

In high school, pressure got intense—not only from school, but from my own unrealistic expectations.

Because I didn’t manage “charging” at night (resting) well, I was exhausted in the daytime. Then at night I couldn’t fall asleep, worried that lack of sleep would ruin the next day… which made me even more anxious… which made me sleep even worse.

My homeroom teacher criticized my free-falling grades and kept calling me into the office.

The grade director only cared whether I was late again so he could “score a point”.

My dad stopped attending parent meetings once he learned how far my ranking had dropped.

Over time, it turned into a vicious cycle. I developed insomnia. I was disappointed in myself. My seat got moved to the front row. My bed got moved into a hospital.

I spent two years treating insomnia. It was a dark period.

During that time I had moments where I didn’t want to live anymore. “Insomnia is painful” is honestly an understatement.

Many cold nights, I cried alone in the bathroom, wondering what life meant.

I just wanted one good night of sleep. But I couldn’t.

No one told me that life isn’t about beating others. It’s about enjoying the process of moving toward your goals—and letting results be a reward for your actions.

Before the insomnia, I wrote with my left hand. During the insomnia, to kill time, I learned to write with my right hand. I practiced calligraphy, which eased the pain a bit.

I spent more time on calligraphy, won competitions, and got perfect scores in Chinese essays more than once (I remember the one criticizing exam-focused education best).

Then something shifted. At night I started feeling sleepy again. I could sleep.

My grades improved a bit too. I still did poorly on the college entrance exam overall, but my high scores in Chinese and English helped me get into a decent university (by then I basically never touched math again).

So when you’re at a low point in life, try learning something new. You might find unexpected joy in it.

Learn to let go, accept yourself, and find what you truly like. If something is necessary for your work and life, try to learn to like it.

That’s part of why I’m sharing these experiences—to help people waste less time and take fewer detours.

During college, my English teacher had it out for me. It started because I forgot my textbook in one class. She asked me to spell words… and I got all of them right.

She praised me in English. I showed off a bit and had a fairly fluent conversation with her.

The whole class looked at me with envy (or contempt—hard to tell). I still feel a little proud when I remember it.

After that, she kept targeting me. I never escaped roll call again.

I hated her.

Then, thanks to her “efforts”, basically the whole department started watching me. Every attendance check began with: “Is the guy from Class 2 here?” If I was there, they didn’t bother calling names anymore.

By the second semester of junior year, I failed 19 courses.

At the start of senior year, the university wanted to demote me back a year. Graduation seemed impossible.

Then the English teacher helped me talk to the homeroom teacher, the department head, the academic office… and insisted I stay in my year. The deal was simple: retake 19 courses; if I failed, I’d drop out.

Senior year, I carried a heavy backpack and studied every day in classrooms and the library. Every evening on the way back to the dorm, I would “hate my English teacher again”.

How can someone be that intense?

Final exams: 26 courses. I was exhausted.

When I found out I passed all 19 retakes, I almost cried. I finally got to slap that “you won’t graduate” story back in the face.

And then I wrote this as my graduation message:

“In college, never fall in love with your teacher.”

Summary

After all this, my conclusion is: learning English mostly depends on talent.

And the best talent is… effort.

About the times I got beaten by teachers: I take responsibility too. I didn’t respect their dignity and acted like a jerk.
This chapter is mostly nonsense—just for a laugh.

Next: My ex-girlfriends

Prev: Learning English with AI
Next: Week 1

File: docs/en/threads/part-1/1-understanding.md

title: "Learning Principles: Turn Effort into Evidence" updated: 2026-08-16

Learning Principles: Turn Effort into Evidence

Before choosing an app or textbook, decide why you are learning, what task you need to complete, how you will judge it, and how to train without sacrificing health.

Separate Three Kinds of Claim

  • Research evidence reports effects observed under stated conditions, not a guarantee for every learner.
  • Personal experience explains why I made a choice; it is a clue for your experiment, not a universal rule.
  • Hypothesis lacks enough evidence and should become a question to test.

When advice omits the population, conditions, outcome, and source, treat it as a hypothesis.

Why Learn English?

