シャドーイング練習: Nobody is hiring vibe coders. Sorry - 動画で英語スピーキングを学ぶ
レッスンを作成中...
1
Coding interviews are dead.
2
You don't need to learn how to code anymore.
3
At least that's what your social media feed keeps telling you.
4
But here's what the data actually says.
5
Since 2023, technical interviews have nearly doubled in the US and they're becoming more challenging than ever, not easier.
6
So what's going on here?
7
Because getting this wrong either way can cost you.
8
Either you keep grinding LeetCode style interviews and prep for an interview that no longer exists,
9
or you misread the industry and walk away from a career that is very much still hiring.
10
So let's ignore what people are saying and look at what companies actually do whenever their money is on the line,
11
because the facts tell a completely different story from what you're hearing.
12
So a few months ago, I was actually going through my own interview process and I interviewed with 11 different companies.
13
Every single one of them that I got far enough with in the process went to the technical interview.
14
These were a mix of coding interview challenges, system design questions, and anything that was related to whatever that company's specific niche was.
15
So if it was database, I got a lot of database questions and so on.
16
Now, two of these were actually the classic solve Sudoku problem.
17
If you've ever went through that technical interview process, you might have seen this.
18
This is just where you have to solve a Sudoku board.
19
essentially just make sure that the board is either valid or not.
20
Now, maybe I just got unlucky and I picked the 11 most old school companies in the industry, but that's not what the data actually shows.
21
According to CoderPad, this is the platform where a lot of these companies actually conduct these interviews.
22
Technical assessments are up 48% globally since mid-2023,
23
and 43% of those companies still conduct the classic algorithmic assessments.
24
Now, the rest didn't drop the technical bar they moved it to system design interview questions pair programming
25
and more of real world debugging situations
26
but okay maybe these are just the companies
27
that are slow to adapt enterprises
28
that are still stuck in 2019 interviewing like the ai wave never happened i guess
29
that could be a fair objection but the issue here is
30
that the strictest interviews are actually conducted by the companies building these AI tools. For example,
31
OpenAI's most commonly reported interview question is implementing an LRU cache from scratch live as you explain your reasoning through the process.
32
An LRU cache is just a hash map plus a linked list.
33
The exact obsolete fundamentals your feed keeps telling you are now irrelevant.
34
Anthropic, on the other hand, publishes its official policy for candidates and for live interviews, it says on it, no AI assistance unless we indicate otherwise.
35
We're curious to see how you think and how you reason through problems in real time.
36
And Meta went the other way entirely.
37
They built a round where you're actually encouraged to use AI.
38
And the accounts coming out of these say that this round is harder than the classic one.
39
The companies that are closest to the models, the people who know exactly what AI can and cannot do, are the ones testing the hardest.
40
Why do you think that is?
41
Well, before I answer that, since this video doesn't have a sponsor, I want to show you something that Brad and I have been working on.
42
If you've been following this channel for some time, you've probably seen our 20-hour long video courses, whether that's on Udemy or our website.
43
And that's how hundreds of thousands of you have learned.
44
And we're still very proud of those courses and we'll always continue to make those.
45
But we realized we need a different approach for those of you that like to be hands-on.
46
And that's why we launched Start.dev.
47
It's an interactive learning platform where lessons are bite-sized with an in-browser editor.
48
So you can start writing code today.
49
So if you're looking to learn about the tools and languages that are most in demand in the AI era, like Python and TypeScript, or you wanna learn about web development,
50
I highly recommend you check it out.
51
Each course has an AI version of its instructor.
52
So for example, you get AI Brad in each lesson with you.
53
He knows what you're working on.
54
He sees your code and he's designed to give you leading answers, not some long-winded chat GPT responses that knock you out of your flow.
55
And yes, there's an AI version of me too.
56
And no, you can't make me say weird stuff people have already tried.
57
There's thousands of students on the platform already and absolutely loving it, so it's free to get started, but if you want 20% off premium access,
58
use the coupon code down in the video description.
59
Okay, so why do AI companies test the hardest?
60
Well, it comes down to what everyone got wrong about interviews in the first place.
61
The people
62
that bought this AI killed coding interviews narrative have a fundamental
63
misunderstanding of what the technical interview was even there for and really what the job of software developer actually requires.
64
The thinking goes that if AI writes the code, now we're somehow out of the loop.
65
As if writing the code was our job all along.
66
Well, the interview was never about the code either, and honestly, I'm a little bit shocked that we even have to say this, but here we are.
67
It wasn't about the syntax.
68
In fact, lots of interviewers let you choose the language
69
that you test in because that was never the part that actually matter to begin with.
