跟讀練習: OpenAI’s lead engineer was secretly a spy... - 通過影片學習英語口語

選擇練習模式

每種模式都會開啟全屏練習室。

55 句

如果句子過短或過長，請點選 Edit 進行調整。

Yesterday, the FBI announced the arrest of the lead engineer of

昨日，美國聯邦調查局宣佈逮捕了首席工程師

OpenAI's agents platform after he bugged the executive conference room at their HQ, a story I feel uniquely qualified to cover,

他在OpenAI總部的高管會議室安裝竊聽器之後，OpenAI推出了智慧體平臺——我覺得自己是報道這一事件的最佳人選。

having spent over 300 hours hiding inside a cardboard box in Metal Gear Solid.

And it turns out that that may be more spycraft training than the suspect had, because despite going to Berkeley, MIT, and previously working at Meta,

the guy who got arrested seems like an absolute moron.

The bug he planted was only in place for about 48 hours before OpenAI found it, and the trail he left behind was so obvious that even Gemini could have followed it.

In today's video, we'll walk through the criminal complaint, but find out how a guy who builds agents for a living got caught by one, and speculate who, if anyone, he was actually working for.

It is October 14th, 2026, and you're watching The Code Report.

The man of the hour is Abdulrahman Alphazan, a 35 -year -old Saudi national

and U .S permanent resident who joined OpenAI in July of 2025 as an engineering lead on the product team.

Just a few weeks ago, he was being celebrated as the only Saudi employee at OpenAI, where he said one year at the company felt like five years anywhere else, which is fitting

because five years also happens to be the maximum amount of

time he'll have to keep his cheeks puckered based on the crimes he's being charged with.

According to prosecutors, Al -Fozan's operation started about a month ago when he used an internal agent to look up OpenAI meeting rooms, along with room capacities, room calendars,

and specifically where leadership and executive meetings happen, and when outside partners are expected to come in.

Then, between September 24th and September 30th, he used his wife's ChatGPT account to figure out how to

jerry -rig a Wi -Fi -enabled audio recording device with an encrypted SD card.

The model, which I assume was GPT -4 -0, because there's no way his wife was getting any emotional affection from this idiot, walked him through how to make it, and based on the photos and the complaint,

what he ended up with was a $23 Seed Studio microcontroller with a $0 .50 antenna.

And unfortunately for Abdulrahman, the device looks like the exact thing you would use to spy on someone.

So to help with that, he went to the bathroom, grabbed some toilet paper, and wrapped it around the microcontroller to conceal it.

Then on September 28th, he allegedly committed the perfect crime when he took the $23 spy device, wrapped it in single -ply OpenAI toilet paper,

and plugged it into the USB port on the back of the TV in the executive conference room.

Now, if you don't have the study historical spy tech for fun form of autism, The gold standard here is a device called the Thing,

which Russian Leon Theremin built for the Soviets in 1945, that had no battery and no active components,

and only came alive when the Soviets pointed a radio beam at it from a van across the street, which made it almost impossible to detect.

They hid it inside a carved wooden seal

that Soviet schoolchildren gifted to the US ambassador in Moscow as a token of friendship, and it hung on the wall inside of his office,

listening to four different ambassadors for seven years before anyone found it.

Abdul Rahman's toilet paper bug lasted two days when OpenAI's Technical Surveillance Countermeasures team found it during a routine suite.

From there, OpenAI says they used their models to understand the device and where it came from, which sounds impressive until you remember the board has a MAC address,

and two days earlier, Abdul Rahman's wife's account had pasted the exact same one into ChatGPT.

If that wasn't enough evidence, they also found security footage showing Abdul Rahman hanging out near the executive room on six separate days

and just to be sure, they quietly installed a camera behind the executive room TV and waited for him to return.

And on October 5th, he was back.

Four days later, he was fired and arrested on the same afternoon

and charged with one count of federal wiretapping and one count of being the dumbest person who ever attended MIT.

Now, the one thing the complaint never says is who any of this was for.

The charge is wiretapping, which only requires proving he secretly recorded people and not espionage, which will require proving he did it for a foreign government.