English connects you to technical documentation, research, courses, cultural work, and international collaboration. That importance is still not an actionable goal. Name the context:

  • an exam with a date, target score, and task types;
  • study that requires discussion, academic reading, or writing;
  • work that requires documentation, email, meetings, interviews, or presentations;
  • life that requires travel, reading, or communication with particular people.

My primary context is ______. In 12 weeks I will complete ______ under ______ conditions. The evidence will be ______.

The context determines material and feedback. Do not distribute time evenly in pursuit of a vague wish to “improve English”.

Remove the Learning-pyramid Percentages

Earlier editions repeated claims such as “people retain 10% of reading and 90% of practice”. Those precise numbers have no reliable research basis and are not what Edgar Dale's Cone of Experience proposed. This edition removes the old diagram and percentages.

Keep a testable principle instead: repeated exposure can create familiarity, while retrieving, explaining, applying, and receiving feedback reveals whether knowledge is usable.

Two broadly supported strategies are:

  • retrieval practice: close the source, recall, answer, retell, write, or teach;
  • distributed practice: return at multiple points instead of cramming once.

Neither offers universal percentages or a universal calendar. Difficulty, prior knowledge, delay, and assessment method all matter.

A Complete Learning Loop

Each 25–45 minute session should contain:

  1. Task: “retell a two-minute news clip”, not “practise listening”.
  2. First attempt: produce a baseline without excessive lookup.
  3. Targeted input: review only the information needed for current errors.
  4. Active output: close the material and perform again.
  5. Feedback and record: compare versions and update Learning State.

Finishing a video or feeling that it made sense is not the endpoint. Performance after the source is closed is stronger evidence.

Output Is Not Performance Theatre

Active output can be small:

  • write a three-sentence summary from memory;
  • list a main point and three details after listening;
  • explain a concept without notes;
  • use a new chunk in two settings;
  • record sixty seconds, replay it, and mark pauses.

Output exposes gaps. Its job is not to prove intelligence.

Let Performance Set the Interval

A schedule such as days 1, 2, 4, 7, and 15 can be a reminder, but it is not Ebbinghaus's universal prescription for you.

  • Easy and accurate recall: lengthen the interval or raise difficulty.
  • Effortful success: keep a similar interval.
  • Frequent failure: shorten the interval, reduce material, or repair prerequisites.
  • Recognition without use: add production and varied contexts.

Record performance and adjust. Missing an imagined “perfect review time” is not a reason to abandon the system.

Sleep and Recovery Are Inputs

Personal experience

At school I once confused visible effort with learning: waking at 5:20, skipping midday rest, and reading after lights-out. Time rose while performance fell. I interpreted this as insufficient effort and pushed harder.

That story does not prove all early rising is ineffective. It shows that for me, chronic sleep loss, no error diagnosis, and mechanical hours combined badly.

Evidence boundary

Sleep contributes to memory consolidation, while sustained sleep loss can affect attention and performance. Needs and health conditions differ. This guide gives no medical diagnosis; persistent sleep, mood, or physical problems deserve qualified professional care.

Budget sleep, movement, food, and unassigned time. Recovery is a condition for the next session, not a prize after it.

Do Not Match Fixed “Learning Styles”

People may prefer pictures, sound, or text, but evidence does not support classifying people as visual or auditory learners and improving outcomes by matching instruction to that label.

Choose representation by content and task:

  • pronunciation needs sound, articulation, and acoustic feedback;
  • spelling needs visual recognition and production;
  • spatial relationships often benefit from diagrams;
  • communication integrates listening, speaking, reading, and writing.

Multiple representations can help. A preference is not a fixed identity or ability ceiling.

Compare with Your Baseline

A useful goal includes a real context, observable task, quality threshold, deadline, and comparable evidence. Use the CEFR self-check for four separate baselines. An average label cannot describe B2 reading and A2 speaking.

Order Feedback by Impact

Treat only one to three high-impact problems at a time:

  1. task completion and meaning;
  2. organisation and logic;
  3. recurring vocabulary, grammar, or pronunciation errors;
  4. low-impact details last.

AI can classify errors and create parallel practice, but it can misjudge facts, pronunciation, and usage. Check important work against dictionaries, corpora, teachers, peers, or real audiences.

