Pinned Tweet
Love these books and here is why:
* Jeff's book is a gem, the "alien notation" and the challenge is a fun start and then it goes on into the practical stuff.
* Mitnick's book is legendary as it taps into the most frequent point of failure in cybersecurity: humans.
* Bluefox is a masterpiece written by a giant in ARM security, something to be consulted carefully and with attention. Absolutely love it.
* Windows Internals is simply a must have for everyone wishing to understand the intricacies of the OS. Much love for Pavel Yosifovich (also my teacher at TrainSec).
What are the books that you're reading now and leaving a mark on your learning?
3
670
Where this paper refutes or proves that Chomsky is wrong?
Translation symmetry in data forcing Fourier like representations is interesting. But in no way it addresses Chomsky's arguments:
* Poverty of the stimulus;
* Statistical induction isn't human linguistic competence;
* Next token prediction vs. Structured knowledge.
That's the problem of diagonally reading a paper. 🫤
Chomsky was wrong.
Researchers from UC Berkeley and Google DeepMind cracked one of the biggest mysteries in AI.
For years, scientists have been finding impossible geometric shapes hidden inside the brains of LLMs.
Calendar months arrange themselves into literal circles.
Historical years bend into smooth, rippled timelines.
Cities organize themselves into actual 2D maps of latitude and longitude.
Everyone assumed this proved massive transformers were spontaneously inventing complex "world models" out of thin air.
They were wrong.
This paper proves you don't need a massive LLM for any of this to happen.
Even basic, decade-old word embeddings like word2vec produce the exact same circles, timelines, and maps.
The geometry isn't coming from the AI's architecture.
It’s coming from a hidden mathematical property inside human language itself.
Researchers call it translation symmetry.
In natural text, how often any two months appear near each other depends almost entirely on the time gap between them.
January and February share the same statistical relationship as July and August.
When a neural network trains on data with that kind of symmetry, the math forces it to learn Fourier representations, literal sine and cosine waves.
The circles? Those are the long waves.
The strange "ripples" researchers found on AI timelines? Just higher harmonics.
Here is the craziest part.
Rresearchers tried to break it.
They deleted every single sentence in the training data where two months appeared together. No "January" next to "February." No "March" next to "April."
And the circle formed anyway.
Why? Because the geometry is a collective phenomenon across the entire language.
Even if two months never touch in a sentence, they are both pulled into orbit by thousands of other words tied to the same seasons, weather, and events.
The rest of human language forces the shape into existence.
18
༺༒ Луди зец ༒༻ retweeted
Pra variar, 3blue1brown com mais uma explicação super elegante de como um modelo de linguagem (em particular uma LLM) funciona como um modelo de compressão de informação, perto do limite teórico da probabilidade média de informação de Claude Shannon. Estudantes de Ciência da Computação, assistam está série
youtu.be/qbsb9DpdZ6s?si=aNYP…
Finally arrived! 😍 Going to dig It.
Co-Authored by the super badass @sherrod_im
༺༒ Луди зец ༒༻ retweeted
We get it. OAI/Ant want a regulatory moat.
LLMs are just probabilistic auto complete. All they do is answer the question of “given a sequence of tokens, what is the next most likely token?”
So please explain to me how we’re getting from that to “33% chance it overthrows the govt in 12 months” for anything other than getting it regulated to make it harder for competitors to enter the market.
We live in a yap epidemic.
༺༒ Луди зец ༒༻ retweeted
A lot of people are confused about the conflicting information coming from OpenAI about their plans for AI safety.
This point by @KelseyTuoc does a good job clarifying what’s going on.
I don't have a reputation to spare so I can speak why this AI hype doesn't resonate with me:
I view models as a collection of tools of varying quality that are extremely good at fulfilling contracts. The deeper issue is the assumption that goes underneath a lot of the hype (ASI, AGI..whatever): that better token prediction is evidence for the Computational Theory of Mind (it isn't). The more immediate damage I see is people outsourcing their own reasoning capabilities to models and then slowly but surely losing whatever critical thinking they have.
(running for cover)😱
I feel people have stopped sharing their real opinions on here, and instead everyone is participating in some informal prediction market where the currency is their reputation. Especially on AI.
It sucks. I want to know what people really think because I want to be exposed to alternative perspectives. But in this degenerate prediction game everyone is playing, there's no room for that. If you post something that diverges from the mainstream, you're basically gambling your entire credibility.