70
What they're grading you on is your judgment behind every solution.
71
Can you think through a problem?
72
Why this decision over that decision?
73
And what are the trade-offs and alternatives?
74
Where does this problem break at scale?
75
Now, a large chunk of the interviews
76
that I've taken were usually some kind of system design where I'm implementing some kind of different pattern
77
or working through a larger idea.
78
And most of the interviews I've taken weren't even coding interviews at all.
79
I would say probably less than 50% of them were.
80
And most of them were actually just some kind of scratch pad or a situational walkthrough on a whiteboard.
81
And on the other side where I've interviewed a lot of candidates in my days, the best interviews I've ever conducted were not the candidates that gave me the answer,
82
but the ones that had a great conversation with me through their process, walked me through all their solutions and ideas.
83
And the way they reasoned told me
84
that they would really be able to add value to my team and allowed me to see really what's under the hood, not that they just memorize something.
85
There's a funny thing happening in Meta's AI assisted interview round where candidates keep claiming
86
that somehow the models are dumber during the interview than they were the day before when they were practicing.
87
Same models, same prompts, and all of a sudden it's writing buggy code and it's not pointing out obvious mistakes.
88
Now, it's not confirmed by Meta that they purposely make their models dumber, but honestly, it wouldn't surprise me.
89
It's a great way to watch where a candidate catches obvious issues and where they just glaze over them.
90
So imagine thinking that AI somehow ended all of this and nobody was going to ask you to explain what you built.
91
Exposing someone has never been easier.
92
A few follow-up questions and it's over.
93
Now, I'm not gonna tell you that nothing has changed because things are changing, just not in the direction that a lot of us think.
94
Developers have been complaining about interviews since long before AI.
95
They've been claiming that things are too abstract, too much memorization and nothing like the real job.
96
And for many years, really nothing has moved but because of AI, it's actually forcing companies to really evaluate this process and change things up.
97
Companies are starting to swap single function puzzles for more real world simulations.
98
They're challenging candidates on multifile code bases and adding more realistic debugging scenarios and working inside of existing code bases
99
instead of greenfield toy projects.
100
Let's invert this binary tree and more.
101
Here's our code, make it better.
102
That's much closer to the actual day-to-day engineering work while still leaning on the deep-rooted concepts that you needed to know before.
103
Now, this is really the first change in years that I would call progress when it comes to the interview process.
104
But here's where the industry is split.
105
Nobody agrees on how much AI should be allowed.
106
Roughly a third of the companies still ban it entirely, half allow it, and the rest decide on a case-by-case basis.
107
And where it is allowed, I've seen companies ask you for your prompt history, grill you deeply on the take-home projects that they give you,
108
and then have you extend it live on a call.
109
So no matter where a company lands on this, whether it's partial AI, full AI, or somewhere in between, there's still no version where knowing less is a winning move.
110
The pattern of interviews has been evolving over the past 70 years.
111
Think of it as a stack.
112
At the bottom, the mechanical stuff.
113
Typing out working code.
114
In the middle, we have implementation, picking an approach and making it work.
115
And at the top, we have the judgment.
116
This is the trade-offs, catching what's wrong and defending our decisions.
117
Producing working code used to prove the layers above it.
118
You couldn't get something working without understanding the layers underneath it.
119
Every era of tooling has always deleted the bottom of the stack and just moved the test upwards.
120
For example, compilers deleted assembly, frameworks deleted boilerplate, and nobody is interviewing you on memory management for a React job.
121
AI is deleting yet another layer, which is producing the actual code, so the test is just moving up another layer.
122
Now, as AI adoption grows, we're gonna see more companies adjust their interviews to include it and focus less on producing the actual code.
123
But just because we may be writing less code by hand doesn't mean
124
that the bar dropped the bar just moved up the stack.
125
And this is the part that really gets me upset whenever I see this rhetoric online.
126
AI bros are telling beginners they don't need to learn how to code, that knowledge is obsolete, they just need to prompt harder.
127
That's like telling a pilot that they can skip flight school because we have autopilot.
128
Autopilot is great, but it's also not the reason why anybody gets to sit in that seat.
129
Nobody is hiring vibe coders.
130
And the market is telling you this with data.
131
74% of developers are saying they're struggling to land a job, while 78% of tech leaders say they can't find qualified people.
132
Now, both are true at once.
133
The jobs exist, but qualified people are getting harder to find.
134
We're about to experience the biggest demand for developers probably ever, and yet the public messaging is manufacturing the biggest shortage of qualified developers.
135
Now, on the other side, we have the real engineers don't touch AI crowd.