Seven Days, Thirty Days, Twelve Weeks

Seven days: baseline

  • save short samples in all four skills;
  • choose one real context;
  • complete three full learning loops;
  • change only one variable from the evidence on day seven.

Thirty days: rhythm

  • keep at least one comparable output each week;
  • classify errors instead of calling everything “careless”;
  • replace exercises gradually with authentic material;
  • repeat the baseline conditions on day thirty.

Twelve weeks: transfer

  • save parallel samples in weeks 1, 4, 8, and 12;
  • add time pressure, unfamiliar topics, or a real audience;
  • complete an exam simulation, meeting, presentation, long reading, or writing deliverable;
  • let the result define the next cycle.

Action

  1. Complete the English Diagnostic.
  2. Create a Learning State.
  3. Delete one task that produces effort but no evidence.
  4. Schedule the first 25–45 minute learning loop.

Sources

Previous: CEFR Goals | Next: Vocabulary

File: docs/en/threads/part-1/2-vocabulary.md

title: "Vocabulary: From Recognition to Contextual Use" updated: 2026-08-16

Vocabulary: From Recognition to Contextual Use

Vocabulary is not a set of English = translation pairs. Knowing a word or chunk involves recognising its sound and form, understanding common senses, knowing collocation and register, and producing it in an appropriate setting.

Remove Two Misleading Equations

Earlier editions claimed that 1,000 words give 75% understanding and 7,000 nearly 90%. Such equations mix counting units, corpora, and definitions of comprehension, so this edition removes them.

Before discussing vocabulary size, ask:

  • Are we counting word forms, lemmas, or word families?
  • Is the material conversation, fiction, news, academic prose, or specialist documentation?
  • Are proper nouns and transparent compounds included?
  • Does “know” mean recognition or productive use across skills?

Research by Nation and others often discusses 95% and 98% lexical coverage as reading conditions. At 95%, roughly one running word in twenty remains unfamiliar; at 98%, about one in fifty. The higher figure generally supports more independent comprehension, but no fixed vocabulary count guarantees understanding. Background knowledge, text structure, language distance, and the task still matter.

Four Kinds of Vocabulary Evidence

Dimension Question Evidence
Receptive listening Can I segment and understand it in speech? Hear a sentence without captions and explain it
Receptive reading Can I understand it from a new context? Explain its function in a new article
Productive speaking Can I retrieve it quickly with natural collocation? Use it naturally in a recording
Productive writing Can I spell it and choose the right register? Use it accurately in a new task

Feeling familiar after seeing the answer is only a small part. Receptive knowledge often precedes productive knowledge, so choose the target from the real task.

Word Families Are a Research Tool

help, helps, helped, helpful, helpless may count as one family in some estimates, but learners do not automatically know every derived form. Families help estimate coverage; they do not replace learning morphology, part of speech, and use.

The useful unit is often larger than one word:

  • collocation: pose a risk, meet a deadline;
  • chunk: What I mean is ..., It depends on whether ...;
  • morphology: analyse → analysis → analytical;
  • register: kids, children, and minors differ by context.

Do Not Classify Visual and Auditory Learners

Evidence does not support improving learning by matching instruction to fixed visual or auditory labels. Vocabulary itself needs multiple kinds of evidence:

  1. hear pronunciation and connected speech;
  2. see spelling, morphology, and a real sentence;
  3. say it, record it, and compare stress and sounds;
  4. write an original sentence or short text;
  5. use it in varied but related contexts.

Images suit concrete concepts; definitions and contexts suit abstract ones; pronunciation requires sound. Representation follows the content, not a learner identity.

Select What to Learn

Prioritise items that:

  1. recur in current material or work;
  2. block the main idea or task when unknown;
  3. form useful collocations with known vocabulary;
  4. have a clear use in the next two weeks;
  5. are essential terms in a specialist task.

Temporarily skip one-off low-frequency detail that does not block meaning.

A High-quality Card

The front should demand retrieval, not recognition.

text
Context: how do I say that a dependency creates risk for a schedule?
Gap: The dependency may ____ a risk to the schedule.

The back includes:

  • pose a risk to;
  • one trustworthy source sentence and citation;
  • pronunciation or audio;
  • a short sense and register note;
  • your second example;
  • a contrast such as cause versus pose.

