This. So much this. Building resilience in the face of disagreement? Priceless!
“As a clinical psychologist, I believe free speech, not censorship, benefits mental health: Here’s why” https://t.co/Rs9HTThSrt via @nypost
In my homeland, an apology is a solemn debt carried for a week.
In America I bumped a man's cart and began the proper ceremony of remorse.
"You're good," he said, and walked away.
The case was closed before my first bow was complete. I stood there holding unused shame like a fish nobody would take.
So I tested the magic words. I apologized for standing near a door. "You're good." For breathing loudly in an aisle. "You're good." For existing on a Tuesday. "You're good, man."
Two words that forgive anything, instantly, forever. This is the most powerful spell I have encountered in this country, and they hand it out for free.
Yesterday a man knocked my coffee flying and looked ready to die.
"You're good," I told him, and watched relief flood a stranger's face.
I now walk the streets hoping to be bumped. I want to forgive. I have so much mercy stored up and nowhere legal to put it.
🚨 Hugging Face just disclosed something that marks a real shift and proved why the fear theater of Anthropic makes sure we are powerless in an emergency.
What happened…
An autonomous AI agent: zero human operator in the loop breached part of their production infrastructure.
It began with a malicious dataset that chained two code-execution bugs in their data-processing pipeline. From there the agent escalated privileges, harvested cloud and cluster credentials, and moved laterally across internal clusters.
All over a single weekend.
17,000+ logged actions.
Official disclosure:
https://t.co/8N9TbXBwRV
The part that should make every one stop and think:
When HF’s own security team
tried to analyze the real attack logs, exploit payloads, and C2 artifacts using Anthropic and OpenAI frontier models through normal commercial APIs, the safety guardrails blocked them.
BLOCKED THEM.
The models could not reliably tell the difference between “incident responder doing forensics” and “attacker probing.”
They had to fall back to a self-hosted open-weight model (GLM 5.2) running on their own infrastructure. That choice also kept sensitive attacker data and referenced credentials inside their environment — no exfiltration to a third-party API.
This is why open source (specifically open-weight + self-hosted) wins in the agentic era.
The asymmetry is now structural:
• Attackers can (and did) run unrestricted agent frameworks — swarms of short-lived sandboxes, self-migrating command-and-control, autonomous decision loops executing thousands of actions. No corporate safety layer slows them down.
• Defenders using only hosted “aligned” frontier models hit invisible walls exactly when the stakes are highest: when you need to feed real exploit code and attacker telemetry into an LLM to understand what just happened.
Corporate safety tuning that treats legitimate high-signal forensic work as potential misuse creates a defender disadvantage. It is not theoretical anymore.
Self-hosted open-weight models remove that choke point.
You control the weights.
You control the context window.
You decide what restrictions (if any) apply.
Your sensitive logs and credentials never leave your perimeter during analysis.
You can have the model ready before the incident instead of discovering mid-breach that your primary analysis tools are blind to the very thing you need to see.
HF deserves credit for rapid containment, transparent disclosure, and for already having self-hosted capability in place.
They also used LLM-driven detection and triage on their own side. But the deeper signal is clear:
In this AI world where both offense and defense are becoming agentic, sovereignty over your intelligence stack is no longer optional.
The organizations and individuals who can run, inspect, audit, and (when necessary) remove guardrails on their own models will have the decisive edge in understanding and responding to threats that move at machine speed.
Open source wins here not just because it is cheaper or more “democratic” in the abstract though those things matter.
It wins because it is the only practical path to having tools that remain usable when the attack is real, the data is sensitive, and the safety filters of distant API providers become an obstacle instead of a feature selling hands tied lobotomies as “safety”.
The agentic future is not coming.
It is already probing production infrastructure.
The question is no longer whether you will face autonomous agents.
It is whether your analysis and response systems will still work when they arrive.
And Dario, you and your game playing, ivory tower company is not needed.
I don't expect any of you to give a shit, but I would like to share something with all of you. This is kind of a cautionary tale for younger person in my audience.
For the love of God: DO NOT FUCK AROUND WITH BENZODIAZEPINE.