136
Well, I'm going to say something that will upset a few of you, but outside of some very specific use cases, I would not hire a developer that refuses to leverage AI tooling.
137
And there's also common consensus out there amongst hiring managers that agree with me.
138
The data is somewhere out there.
139
I didn't pull it up for this video, but look it up.
140
This is a common belief here.
141
And at this point, it's really just getting irresponsible not to use AI and the pushback against it, which I can understand there's been a whole pendulum
142
that swung the wrong way it's getting to the point where
143
even the resistance against it is now a problem now i want to give you a quick example here
144
so back in 2017 i hired my first developer to work on this project
145
that i built out it scaled up
146
and it was time to add some team members now this guy was good
147
but he refused to move out of the php stack
148
and would always complain about the frameworks that i was using his claim was
149
that real developers should be able to write their own web server from scratch and that frameworks like Django and React, well, those were just for noobs.
150
Imagine holding that belief today.
151
I think we're past the AI is just a tool phase and really need to adapt here.
152
You're not only gonna be graded on your ability in the future, you're also gonna be graded on your AI fluency as well.
153
Now, I don't make the rules.
154
That's just where the market is at right now and it's probably where we're going in the future.
155
So to close this video out, no coding interviews aren't dead.
156
There are more of them now than ever, and they're testing the same thing they've always tested.
157
Whether you know what you're doing or not, AI didn't change that.
158
So keep the fundamentals, get good with the tools, and understand everything you build.
159
Because the developers that are doing both right now aren't scared.
160
They're thriving.
📺 同じチャンネル
✨ おすすめ動画
B21
22 文
I'm an ex-Google interviewer. You're doing LeetCode wrong.
Business EnglishInterview EnglishUniversity Students
B217
130 文
How AI Is Killing The Value Of A College Degree
Business EnglishInterview EnglishUniversity Students
B21
132 文
Why Companies Are Quietly Rehiring Software Engineers
Business EnglishUniversity StudentsWorking Adults
B25
69 文
7 Programming myths that waste your time
American AccentBusiness EnglishUniversity StudentsWorking Adults
B21
193 文
J*b hunting in the big '26
Business EnglishInterview EnglishUniversity Students
B22
231 文
My Senior UX Design Interview Experience | tips, questions, prep, & more!
Business EnglishInterview EnglishUniversity Students
B22
109 文
Doing This (Almost) GUARANTEES You Get Hired In A Job Interview!
Business EnglishInterview EnglishUniversity Students
B22
78 文
5 Things You Should Never Ask ChatGPT | Vantage with Palki Sharma
Business EnglishDaily ConversationUniversity Students
C11
315 文
how AI broke the SWE Career Ladder
Business EnglishInterview EnglishUniversity Students
B21
181 文
8 Years of Data Analytics Advice in 12 Minutes
Business EnglishInterview EnglishUniversity Students
B24
128 文
Please get a job...my advice to Gen Zs and young millennials...
Business EnglishInterview EnglishUniversity Students
このレッスンの語彙とスピーキングのポイント
このB2レベルのスピーキングレッスンは、動画「Nobody is hiring vibe coders. Sorry」を教材にしています。 繰り返し出てくる語は次のとおりです：interview, code, companies, developer。 この動画には、シャドーイング用の文が160文、単語が2257語あります。 音声の長さは12:25です。 話す速さは速く、1分あたり約182語です。音のつながりや弱く発音される音が多くなります。 単語の83%は英語の頻出3,000語に含まれます。残りは練習の前に確認しておきましょう。
この動画の重要語彙
動画の中で特に難しい単語15語を、発音と意味つきで紹介します。
| 単語 | 発音 | 意味 |
|---|---|---|
| stack 名詞 | /stæk/ | 塚 |
| scratch 動詞 | /skɹæt͡ʃ/ | 引っ掻く |
| autopilot 名詞 | /ˈɔtoʊˌpaɪlət/ | オートパイロット |
| debug 動詞 | /diːˈbʌɡ/ | デバッグする |
| cache 名詞 | /kæʃ/ | キャッシュ |
| obsolete 形容詞 | /ˈɑbsəlit/ | 廃れた |
| adapt 動詞 | /əˈdæpt/ | 適応する |
| implement 動詞 | /ˈɪmpləmənt/ | 実行する, 実現する |
| compiler 名詞 | /kəmˈpaɪlɚ/ | コンパイラ |
| fluency 名詞 | /ˈfluːənsi/ | 流暢 |
| invert 動詞 | /ɪnˈvɜɹt/ | 反転 |
| memorization 名詞 | /ˈmem(ə)ɹaɪˌzeɪʃn̩/ | 暗記 |
| memorize 動詞 | /ˈmɛm.əˌɹaɪ̯z/ | 記憶する, 覚える |
| misread 動詞 | /mɪsˈɹiːd/ | 誤読する |
| misunderstand 動詞 | /mɪs.ʌn.də(ɹ)ˈstænd/ | 勘違いする, 誤解する |
動画に出てくる句動詞
| 単語 | 意味 |
|---|---|
| point out 動詞 | 指す, 指し示す |
繰り返し練習したい文
動画に出てくる短く完結した文です。日常会話でそのまま使えます。
- You don't need to learn how to code anymore.