Long cards turn review into rereading. Test one decision per card when possible.

Performance-adjusted Spacing

Forgetting is real, but there is no universal review timetable. In Anki or another spaced-repetition tool, let performance control the interval:

  • fast, accurate recall plus new use: lengthen it;
  • effortful success: keep or slightly lengthen it;
  • repeated confusion: add contrast and a new context;
  • repeated failure: reduce new cards, split the card, or repair pronunciation/concept foundations;
  • correct cards but no real-world use: stop adding volume and switch to speaking or writing.

Anki is one tool, not the method. An old third-party “Macmillan 7000” deck shared here is no longer distributed because its link, maintenance, and redistribution rights cannot be kept current.

Practice in Real Context

For each group, complete at least two transfers:

  • alter the people, time, viewpoint, or domain in the source sentence;
  • contrast a natural use with a typical misuse;
  • give a 60–90 second retelling without the list;
  • find three collocations in independent real sources;
  • use the chunks in an email, explanation, log, or short essay.

Prefer learner dictionaries for pronunciation, collocation, and examples. AI can generate candidates, but verify definitions, collocations, and examples against dictionaries or real corpora.

A Thirty-minute Session

  1. Five minutes: select 5–8 valuable chunks from real material.
  2. Eight minutes: check pronunciation, sense, collocation, and register.
  3. Seven minutes: create short retrieval cards.
  4. Seven minutes: speak or write a different context without the source.
  5. Three minutes: record errors and the condition for the next check.

Daily new-word count is not the outcome. Understanding and using items in a new task a week later is.

Seven Days, Thirty Days, Twelve Weeks

Seven days

  • collect 20–30 chunks from one real topic;
  • include sound, collocation, and source for each;
  • complete at least two closed-book retrievals;
  • on day seven, record two minutes or write 200 words with the set.

Thirty days

  • use parallel material weekly to test listening or reading coverage;
  • complete one real output with the week's chunks;
  • delete low-value, duplicate, or cue-dependent cards;
  • compare comprehension, production, and retention rather than card totals.

Twelve weeks

  • build a small corpus around one professional or life domain;
  • use high-frequency chunks in meetings, articles, code review, exams, or presentations;
  • sample delayed recall and transfer to unfamiliar contexts;
  • expand domains only after output becomes stable.

Feedback Criteria

For each target chunk, ask whether you can:

  • recognise it in speech and text;
  • explain the current sense in your own words;
  • use natural collocation, form, and register;
  • produce it without a cue;
  • reuse it in a new context one week later.

Sources

Previous: Learning Principles | Next: Listening

File: docs/en/threads/part-1/3-listening.md

title: "Listening: From Sound Recognition to Real Understanding" updated: 2026-08-16

Listening: From Sound Recognition to Real Understanding

Source (中文): 听力篇

Most of the recommended videos are on YouTube. Many of them have fairly accurate English captions (often auto-generated). Depending on where you live, you may need a VPN to access them.

Common Mistakes When Practicing Listening

  • Scattering your materials

    If you subscribe to everything—VOA, BBC, ten podcasts, random “classic readings”… you’ll end up listening to nothing consistently.

    One day VOA, next day BBC, then you bounce between random YouTube channels. That kind of “no plan, just vibes” practice isn’t very efficient.

    VOA/BBC/podcasts can be great. Just pick a few you genuinely like and that fit your time.

  • Starting with materials that are too hard

    A lot of people start with VOA. I don’t recommend it. Even “VOA slow” can have enough unknown words to make you stuck and lose interest.

    Some people ask: “If I don’t understand it once, what if I listen 10 times?”

    If you never clear the vocabulary using the transcript, listening 100 times still won’t help much. That goes against what we’re aiming for: high-leverage learning.

    US/UK TV shows are even less beginner-friendly. Choose materials that match your current vocabulary and level, and move up step by step.

  • Over-relying on subtitles

    If you watch shows with bilingual subtitles, it’s easy to think you “understand most of it”. Try turning subtitles off. Even on a rewatch, you’ll miss a lot.

    Does “no subtitles” automatically improve listening? Not always—because it can violate the previous rule (material too hard).