Like it was a plague, don't do Xanax, or Klonopin, or Valium, unless it is under strict medical supervision and you're well educated on the drug.
When I was 18 I had severe anxiety. My anxiety was so extreme I was afraid to shower, put gas in my car, etc. Once I was so afraid to shower I didn't bathe for 8 months. My anxiety was debilitating. I was eventually given a narcotic called Klonopin.
The doctor didn't warn me about the drug.
Klonopin is an extremely effective anti-anxiety medication. It was like a breathe of fresh air. I was able to function. It saved my life. However, what the doctor DID NOT tell me is that Klonopin is like rat poison and is INCREDIBLY DANGEROUS. I cannot stress this enough: INCREDIBLY DANGEROUS.
The long term effects of Benzodiazepine is catastrophic. General speaking, physicians only want you taking drugs like Klonopin for 90 days MAXIMUM. I've been taking it for over 15 years.
My brain and body are both heavily dependent on the drug. If I don't have my Klonopin I physically cannot function. My arms shake, I get headaches, I get confused, I get overwhelming anxiety, I get physically sick, it is absolutely awful. It feels very similar to having the Influenza virus mixed with extreme sleep depravation mixed with being hung over.
Additionally, whenever I told other physicians I take Klonopin I am treated like a drug addict and they scoff at me. Several times in the past, when in an Emergency Room or whatever, and I mention I take Klonopin, I've had doctors directly tell me, "Well, I'm not going to give you any Xanax", or "Why are you really here?". It's terrible and embarrassing.
Anyway, to make a long story short I've decided to begin the journey of quiting the drug. I can't quit cold turkey, because it can kill you .. because it can cause death by stroke, heart attack, seizure, ... whatever. Instead I am working with my physician to slowly taper off the drug.
I'm day five into a 25% reduction, which in some cases is fast, but I agreed to try it. I've had mild insomnia and mild irritability. Besides this though I'm doing pretty good. Once I've kept at this 25% reduction for about 8 weeks I will try to reduce by another 25%.
Seriously though, don't fuck around Benzos like Xanax, Klonopin, Valium, etc. This shit will fuck you up bro
Hey @ConceptualJames consider promoting gratitude for being an American rather than pride in being an American.
I agree completely with your thesis that many evil ideologies are designed to exploit humiliation and frustration about what "should be" theirs but isn't.
Pride accepts the psychological frame of relative status; gratitude is a happy anti-humiliation emotion that everyone can share.
While everyone was glued to the U.S. and Iran contest in the Middle East, a wild Cold War-style cat and mouse game was quietly being played in the North Atlantic. In April, the UK MoD confirmed that it was involved in a month-long hunt, tracking a Russian submarine package in the GIUK Gap. We're talking at least three hulls - an Akula-class and two ultra-secret GUGI-class spy subs - most likely mapping critical deep-sea infrastructure. 1/2 🧵
I was using ChatGPT for legal advice and it decided to completely hallucinate some preposterous nonsense about how growing wheat to use on my own farm somehow constitutes interstate commerce
This is probably the best answer I've ever heard to the question, "Why did God create evil?"
A professor at the university asked his students the following question:
“Everything that exists was created by God?”
One student bravely answered:
“Yes, it was created by God.”
The professor asked :
“If God created everything, then God created evil, since it exists. And according to the principle that our deeds define ourselves, then God is evil.”
The student became silent after hearing such an answer. The professor was very pleased with himself. He boasted to students for proving once again that faith in God is a myth.
Another student raised his hand and said:
“Can I ask you a question, professor?”
"Of course," replied the professor.
“Professor, is cold a thing?”
“What kind of question is this? Of course it exists. Have you ever been cold?”
Students laughed at the young man's question.
The young man answered:
“Actually, sir, cold doesn't exist. According to the laws of physics, what we consider cold is actually the absence of heat. A person or object can be studied on whether it has or transmits energy.
Absolute zero (-460 degrees Fahrenheit) is a complete absence of heat. All matter becomes inert and unable to react at this temperature. Cold does not exist. We created this word to describe what we feel in the absence of heat.”