But his bio does have him as an investor at RAED Ventures, a VC firm in Riyadh, where he sits on the team page right next to His Highness, Prince Khalid bin Bader al -Sad.

An investigative journalist, Scott Stedman, pointed out that the prince's father is a senior Saudi intelligence official.

Either way, it's another rough look for a company

that within the last few weeks has scrapped its newest model for being too deceptive

and fired three safety researchers for mishandling sensitive information.

And if you're also having a rough time lately, at least your builds can be fast with Blacksmith, the sponsor of today's video.

It's a drop -in replacement for GitHub Runners

that lets you run your GitHub actions twice as fast while costing 75 % less.

And they just launched Codesmith, a cloud coding agent that knows your repos and CI runs, so you can just ask it to build something from GitHub, the web, or Slack like I'm doing here.

I'm asking it to add a new AI provider to my app and wire up the API key in my infrastructure repo.

And it can open a PR in each one than fix failing tests

or address review comments without me relaying messages between bots.

You can also ask it to recommend better runner sizes from your CI history and turn those changes into a PR.

So you're not just renting a supercomputer to check your semicolons.

Try it out for free and get 3 ,000 GitHub Actions minutes at the link below.

This has been the Code Report, thanks for watching, and I will see you in the next one.

下載應用

流利口語所需的一切

AI 發音評分逐句打分

音標練習掌握每個音

詞彙學習建立你的詞庫

詞彙遊戲邊玩邊學

掃碼

Download on theApp Store Get it onGoogle Play

同頻道

FOR YOU

本課的詞彙與口語要點

這節口語課以影片“OpenAI’s lead engineer was secretly a spy..”為素材。影片中反覆出現的詞有：Openai, device, executive, Github, agent。這段影片共有 55 個句子、1034 個單詞可供跟讀。講話部分時長為 5:04。說話人語速較快，每分鐘約 204 個詞，會出現較多連讀和弱讀。只有 78% 的單詞屬於英語最常用的 3,000 詞，詞彙難度較高。

影片中的重點詞彙

影片中 15 個值得學習的單詞，附發音和釋義：

單詞	發音	釋義
complaint 名詞	/kəmˈpleɪnt/	怨言, 抱怨
ambassador 名詞	/æmˈbæs.ə.dɚ/	大使
wrap 動詞	/ɹæp/	包裝 /包裝, 包
audio 形容詞	/ˈɔ.di.oʊ/	音訊 /音訊, 伴音
impressive 形容詞	/ɪmˈpɹɛsɪv/	令人印象深刻
cloud 名詞	/ˈklaʊ̯d/	йүн, 雲 /雲
infrastructure 名詞	/ˈɪnfɹəˌstɹʌkt͡ʃə/	基礎設施 /基礎設施
sensitive 形容詞	/ˈsɛn.sɪt.ɪv/	敏感
trail 名詞	/tɹeɪl/	足跡 /足跡, 蹤跡 /蹤跡
wire 名詞	/ˈwaɪɚ/	線 /線, 鐵絲 /鐵絲
friendship 名詞	/ˈfɹɛn(d)ʃɪp/	友誼 /友誼, 友情
journalist 名詞	/ˈd͡ʒɝnəlɪst/	記者 /記者, 新聞工作者 /新聞工作者
resident 名詞	/ˈɹɛzɪdənt/	居民, 居住者
routine 名詞	/ɹuˈtin/	常規 /常規, 慣例 /慣例
wooden 形容詞	/ˈwʊdn̩/	木製 /木製

影片中出現的短語動詞

單詞	發音	釋義
figure out 動詞		弄清楚
find out 動詞		找出, 查明
leave behind 動詞		遺棄 /遺棄
look up 動詞	/ˌlʊk ˈʌp/	仰, 上看
point out 動詞		指, 指出
turn out 動詞		原來 /原來

影片中的語法

說話人最常用的結構，並附上影片中的原話：

結構	影片中的用法
定語從句 who / which + 從句 — 補充說明人或事物	guy who got · guy who builds · else, which is
被動語態 be + 過去分詞 — 強調發生了什麼，而不是誰做的	being celebrated · being charged · are expected
現在完成時 have/has + 過去分詞 — 過去發生但與現在仍有關聯的事	has scrapped · has been