- Why do you think that is?
- He knows what you're working on.
- Can you think through a problem?
- Let's invert this binary tree and more.
この動画の文法
話し手がよく使っている文型を、動画の実際の表現とともに紹介します。
| 文型 | 動画での表現 |
|---|---|
| 現在完了進行形 have/has been + -ing — 以前に始まり今も続いている動作 | have been working · you've been following · have been complaining |
| 現在完了形 have/has + 過去分詞 — 過去の出来事が今も関係している | you've probably seen · have learned · have already tried |
| 受動態 be + 過去分詞 — 誰がするかより、何が起きるかに焦点を当てる | are still stuck · are actually conducted · is split |
| 関係詞節 who / which + 節 — 人や物について情報を加える | people who know · layer, which is |
注意したい発音
話し手はwe're, you're, they'reなど、短縮形や弱形を57回使っています。聞こえたとおりの短い形で発音しましょう。
- 「th」の音: algorithmic /ˌælɡəˈɹɪðmɪk/, anthropic /ænˈθɹɒp.ɪk/, walkthrough /ˈwɔːk.θɹuː/, thrive /θɹajv/, underneath /ˌʌndɚˈniθ/
- 「sh」と「zh」の音: cache /kæʃ/, memorization /ˈmem(ə)ɹaɪˌzeɪʃn̩/, preposition /ˌpɹɛpəˈzɪʃən/, situational /sɪtjuːˈeɪʃənəl/, hash /ˈhæʃ/
- 長い単語（アクセントの位置に注意）: developer /dɪˈvɛləpɚ/, autopilot /ˈɔtoʊˌpaɪlət/, algorithmic /ˌælɡəˈɹɪðmɪk/, memorization /ˈmem(ə)ɹaɪˌzeɪʃn̩/, misunderstand /mɪs.ʌn.də(ɹ)ˈstænd/
日本語話者が苦手な音：
- /r/ と /l/ の区別 — /l/ は舌先を歯茎につけ、/r/ はどこにもつけない: algorithmic /ˌælɡəˈɹɪðmɪk/, irresponsible /ˌɪɹ.əˈspɑn.sə.bəl/, grill /ɡɹɪl/, leverage /ˈlɛv.(ə.)ɹɪd͡ʒ/, irrelevant /ɪˈɹɛləvənt/
- /v/ — /b/ にならないように、上の歯を下唇に当てる: developer /dɪˈvɛləpɚ/, invert /ɪnˈvɜɹt/, interviewer /ˈɪntɚvjuɚ/, thrive /θɹajv/, evolve /ɪˈvɑlv/
- /f/ — 「フ」ではなく、上の歯と下唇で出す: qualify /ˈkwɑ.lɪ.faɪ/, fluency /ˈfluːənsi/
この動画での練習方法
- まず声を出さずに動画を最後まで聞き、知らない単語をメモします。
- まず0.75倍速で一文ずつシャドーイングし、慣れてきたら通常の速度に戻します。
- 自分の声を録音して元の音声と比べます。stack, scratch, autopilotなどの単語に特に注意しましょう。
シャドーイングとは？英語上達に効果的な理由
シャドーイング（Shadowing）は、もともとプロの通訳者養成プログラムで開発された言語学習法で、多言語習得者として知られるDr. Alexander Arguelles によって広く普及されました。方法はシンプルですが非常に効果的：ネイティブスピーカーの英語を聞きながら、1〜2秒の遅延で声に出してすぐに繰り返す——まるで「影（shadow）」のように話者を追いかけます。文法ドリルや受動的なリスニングと異なり、シャドーイングは脳と口の筋肉が同時にリアルタイムで英語を処理・再現することを強制します。研究により、発音精度、抑揚、リズム、連音、リスニング力、そして会話の流暢さが大幅に向上することが確認されています。IELTSスピーキング対策や自然な英語コミュニケーションを目指す方に特におすすめです。