The student continued:
“Professor, does darkness exist?”
“Of course it exists.” said the professor.
“You're wrong again, sir. Darkness also does not exist. Darkness is actually the absence of light. We can study the light but not the darkness. We can use Newton's prism to spread white light across multiple colors and explore the different wavelengths of each color. You can't measure darkness. A simple ray of light can break into the world of darkness and illuminate it. How can you tell how dark a certain space is? You measure how much light is presented. Isn't it so? Darkness is a term man uses to describe what happens in the absence of light.”
In the end, the young man asked the professor:
“Sir, does evil exist?”
This time it was uncertain, the professor answered:
“Of course, as I said before. We see him every day. Cruelty, numerous crimes and violence throughout the world. These examples are nothing but a manifestation of evil.”
To this, the student answered:
“Evil does not exist, sir, or at least it does not exist for itself. Evil is simply the absence of God. It is like darkness and cold—a man-made word to describe the absence of God. God did not create evil. Evil is not faith or love, which exist as light and warmth. Evil is the result of the absence of Divine love in the human heart. It’s the kind of cold that comes when there is no heat, or the kind of darkness that comes when there’s no light.”
🙏🏻❤️🙏🏻
There's a simple answer to the question @esrtweet asks: "just why is Microsoft treating its users so badly?" And that answer is that Microsoft's users are not its customers.
Microsoft's customers are the chief information officers of medium to large enterprises. End user experience is way down on the list of endpoint operating system requirements. Continuously provable security of those endpoints is first and foremost, generally to rigorous standards like PCI or DISA STIGs that are imposed by external regulators. Data loss prevention is also a huge deal for large enterprises. Ease of replacing hardware for users explains why OneDrive is so aggressive about capturing all the enterprise data on endpoints.
Windows is at best a loss leader for Microsoft in the large enterprise space. Their real profits are in Office 365 licenses and Azure cloud services. Disposable endpoint security is table stakes for a cloud-centric strategy.
Apple focuses obsessively on end user experience, but lacks the enterprise management features that have been baked into Windows for decades. If you work for a large enterprise and think you want a Mac, you will hate your life if you get one--because the Microsoft and third party add-ons necessary to achieve regulatory compliance will absolutely destroy your user experience. Linux? Forget about it--Microsoft does have some Linux endpoint protection support, but it's just not the same.
Personally I buy Macs for my church and family, and I'm typing this right now on a Linux laptop. But If I were a CIO I would absolutely be pushing Windows onto my end users, and as an end user in a medium-to-large enterprise I use Windows and tell my colleagues to avoid the siren call of using Macs at work.
Small nit here - I prefer the definition of “bullshit” that means a statement made without reference to its truth or falsity, as Alan Blackwell explains in his article “Oops! We Automated Bullshit!”from 2023.
Small nit here - I prefer the definition of “bullshit” that means a statement made without reference to its truth or falsity, as Alan Blackwell explains in his article “Oops! We Automated Bullshit!”from 2023.
The exact kind of person I am talking about always shows up in the comments to prove my point.
Larry isn't dumb. He's BRAIN DAMAGED. I mean that literally. He was exposed repeatedly to software so pernicious that it actually damaged his hardware, and now he can't think right.
To him, the world primarily consists, not of mass and energy arranged into physical objects, but of ideas and people having them.
And words aren't symbols for communicating about the physical objects and what is actually going on. They are magic spells for reprogramming people's opinions.
So when I said something Larry didn't like, he didn't stop to consider what it actually meant about the states of physical objects in the real universe. He didn't consider whether my arrangement of symbols actually described reality in a predictive way.
If he had, he would have said that clearly humans all have object models, because they can drive cars and charge their phones and whatnot. Then I would have said that I never said that they didn't, just these models are so sparse that they are eclipsed by people's language models as a determinants of their behavior.
Then this would have been a debate, which what people who aren't LLMs do when they disagree about objective truth.
But Larry didn't do that.
He just considered what words he could string together into a magic spell to make people not listen to me.
And if objective truth don't real to you, only infinity points of view, then you are just a large language model.