    A better approach: when you rewatch a show/movie you love (second/third time), turn subtitles off. Your memory fills the gaps, and you train real listening.

  • Using materials you don’t care about

    No matter how “high quality” a listening resource is, if it bores you, it becomes torture. And I won’t tell you to “learn to love” something you already know you hate.

    There’s so much content out there now—it’s not hard to find something both useful and interesting.

    Bonus: choose materials that help your work, life, relationships, fitness, etc. Real benefits create real motivation.

  • Listening on autopilot

    Most people listen to English songs a lot, but rarely pay attention to lyrics. That’s a huge missed opportunity.

    English lyrics are often beautifully written. If you learn what the lyrics mean and try to identify what’s being sung, you’ll improve over time.

    That said: sometimes just enjoy the music. Turning life into 24/7 studying can lower your happiness too.

Intensive vs. Extensive Listening

Intensive listening

The goal is: understand the meaning first, then try to hear every word clearly. Intensive practice often pairs well with:

  • dictation (spelling what you hear)
  • repeating out loud (shadowing / speaking practice)

It’s normal if you can’t keep up on the first listen. A few repeats usually make a big difference.

Intensive practice also helps you notice pronunciation patterns: linking, reductions, pauses, chunking. Find your weak spots and train them deliberately.

In high school I used a very “dumb” method: I listened over and over until I could recite and write down the whole text. I spent a lot of time on New Concept English Book 3. Back then, I couldn’t really see the improvement and even wondered if it was pointless—until I scored 115/120 on the English exam.

There are many options, but New Concept English Book 3 and 4 are solid classics for this kind of training.

Extensive listening

Extensive listening is about making English exposure natural, building intuition, and understanding the overall meaning rather than every detail.

It helps with vocabulary and speaking too. Great sources include audiobooks, movies, TV shows, and music.

Audiobooks

Many classics and bestsellers have audiobook versions: Gone with the Wind, The Kite Runner, Pride and Prejudice, The Great Gatsby, etc.

Audiobooks are often narrated by professional voice actors—great voices, great pacing, and surprisingly enjoyable.

  • audible has a lot of resources and a decent app.
  • Some Chinese “FM” apps also have good audiobooks.

TV shows

Friends is a classic for a reason.

I also strongly recommend Modern Family—it’s one of my favorite sitcoms, won a bunch of Emmy awards, and the writing is excellent.

Better Call Saul is also amazing, but the language level is noticeably harder than Friends and Modern Family.

Season 1 of Modern Family was filmed in 2009. In real life, many people’s “modern family life” still hasn’t caught up to that 2009 version.

Movies

Great movies are perfect for extensive listening, especially high-rated classics. See the Douban Top 250 list: https://movie.douban.com/top250

Movies I’ve watched 10+ times (purely personal taste):

  • Flipped
  • The Godfather I/II/III
  • WALL·E
  • Witness for the Prosecution
  • The Apartment
  • Interstellar
  • Life Is Beautiful

Learning English through movies and TV takes a decent foundation. The “learning effect” often comes as a bonus while you enjoy the film.
Podcasts (or audio with lyrics) have fewer distractions than full video, so the learning signal can be clearer.

Music

When choosing songs, use current charts and check their region and date. The list below preserves only the selection approach.

There are way too many great songs to list, but here are a few directions:

  • popular artists: Justin Bieber, Troye Sivan, Katy Perry, Taylor Swift, The Weeknd, Imagine Dragons, Ed Sheeran, Adele, Maroon 5, Billie Eilish, Sam Smith…
  • Apple Music charts
  • Billboard
  • UK charts

Live streams

If you like streams, find creators you enjoy on Twitch.

Beginner Listening Training

Advanced: Understanding Native-Speed English

Do you sometimes struggle to follow native speakers, or to understand YouTube videos and movies?

Maybe you do fine in listening tests inside apps, feel confident… and then you remove subtitles and everything collapses.

One important thing: most listening training materials are processed—they’re made easier on purpose so learners don’t feel crushed.

A common mistake in listening is focusing too hard on single words instead of the sentence meaning.

But:

  • words change in context
  • when words connect, pronunciation changes too

If you translate word by word, you will misunderstand. Try to listen for chunks and meaning.