I find myself moderating my own posts, even. I don't want to say something too strongly and then have it thrown in my face. So my more unique opinions, the ones where I don't think what everyone else thinks, I don't share.
I'm tired of it. I don't learn anything on here because everyone is just like "ASI by 2028" because someone else said it and they don't want to be seen as an AI skeptic. Please stop. Stop treating every opinion as a bet you can win against someone. All it's doing is reducing the quality of discourse by forcing everyone to shut up and conform.
1
2
122
༺༒ Луди зец ༒༻ retweeted
Enquanto deputados do PSOL do alto do seus gabinetes promovem antissemitismo nas redes sociais contra turistas israelenses no Brasil, o povo de Itacaré, na Bahia, se manifesta em solidariedades aos turistas judeus.
PSOL e antissemitismo caminham lado a lado.
༺༒ Луди зец ༒༻ retweeted
November Black Friday on trainsec!
All individual courses are 40% off!
trainsec.net/cybersecurity-c…
༺༒ Луди зец ༒༻ retweeted
If you manage Active Directory today and are considering your career options, and you’re not retiring soon, here’s my advice.
Get a job as an internal security resource. A generalist role. Find a company and a team you really jive with.
Your super power should be that you actually understand IT.
Networking. Windows, Linux, Mac. Infrastructure. Backups. Hypervisors. Firewalls. Technical Support.
Having a strong IT background is a super valuable skill when transitions to cybersecurity.
🚨🚨🚨
HEADS UP re: RAILWAY INCIDENT: Josh @joshmanders needs some help with incident response so I am posting this on his behalf and with his permission asking for those of you in my orbit who are qualified, and many of you are, to reach out to him with serious intent about helping him with the recent Railway incident.
Triage.
Log analysis.
Correlation.
Josh will give you the specifics.
And no trolling or I'm going to lose my shit on you. This is a serious tweet and inquiry.
Make me proud.
P. S. Retweet far and wide please.
80
🎯🎯🎯
Yes we should talk about this so everybody is very clear on what it is you have said and what I've said and what others have said here.
This behavior is illegal, an actual crime was committed, and the people who ran these agents are 100% responsible and accountable for what the agent did.
STOP RUNNING AGENTS WITH FULL PRIVS AND WITHOUT OVERSIGHT.
AND DON'T HIRE IDIOTS WHO PORTSCAN YOUR COMPETITORS AND PASS IT OFF AS TESTING.
You're not allowed to do that!
66
༺༒ Луди зец ༒༻ retweeted
Replying to @ZackKorman
A portion of security complains about being heard and another portion are elitist. That manager, director, GRC finding gets you headcount and tools. All that takes influence. It might be harder to influence an entire culture and navigate political barriers than find a vuln.
Outside the bubble no one gives a shit about security. If they did we wouldn’t need sanctioned security awareness programs that everyone hates. We might not have mountains of IT debt. Stay humble.
༺༒ Луди зец ༒༻ retweeted
We have a golden opportunity here to do the funniest thing and respond to everything Sam writes on the internet with pictures of our RAM. Like take a selfie with it. And I mean like everything this dude writes needs to just be filled with pictures of RAM. Just flood his feed with it!
OpenAI CEO Sam Altman says it’s “quite terrible” that people have so much DRAM sitting in their own PCs instead of being used in data centers.
“Everyone have a bunch of the world’s limited supply of DRAM sitting in their own computer. And be used 5% of the time,”
He argues that cloud data centers should make better use of limited memory supplies instead of people running AI models locally.
😂😂😂😂😂😂. Todo mundo gosta de se sentir "único" e "especial", acreditar que tem um superpoder ou algo que faça se destacar dos demais.
As armadilhas : o pedantismo, o efeito Dunning-Kruger, a indução vulgar e o perigo de ficar só arrotando grandeza no seco, por puro medo de mergulhar e se molhar.
superdotação é uma desgraça. Existe uma discriminação absurda na vida adulta quando a maioria tá satisfeito em ser pouca merda e a gente que é superdotado sempre tá lá em cima e não aceita esse sistema. Daí começa o boicote.
queria mesmo era ser burra q nem uma porta
54





