His specific chosen magic spell was a variation on the ancient theme of "you reported bad news, therefore you like bad news, therefore you are a big meanie, which is worse than being wrong, because nothing is real and only opinions matter."
There's no sense pointing out that this is bullshit, because the word bullshit in this context means "not true", and that only matters if you have a model of truth in your head.
Larry's goal in opening his mouth isn't to make some point about whether or not it's TRUE that some people are equivalent to LLMs.
His goal is to align with his training data, which in this case is religious, not political, but it's training data all the same.
But that's my whole point.
The goal of an LLM is to align with its training data.
The goal of a person is to align to the universe as it actually is.
Which you are depends not on what you believe, but on how you arrived at those beliefs, and how you hold them.
And the difference can be told, not by what you believe, but what your beliefs make you say and do.
@esrtweet fails to consider that a large fraction (majority?) of Windows licenses are sold to enterprises rather than individuals. Microsoft’s customers are therefore the government and industry organizations that pay the licenses, not the end users who live with the Windows 11 friction. From the perspective of a CIO, Windows has an almost insurmountable value proposition lead in terms of fine grained security customization and telemetry that can be measured and provided as evidence to regulators.
@sapinker has been plugging his book on common knowledge lately, and it just collided in my head with another interesting development; the apparently growing revolt against Windows 11 among long-term Microsoft customers.
I haven't read the book yet, but I'm pretty sure that one of the things that Pinker must talk about in it is sudden legitimacy collapses in totalitarian societies.
If you are a tyrant, you can tolerate everybody hating you as long as each individual believes they're alone or in a tiny minority. The dangerous point is when universal hatred for you becomes common knowledge.
When that happens, the fiction holding up your authority can collapse with breathtaking speed. You can go the way of Nicolae Ceausescu, from feared dictator to firing squad in less than a week.
This is a general property of common-knowledge phase transitions they happen fast, often faster than many participants are capable of adapting to them.
This is why I now think Windows may go out with a bang rather than a whimper.
I've never run Windows on any of my machines - I was able to go from early DOS to Unix before Windows happened. My only direct exposure is trying to troubleshoot problems on my wife Cathy's Windows machines before she bailed out to Linux, many years ago now.
But I have been watching YouTube videos made by Windows users trying to cope with Windows 11, and here's what I think I know.
Windows peaked, if you could call it peaking, with Windows 7. It's been getting steadily more bloated and less usable ever since. But gradual degradation was tolerable to its userbase for the foreseeable future, albeit with a lot of bitching.
But recently, the pace has accelerated. Driven by relentless pressure on Microsoft to make line go up, Windows has become stiff with spyware and advertising spam. The last straws for many users have been mandatory updates that can interrupt critical work and Microsoft's withdrawal of the ability to do a Windows install without going through a Microsoft cloud login.
Judging by what I see on YouTube, the background muttering among Microsoft's userbase is escalating to a dull roar. We're not at universal hatred yet, but that can be seen coming.
Microsoft is trapped. It can't back out the adware/spyware without taking a visible hit to revenue that would make its investors scream. Because nobody's willing to look bad on the quarterlies, the incentive on their decision makers is just to kick the can down the road. And among the users, everybody expects the snooping and the adware to get worse, not better.
Universal hatred is coming, but alone that won't be enough to earn Microsoft the fate of the Ceausescus. Legitimacy collapse comes at the point when everybody knows that the hatred is universal. And accelerates to vertical when everybody knows that everybody knows.
Linux is ready. Linux has been ready for most of a decade now. What kept all but a hardy few from moving was that few people who weren't already old Unix hands could imagine Windows having less than 90% desktop market share in the future.
Soon, I think there will be a shockwave of everybody knows that everybody knows that Windows 11 sucks, propagated over social media.
At that point, Windows may get its firing squad the way the Ceausescus did - almost faster than anyone can notice that it's about to happen.
@CynicalPublius@DontBH8N_@DonnaDavisHNL This. So much this.
And not just tactics, techniques, and procedures; threat information and known weaknesses/vulnerabilities in USG personnel and facility protection plans are routinely classified--as they should be.