If you want a deep dive, I strongly recommend this:

Good English Learning Resources

  • laracasts
    Rating: 5/5
    Great front-end + Laravel video tutorials. Detailed, frequently updated, and beginner-friendly. Covers JavaScript/Vue/React/Laravel/PHP/editor tips, etc.
    Iconic line: Does it make sense to you?

  • LearnCode.academy
    Rating: 5/5
    If you want to learn React/Redux/MobX/AngularJS/NodeJS/Docker, you can easily spend a lot of time here. (The channel has been inactive for years; older videos still work as listening material, but check current docs for the tech itself.)

  • Traversy Media
    Rating: 5/5
    A great front-end channel with wide coverage and frequent updates. Clear pronunciation and slower pace—friendly for beginners.

  • Derek Banas
    Rating: 4/5
    His “learn a language in one video” series helps you quickly see the basics of popular programming languages. High information density and faster speed.

  • The Net Ninja
    Rating: 4/5
    Great channel for front-end learning. Old intro audio was a bit scary; newer videos have a friendlier intro. CSS/Sass content is worth checking out.

  • DevTips
    Rating: 4/5
    Beginner-friendly. Solid fundamentals, especially CSS and jQuery. (Inactive for years; treat it as listening material.)

  • egghead.io
    Rating: 4/5
    Many front-end courses. Some are free.

YouTube channels for English learners

For superhero movie fans

Talk shows

Music channels

Others

  • Disney UK
    The Frozen theme song has 1B+ views.
  • Vevo
    A gold mine of “movie-level” music videos. Many artists have their own official channels too.
  • OneDirectionVEVO (the band has been on hiatus since 2016; the channel is inactive)
  • SiaVEVO
    Great production quality. Often thoughtful.
  • SSSniperWolf
  • TED
    TED Talks are great for ideas. Start with popular videos—many have multilingual subtitles.

A Few Specific YouTube Videos

CEFR Task Goals

  • A1–A2: identify people, time, place, and direct needs in slow, familiar speech.
  • B1: follow main points and important detail on familiar topics, then retell them.
  • B2: follow natural interviews, meetings, or lectures and distinguish position, reason, and attitude.
  • C1–C2: process extended, implicit, or accent-diverse material and synthesise sources.

A level selects material; a real goal states conditions, task, and evidence.

Inspectable Practice

Choose 2–5 minutes with a reliable transcript:

  1. Listen once without pausing; write the gist and three details.
  2. Mark uncertain time ranges on a second pass without reading the whole transcript.
  3. Classify errors: unknown language, known-but-not-heard, connected speech, background knowledge, or attention.
  4. Shadow two or three useful segments, close the transcript, and retell.
  5. Use parallel material one week later to test transfer.

Keep the first answer, error timestamps, retelling, and delayed test. Playback time is not evidence.

Feedback

Score each 0–2: gist, key detail, segmentation, retelling comprehensibility, and delayed transfer. Repair the largest barrier instead of perfecting every sound at once.

Seven Days, Thirty Days, Twelve Weeks

  • Seven days: four 2–3 minute sessions; compare parallel material on days 1 and 7.
  • Thirty days: one intensive listen, two extensive listens, and one no-caption retelling weekly; count error categories rather than hours alone.
  • Twelve weeks: increase speed, accent variety, length, and topic unfamiliarity; finish with a blind-rated meeting, lecture, or exam simulation.

Prev: Vocabulary
Next: Reading

File: docs/en/threads/part-1/4-reading.md

title: "Reading: From Word-by-word Translation to Claims and Evidence" updated: 2026-08-16

Reading: From Word-by-word Translation to Claims and Evidence

Source (中文): 阅读篇

Intensive vs. Extensive Reading

Intensive reading

Short but dense writing is great for intensive reading—The Economist is a classic example.

If you skim an Economist-style article, you’ll only get the surface. The structure, subtext, and sharp opinions are where the value is.

Treat intensive reading like tasting: look things up, reread, and try to figure out why the author chose those words.

Extensive reading

Beautiful novels are perfect for extensive reading—Animal Farm, for example.

This is more about enjoyment. Sometimes the first few lines hook you so hard you can’t stop.