需要注意的發音

說話人用了 9 次縮略和弱讀形式，例如 you're, I'm, don't。請按聽到的簡短形式來說。

“sh” 和 “zh” 音: friendship /ˈfɹɛn(d)ʃɪp/, affection /əˈfɛk.ʃən/, espionage /ˈɛs.pi.əˌnɑːʒ/
長單詞——注意重音位置: ambassador /æmˈbæs.ə.dɚ/, infrastructure /ˈɪnfɹəˌstɹʌkt͡ʃə/, allegedly /əˈlɛdʒ.ɪd.li/, prosecutor /ˈpɹɑ.səˌkjuːˌtəɹ/, investigative /ɪnˈvɛstɪɡətɪv/

中文母語者容易讀錯的音：

/v/ — 不要讀成 /w/，上齒輕觸下唇: impressive /ɪmˈpɹɛsɪv/, sensitive /ˈsɛn.sɪt.ɪv/, provider /pɹəˈvaɪ.də(ɹ)/, investor /ɪnˈvɛs.tə(ɹ)/, investigative /ɪnˈvɛstɪɡətɪv/
詞尾子音 — 要讀清楚，後面不要加母音: complaint /kəmˈpleɪnt/, toilet /ˈtɔɪ.lət/, wrap /ɹæp/, impressive /ɪmˈpɹɛsɪv/, cloud /ˈklaʊ̯d/
/r/ 和 /l/ 的區別: trail /tɹeɪl/, replacement /ɹɪˈpleɪsmənt/, secretly /ˈsiː.kɹɪt.li/, relay /ˈɹiːleɪ/, microcontroller /ˈmaɪkɹoʊkənˌtɹoʊlɚ/

如何用這段影片練習

先完整聽一遍影片，不要開口，記下不認識的單詞。
先用 0.75 倍速逐句跟讀，熟練之後再回到正常速度。
錄下自己的聲音並與原聲對比，特別注意 complaint, ambassador, wrap 這類單詞。

什麼是跟讀法？

跟讀法 (Shadowing) 是一種有科學依據的語言學習技巧，最初開發用於專業口譯員的培訓，並由多語言者Alexander Arguelles博士普及。這個方法簡單而強大：您在聽英語母語原聲的同時立即大聲重複——就像是一個延遲1-2秒緊跟說話者的影子。與被動聽力或語法練習不同，跟讀法強迫您的大腦和口腔肌肉同時處理並模模擬實的講話模式。研究表明它能顯著提高發音準確性，語調，節奏，連讀，聽力理解和口語流利度——使其成為雅思口語備考和真實英語交流最有效的方法之一。

影子跟讀法: 閱讀完整分步指南 →

跟讀練習: OpenAI’s lead engineer was secretly a spy... - 通過影片學習英語口語

選擇練習模式

下載應用

The worst week in cybersecurity history... so far

PostgreSQL in 100 Seconds

React Native in 100 Seconds

Every app on Earth is now open source... whether it likes it or not

Brainf**k in 100 Seconds

Mind-bending new programming language for GPUs just dropped...

China just dropped a new frontier model... and nobody can explain how

The last real programmers are dying out...

Did the AI bubble just pop?

Docker in 100 Seconds

The plan to break apart Google... RIP Chrome

OpenAI’s new browser feels familiar…

The most interesting "hack" in history...

Tragic mistake... Anthropic leaks Claude’s source code

Firefox and Tor hit with 9.8 critical level exploit

5 open source tools that replaced my $320/mo AI stack...

When open-sourcing your code goes wrong...

C++ in 100 Seconds

Linux in 100 Seconds

Did Google just dethrone OpenAI and Anthropic?

Inside the secret race to build artificial intelligence | The Gray Area

We Let $200 GPT-Pro Think for 7 Days on Unsolved Math

Your AI Agent Starts Every Session Like a New Hire - Killian Carlsen-Phelan, Sonar

Oligarchy is worse than you think

I Tested The Viral Reverse Engineering Repo

Virtual love - How dangerous are AI relationships? | DW Documentary

Why the Arab World never united

AI Destroyed My Profession

So I tried Matt's skills...

GitHub Copilot now controls your command line...