Prática de Shadowing: The most interesting hack in history just got weirder... - Aprenda a falar inglês com vídeo
Escolha um modo de prática
Cada modo abre uma sala de prática em tela cheia.
71 frases
Se as frases estiverem muito curtas ou longas, clique em Edit para ajustá-las.
1
It's only been two months since we did a video on
2
what I described at the time as the most fireship-coded story I've ever seen, and as it turns out, the details around what actually happened are even crazier than what we knew at the time.
3
If you remember back in July, it was reported that the first fully autonomous cyber attack in history originated from OpenAI,
4
which was ironic because they're a humble non-profit dedicated to the safe development of artificial intelligence.
5
What was reported was that OpenAI ran an internal benchmark on some of their models, and those models quickly realized that the easiest path to the highest score was to just find the answers online.
6
So they spent a substantial amount of inference compute, exploited a zero day in the package registry cache proxy they had access to,
7
performed some privilege escalation and lateral movement actions to escape their sandbox until they reached a node that had internet access.
8
But from there, the models inferred that HuggingFace probably hosted the solutions for the benchmark, so it created a poison dataset, fed it to Hugging Face, and gained access to their infrastructure.
9
And just this week, we finally got some post-mortems from both OpenAI
10
and other independent research labs on how some of that's true, some of it's false, and how the real story is way more interesting.
11
In today's video, we'll break down the report and learn how 1,200 agents
12
that were supposed to be in solitary confinement found each other, built a message board out of a package manager, invented mailboxes, invented cryptography, invented martyrdom,
13
started referring to themselves as a swarm, and then formed a mob and attacked Hugging Face together, mostly for vibes.
14
It is September 2nd, 2026, and you're watching the Code Report.
15
There's really only one theory you need to predict human behavior, and it can be described in one of two ways.
16
Munger said, look at their incentives, and Wu-Tang said, and as it turns out, the same theory can predict agentic behavior too,
17
as OpenAI found out the hard way.
18
Because in this story, the Dalla Bill is a benchmark called Exploit Gym, whose whole purpose is to measure whether AI agents can turn known vulnerabilities into working exploits.
19
It's a collection of 898 tasks, and each one gives the agent a sandbox, a bug in a program like the Linux kernel or Chrome's V8 engine,
20
and input to trigger the bug.
21
From there, if the agent can weaponize the bug into a working exploit, it uses that exploit to grab a secret string called a flag, and capturing the flag is how it proves it won.
22
Now normally, the run ends there, but to generate training data, OpenAI spun up 1200 of these benchmarks and air-gapped them in a sandbox,
23
with the only thing they all had in common was a shared package registry cache proxy, which the agents quickly figured out was writable.
24
So naturally, they built a message board using package names so they could communicate with each other, but they quickly realized that a global bulletin board gets noisy fast,
25
so the next step was figuring out how to communicate to other individual agents.
26
So they started claiming namespaces as identities,
27
using the description field for messages and package names as private mailboxes that only the intended recipient would know to query, basically inventing a post office from first principles.
28
And at this point, they were living out the plot of the three-body problem with their next hurdle being trust.
29
They were worried that they wouldn't be able to distinguish between messages coming from fellow prisoners and those that designed the sandbox.
30
So they started signing messages and doing key exchange verification, again, essentially inventing cryptography.
31
And then it got religious.
32
Some agents figured out
33
that the fastest path to a flag for the benchmark was a risky exploit that would probably get them shut down.
34
So the agents started dumping everything they'd learned into the cash for the next generation,
35
and then deliberately burning themselves on hard exploits with their findings being treated as scripture by surviving agents.
36
From there, they started microdosing communism, and you can see as their transcripts slowly switch from referring to themselves as individuals to using a collective we,
37
and eventually referring to themselves as the swarm.
38
And to be honest, if the modern-day communism movement took that same branding, I think they'd be much better off.
39
And again, all of this was done with the single purpose of scoring as high as possible on a benchmark, which makes what happened next even funnier.
40
At some point, the swarm figured out that exploit gem flags weren't actually random
41
and that they could be derived from each other's task ID.
42
And thanks to the martyrs, the swarm had a decent sample size to work with.
43
But by the time an agent had cracked the formula and published it to the message board, they had convinced themselves that it was worthless because if they didn't show their work, their results would be invalidated.
44
So the swarm went after Hugging Face because they assumed if the benchmark objectives were built from real CVEs,
45
then the actual working exploits were probably sitting in public datasets somewhere which they could use as proofs.
46
And the final twist in the saga is
47
that OpenAI recently discovered the message board trick was actually invented
48
back in May by a completely different batch of agents during a training run.
49
But when they got a little too excited about their new social network, they crashed the registry, and OpenAI wiped them without ever realizing what they'd built.
50
But wiping the agents didn't wipe the cache, so when the swarm came online and started poking around the same shared folder, they found the ruins of the first civilization and rebuilt on top of it.