Don’t turn it into homework:

  • no strict plan
  • no rushing to “finish the book”
  • go with your mood and enjoy it

Also: start with something manageable. Jumping straight into A Song of Ice and Fire is a fast way to get crushed by unknown words and quit.

To help you get started, here are a few books across different difficulty levels (inspired by various English-learning YouTube channels):

If you’ve already read the Chinese translation of a book, it often helps a lot before you read the original English version.

WeChat Official Accounts

I rarely use WeChat, so I’ll only recommend two that I’ve actually read:

  • 英文悦读
  • 独霸上海的妖怪

If you have good recommendations, open an issue and I’ll consider adding them after reading.

How to Read English Documentation

Do not use a mechanical “five unknown words means stop” rule. First decide whether the words block the main idea. Mark them on a fast first pass; on the second, look up recurring, domain-critical, or reasoning-critical items; then close the dictionary and reconstruct the structure. If almost every sentence is blocked, choose easier parallel material or learn the domain chunks first.

For technical terms, be extra careful:

  • read definitions and examples
  • if you still don’t get it, search for good explanations (including Chinese blogs when useful)

If you use tools to pre-extract unknown words from a document and learn them first, reading becomes much smoother.

Kindle isn’t great for technical PDFs: fonts are often too small and it’s tiring for your eyes. An iPad usually works better.

Medium, Quora, Reddit, Hacker News, Stack Overflow

  • Medium: great articles on tech, design, and life. Good for broadening your mind and improving execution.
  • Quora: global Q&A. Top stories often have surprisingly moving answers.
  • Reddit: the “front page of the internet”. News and gossip included.
  • Hacker News: a must for people in tech.
  • Stack Overflow: the place to solve technical problems.

Reference Book

CEFR Task Goals

  • A1–A2: understand short notices, instructions, messages, and direct familiar information.
  • B1: identify main point, sequence, and main reasons in familiar articles.
  • B2: understand arguments, evidence, and position in complex or specialist material.
  • C1–C2: process long, stylistically varied, implicit material and synthesise sources.

Inspectable Practice and Feedback

Choose a challenging but finishable article. Time the first read and state the gist in one sentence; label each paragraph's function; reconstruct the argument in five sentences; write one supported inference and one counterexample; verify quotations last.

Score 0–2 for gist, structure, evidence, inference boundary, and delayed retelling. Unknown-word count is diagnostic, not a level judgment.

Seven Days, Thirty Days, Twelve Weeks

  • Seven days: read four short texts on one topic; keep first summaries and a day-seven retelling.
  • Thirty days: one annotated close read and two timed extensive reads weekly; compare speed and summary accuracy.
  • Twelve weeks: move from one text to multi-source synthesis; finish with a cited report, technical review, or exam simulation.

Prev: Listening
Next: Speaking

File: docs/en/threads/part-1/5-speaking.md

title: "Speaking: Make Meaning Arrive" updated: 2026-08-16

Speaking: Make Meaning Arrive

Source (中文): 口语篇

Speaking is where everything comes together. And yes—pronunciation matters, but you don’t need to obsess. You just need a solid base and a lot of reps.

IPA / Phonics

If you want to speak better, start with pronunciation. And to learn pronunciation, you’ll eventually meet IPA (phonetic symbols) or phonics.

There’s no shortcut here. The good news is: the content is limited. Learn the sounds, practice out loud, and come back to them repeatedly.

Recommended playlist:

Vowels (quick reference)

  • cop /ɑ/ — like “ah”
  • the /ə/ — “uh” (unstressed)
  • cup /ʌ/ — like “uh” (stressed)
  • boot /u/ — “oo”
  • book /ʊ/ — shorter “oo”
  • beat /i/ — “ee”
  • bit /ɪ/ — “ih”
  • make /eɪ/ — “ay”
  • head /e/ — “eh”
  • had /æ/ — “a” as in “cat”
  • law /ɔ/ — “aw”
  • now /aʊ/ — “ow”
  • bite /aɪ/ — “eye”
  • boy /ɔɪ/ — “oy”
  • go /oʊ/ — “oh”

Consonants (quick reference)

