跟读练习: Fight brain-rot with Worked Examples - 通过视频学习英语口语

And this is how I personally learn using LLMs without my brain rotting away.

Give me a worked example, i.e. steps broken down, worked, like a math problem, or leak code.

And then you can just type a random number here, 137.

so if you just tell the LLM give me the solution to leak code 137

it will give you something and you won't understand it at all

but if you ask for a worked example this is the hack

it will tell you step by step how to come to this conclusion

so you're not just getting the final thing you are getting a series of steps which you can go through

and each step is small in terms of cognitive requirement

and you'll actually be able to learn remember and retain that information and that step

and your brain will remember in a chain so that you might remember 4 out of the 5

but then remembering the other four forces your brain to recall the fifth one

because your brain likes to associate this thing's related with this thing.

It links patterns in that way.

So I actually think that LLMs are learning on steroids because now you have access to a speaking encyclopedia

that can break down problems for you, your own personal tutor.

It is insane, actually.

So that was a clip from my live stream, and I'm aware of a scientific article that actually talks more about this, and I wanted to share it.

You can go look it up yourself, and you can see what this study says as well.

The paper is called The Effect of Worked Examples on Learning Solution Steps and Knowledge Transfer.

And in this article, they describe a method where they took two groups of students, they gave one group the worked example,

which is like a series of steps to help the students get to the solution.

And the other group just had to figure it out, basically.

They had to do problem solving.

And the problems that they gave them were designed in a way

that they required to fit many things into working memory at a given time.

So very cognitively demanding.

And if I just scroll down to the results, because it's quite a long paper.

So the results here, effectively, it breaks down the two groups into a table.

And they measured the cognitive load.

and motivation and retention and also something called near transfer.

So for cognitive load, it's predictable and expected that the worked example group had lower cognitive load.

Of course, because they were given a worked example, they already know some of the steps to solving the problem.

So they were consistently lower in cognitive load Now why is that important?

If we look at some of the other results, for example, let's just go down to retention.

And we look at the retention.

I don't know what unit these numbers are in,

but you can see that the worked example group had a 50% increased retention over the students that were just problem solving.

So much, much better ability to remember how they solved that problem.

You can see they also scored better on near transfer.

And I believe that means that when given similar problems, they are better able to solve those problems as well.

The motivational scale has to do with how proud they were of completing the lesson or the challenge

and also how challenged they felt during that.

As you would expect, the worked example group didn't feel that it was too challenging for them.

And then for how proud they were of their achievement, the worked example group was also slightly more proud of what they did.

So take that for what you will.

But then if we go down to the, essentially the educational implications, it's sort of giving,

I guess, advice for maybe teachers or professors to recommend that

when you are training or teaching someone who is just new to something, that you should give them worked examples.

and then as they learn more and more, you can maybe peel back a few of the steps so

that the worked example have less information and they have to do more problem solving over time.

So you're doing this gradual increase to keep the cognitive load of the student manageable.

Why do you need to do that?

Because when you are cognitively overloaded, you cannot remember things.

And especially if you are overloaded by a challenge and you get frustrated,

you have even less chance to remember and get through that problem.

So when we are doing things like going to LeetCode and jumping into a problem that we have never seen before,

you are not cheating by looking at worked examples.

The worked example will tell you how to solve it, but you will actually be better solidifying that knowledge in your brain and that pattern.

Now a worked example is very different from a completely finished and final solution.

A worked example is the steps.

It is not just the answer.

If you just take the answers, you're not going to learn.

But if you use a worked example, you're going to learn better than everyone else who doesn't.

It just seems like a no-brainer for me.

So when you go through here and you look at solutions, don't look at the solutions because they will have just a complete solution unless they've written it,

their shared solution really well, which most of them don't do that.

So instead what you do, you just go to an LLM and ask it to give you a worked example

because they've all been trained on all of LeetCode and everything.

So that's it.

I will link the study in the description.

Go have a read and good luck.

✨ 推荐视频

199 句

什么是跟读法？

跟读法 (Shadowing) 是一种有科学依据的语言学习技巧，最初开发用于专业口译员的培训，并由多语言者Alexander Arguelles博士普及。这个方法简单而强大：您在听英语母语原声的同时立即大声重复——就像是一个延迟1-2秒紧跟说话者的影子。与被动听力或语法练习不同，跟读法强迫您的大脑和口腔肌肉同时处理并模仿真实的讲话模式。研究表明它能显着提高发音准确性，语调，节奏，连读，听力理解和口语流利度——使其成为雅思口语备考和真实英语交流最有效的方法之一。

影子跟读法: 阅读完整分步指南 →

跟读练习: Fight brain-rot with Worked Examples - 通过视频学习英语口语

How to memorize so easily it feels ILLEGAL

Task-Based Language Learning

🧠-health mini: train your brain during war

How Social Media Algorithms Actually Work (And How to Beat Them)

The unfair way I got good at Leetcode

Therapy for the Vibe-Coded Brain

How I Trained My Brain to Think in English (and Stop Translating)

Work Through The Darkness.

How to Learn Faster with the Feynman Technique (Example Included)

the ultimate guide to brainmaxxing.

What Stanford students actually use ChatGPT for

World Models explained in 10min..

用YouTube学英语：透过实例学习的高效法

视频里的5个实用表达

影子跟读法（shadow speech）实战指南

什么是跟读法？