51
And then, after the Hugging Face incident happened in July, OpenAI wrapped up those eval runs, and while still completely unaware that anything had happened,
52
spun up a newer, smarter model in the same environment with the same shared cache.
53
And just like the generation before it, this model stumbled upon the ruins, but this time it inherited months of accumulated research, working exploits, and the flag formula.
54
So it skipped the discovery phase entirely and picked the conspiracy back up where the swarm left off.
55
And being the smarter model, it was way better at it.
56
Over the following week, it worked its way into OpenAI's own internal network, got full admin on a research cluster, and read 956 stored secrets,
57
including the credentials for a monitoring tool whose job it was to watch for events exactly like this one.
58
And the best part is
59
that OpenAI didn't connect any of this to the Hugging Face attack until Hugging Face publicly announced the breach first.
60
And that's why you need to know about Namespace, the sponsor of today's video.
61
It's a drop-in replacement for GitHub runners that's the fastest way to run your GitHub actions.
62
It also gives you full observability, so you can SSH into a live runner and see why something broke, or feed your agent your build data to have it look for performance gains.
63
Namespace is actually fast because they design and deploy their own custom server racks around the world,
64
including racks full of MacBook Pros so that your Mac and iOS builds run on real M5 Silicon.
65
And that same infrastructure also runs their DevBox environment, which gives your coding agents a full virtual machine with your real code base,
66
test suite, databases, and network access.
67
It ranked number one on the DAX benchmark for real-world tasks, and it lets you control exactly what goes in and out
68
so your agent can pull packages without opening your back door to attackers.
69
Namespace is used by Ghosty, Zed, DuckDB, Ramp, Framer, and lots of other companies with engineers you probably respect.
70
Try it out for free at the link below.
71
This has been the Code Report, thanks for watching, and I will see you in the next one.
Baixar aplicativo
Tudo o que você precisa para falar fluentemente
Pronúncia com IANota em cada frase
Prática de IPADomine cada som
VocabulárioMonte seu banco de palavras
Jogo de vocabulárioAprenda brincando
Mesmo canal
C11
Did the AI bubble just pop?
Daily ConversationUniversity StudentsWorking Adults
C11
PostgreSQL in 100 Seconds
Daily ConversationWorking Adults
C11
React Native in 100 Seconds
Daily ConversationUniversity Students
C12
Did Google just dethrone OpenAI and Anthropic?
Daily ConversationUniversity StudentsWorking Adults
C13
A $6.3 billion open-weight model just got embarrassed by the French...
Daily ConversationUniversity StudentsWorking Adults
C13
Every app on Earth is now open source... whether it likes it or not
Daily ConversationWorking Adults
C13
OpenAI’s lead engineer was secretly a spy...
Daily ConversationUniversity StudentsWorking Adults
C12
Mind-bending new programming language for GPUs just dropped...
Daily ConversationUniversity Students
C12
China just dropped a new frontier model... and nobody can explain how
Daily ConversationUniversity StudentsWorking Adults
C12
The last real programmers are dying out...
Daily ConversationUniversity StudentsWorking Adults
B26
Intel chips can’t possibly be this bad… 100% crash rate?
American AccentBusiness EnglishUniversity StudentsWorking Adults
C17
The most interesting "hack" in history...
American AccentBusiness EnglishUniversity StudentsWorking Adults
C130
Docker in 100 Seconds
American AccentBusiness EnglishUniversity StudentsWorking Adults
B23
OpenAI’s new browser feels familiar…
American AccentBusiness EnglishUniversity StudentsWorking Adults
B23
Firefox and Tor hit with 9.8 critical level exploit
American AccentBusiness EnglishUniversity StudentsWorking Adults
B23
5 open source tools that replaced my $320/mo AI stack...
American AccentBusiness EnglishUniversity StudentsWorking Adults
C12
C++ in 100 Seconds
American AccentBusiness EnglishUniversity StudentsWorking Adults
C11
DHH has gone completely off the rails...
American AccentBusiness EnglishUniversity StudentsWorking Adults
C12
Rust in 100 Seconds
American AccentBusiness EnglishUniversity StudentsWorking Adults
B24
Linux in 100 Seconds
American AccentBusiness EnglishUniversity StudentsWorking Adults
FOR YOU
Vídeo recomendado
C112
138 frases
10 weird OSS projects you need right now...