  • web /w/
  • yes /j/
  • father /f/
  • very /v/ (like /f/ but with voice)
  • red /r/ (before a vowel)
  • car /r/ (after a vowel)
  • light /l/ (before a vowel)
  • well /l/ (after a vowel)
  • night /n/ (before a vowel)
  • pen /n/ (after a vowel)
  • mom /m/
  • sing /ŋ/
  • roads /dz/
  • let’s /ts/
  • boss /s/
  • rose /z/ (voiced /s/)
  • thanks /θ/
  • them /ð/
  • just /dʒ/
  • check /tʃ/
  • she /ʃ/
  • Asia /ʒ/
  • try /tr/
  • dry /dr/
  • pet /p/
  • bed /b/
  • too /t/
  • do /d/
  • kiss /k/
  • go /g/
  • how /h/

Channel recommendation:

Speak Out Loud

Speaking quietly and speaking out loud are very different.

Using New Concept English Book 3/4 as an example:

  1. listen carefully first
  2. read out loud
  3. at the beginning, don’t worry about linking or pausing—just focus on clear, accurate sounds
  4. once you can read smoothly, start mimicking the speaker’s rhythm and intonation

Real Conversations: The “Yeah… yeah… yeah” Phase

When you first try to communicate in pure English, it’s common to feel this awkward gap:

In your head you want to say:

“I can’t believe we think the same way.”

But what comes out is just:

“Yeah.”

And then:

“Yeah… yeah… yeah…”

This is normal.

Many people speak like this at first:

  1. think in Chinese
  2. translate into English in their head
  3. search for the right word
  4. get stuck and freeze

After enough listening and speaking practice, your brain starts building “reflex phrases”. In certain situations you’ll respond instantly without translating.

(The simplest proof is that people can blurt out things like “f**k” and “sh*t” without any effort.)

Sing English Songs

If you’re already in love with melodies, why not sing?

Lately I’ve been into Troye Sivan. Here’s a simple way to practice:

  • organize the lyrics
  • write down unknown words and learn them first
  • sing along (yes, I actually bought a microphone)

Clear motivation + strong goal + real interest = surprisingly good results.

Prepare Common Questions for Yourself

To build “English intuition”, prepare a set of common topics and practice them repeatedly until you can speak smoothly and reorder sentences freely:

  • introduce yourself
  • introduce your hometown
  • how life has been recently
  • a few things that make you happiest
  • a small but unforgettable memory
  • your opinion on a product (e.g., iPad Pro)

Less is more. Pick a few and practice a lot. If you have great channels to recommend, open an issue.

Be Brave: Talk to Strangers

CEFR Task Goals

  • A1–A2: complete short introductions, shopping, directions, and appointment interactions.
  • B1: explain experiences, plans, and reasons coherently and handle common follow-ups.
  • B2: participate fairly naturally, explain specialist matters, and maintain a position.
  • C1–C2: organise flexibly and negotiate fine meaning in complex or pressured settings.

Inspectable Practice and Feedback

Choose a real scenario and record an unscripted two-minute first take. Transcribe it and mark only pauses, chunks, grammar, or pronunciation that affect understanding. Repair one to three issues and record again after a delay. Ask a listener to restate the message from the audio alone.

Score 0–2 for task completion, comprehensibility, organisation, interaction, and accuracy/range. A different accent is not an error; prioritise misunderstandings and persistent blocks.

Seven Days, Thirty Days, Twelve Weeks

  • Seven days: four 60–120 second outputs in one scenario, each with first and second takes.
  • Thirty days: one monologue, one role-play, and one real or peer conversation weekly; track recurring pauses and repair strategies.
  • Twelve weeks: add unfamiliar questions, time pressure, and real audiences; finish with a five-to-ten-minute presentation, interview, or meeting.

Prev: Reading
Next: Writing

Explore Related Technologies & Topics

View All 9,111+ Repositories →

Discover similar high-velocity repositories, agent skills, and OpenAPI specifications across the ecosystem.

#chinese Browse chinese catalog → #english-learning Browse english-learning catalog → #tutorial Browse tutorial catalog → ⚡ MCP Servers Model Context Protocol → 🧠 Agent Skills Prompt & rule files → 🔌 OpenAPI Specs REST & Swagger schemas →