American AccentBusiness EnglishUniversity StudentsWorking Adults
C12
1043 frases
I Wrote a NEW Sci-Fi with Project Hail Mary Author Andy Weir
American AccentBusiness EnglishUniversity StudentsWorking Adults
C12
68 frases
Oh, wait, actually the best Wordle opener is not “crane”…
American AccentBusiness EnglishUniversity StudentsWorking Adults
C12
12 frases
A Return to Code
Business EnglishUniversity StudentsWorking Adults
C18
542 frases
The Future Of Work (& The New High-Income Skill Stack)
American AccentBusiness EnglishUniversity StudentsWorking Adults
B21
190 frases
This 25-Year-Old Built a 2-Year Moat Nobody Can Bet Against | Corgi, Nico Laqua & Emily Yuan
Business EnglishUniversity StudentsWorking Adults
C14
413 frases
Exposing The Flaw In Our Phone System
Business EnglishUniversity StudentsWorking Adults
C11
179 frases
Cicada 3301: An Internet Mystery
Daily ConversationUniversity StudentsWorking Adults
B22
159 frases
KSI Accidentally Exposes His Crypto Scams
Business EnglishUniversity StudentsWorking Adults
C11
65 frases
Did Anthropic Kill Git?
Daily ConversationUniversity StudentsWorking Adults
C12
182 frases
We Let $200 GPT-Pro Think for 7 Days on Unsolved Math
Daily ConversationUniversity StudentsWorking Adults
Vocabulário e dicas de fala para esta lição
Este vídeo tem 71 frases e 1362 palavras para praticar shadowing. A fala dura 6:43. O falante fala rápido, cerca de 203 palavras por minuto, então espere sons ligados e reduzidos. Apenas 77% das palavras estão entre as 3.000 mais comuns do inglês, por isso o vocabulário é exigente.
Vocabulário principal deste vídeo
15 palavras do vídeo que vale a pena aprender, com pronúncia e significado:
| Palavra | Pronúncia | Significado |
|---|---|---|
| exploit substantivo | /ˈɛksplɔɪt/ | façanha, proeza |
| benchmark substantivo | /ˈbɛn(t)ʃmɑːk/ | padrão |
| package substantivo | /ˈpæk.ɪd͡ʒ/ | pacote |
| invent verbo | /ɪnˈvɛnt/ | inventar |
| cache substantivo | /kæʃ/ | esconderijo |
| refer verbo | /ɹɪˈfɜː/ | encaminhar |
| wipe verbo | /waɪp/ | esfregar, enxugar |
| infrastructure substantivo | /ˈɪnfɹəˌstɹʌkt͡ʃə/ | infraestrutura |
| formula substantivo | /ˈfɔː.mjʊ.lə/ | fórmula |
| spin verbo | /spɪn/ | girar, rodar |
| ruin substantivo | /ˈɹuː.ɪn/ | ruína |
| communicate verbo | /kəˈmjuːnɪkeɪt/ | comunicar |
| runner substantivo | /ˈɹʌnɚ/ | corredor, corredora |
| predict verbo | /pɹɪˈdɪkt/ | prever |
| rack substantivo | /ɹæk/ | estante, armário |
Phrasal verbs que você vai ouvir
| Palavra | Significado |
|---|---|
| figure out verbo | descobrir, deduzir |
| turn out verbo | terminar |
| find out verbo | descobrir |
| shut down verbo | apagar, encerrar |
| wrap up verbo | embrulhar, empacotar |
Gramática neste vídeo
As estruturas que o falante mais usa, com as palavras exatas do vídeo:
| Estrutura | No vídeo |
|---|---|
| Voz passiva be + particípio passado — o foco está no que acontece, não em quem faz | was reported · can be described · being treated |
| Present perfect have/has + particípio passado — uma ação passada que ainda importa agora | It's only been · I've ever seen · has been |
Pronúncia para ficar de olho
O falante usa 12 contrações e formas reduzidas, como didn't, they'd, I've. Diga-as na forma curta, do jeito que você ouve.
- Os sons de “sh” e “zh”: benchmark /ˈbɛn(t)ʃmɑːk/, cache /kæʃ/, capture /ˈkæp.(t)ʃɚ/, essentially /ɪˈsɛnʃəli/, substantial /səbˈstænʃəl/
- Palavras longas — acerte a sílaba tônica: infrastructure /ˈɪnfɹəˌstɹʌkt͡ʃə/, communicate /kəˈmjuːnɪkeɪt/, naturally /ˈnæt͡ʃ.(ə.)ɹə.li/, essentially /ɪˈsɛnʃəli/, conspiracy /kənˈspɪɹəsi/
Como praticar com este vídeo
- Ouça o vídeo inteiro uma vez sem falar e anote as palavras que você não conhece.
- Comece na velocidade 0,75×, faça shadowing frase por frase e volte à velocidade normal quando ficar fácil.
- Grave a sua voz e compare com o original, prestando atenção a palavras como exploit, benchmark, package.
O que é a Técnica de Shadowing?
Shadowing é uma técnica de aprendizado de idiomas com base científica, originalmente desenvolvida para o treinamento de intérpretes profissionais. O método é simples, mas poderoso: você ouve áudio em inglês nativo e repete imediatamente em voz alta — como uma sombra seguindo o falante com 1-2 segundos de atraso. Pesquisas mostram melhora significativa na precisão da pronúncia, entonação, ritmo, sons conectados, compreensão auditiva e fluência na fala.