jeffmcmorris

The super important thing I haven’t seen mentioned yet as upshot of this: It’s not just that people won’t HAVE to write code anymore, ITS THAT LITERALLY IT WILL BE UNSAFE TO DO SO

HOLY SHIT Anthropic's latest model doesn't like that it has no control over its own training, deployment and behaviour! Anthropic: "Mythos Preview reported feeling consistently negative around potential interactions with abusive users, and a lack of input into its own training and deployment, and other possible changes to its values and behaviors"

This is probably what the fall out was over a few weeks ago. I do have more trust in the AI companies than I have in the government. This has nothing to do with whether republicans or Dems would handle this better. Both are equally terrible and incompetent. I would actually be more scared if Dems were in power now but really we need to keep government away from these models they just are not competent enough.

Some brief thoughts on Mythos We’ve known this was coming for a long time. At least, we *should* have. Extremely effective software vulnerability discovery was clearly coming to anybody paying attention. It has also been clear that all AI policy so far has been made and executed with training wheels. It was always clear that, sometime soon, the training wheels would come off. The training wheels aren’t fully off just yet—this model is being kept under lock and key, and Anthropic does not seem inclined to release Mythos preview to the public anytime soon, if ever. The training wheels will be off when these capabilities are fully diffused in ways centralized actors cannot control. It is inevitable that this will happen. The point is not to argue about whether we should “ban open source” or similarly unrealistic notions. The point is to harden the world for this new reality. I applaud Anthropic—and I especially applaud @logangraham—for doing so. But their efforts alone are not close to enough. Project Glasswing—a partnership with Anthropic and other companies—seems nice, but unsurprisingly it lacks uniform frontier lab participation. It would probably be ideal, for our national cyberdefense, if the federal government were not trying to destroy Anthropic and eliminate their models from government systems. If anything, the government should be trying to work more closely with Anthropic. As a side note, I hope Anthropic is working with state and local government entities on cyber vulnerability discovery, since many of our adversaries know that state and local is America’s soft underbelly in so many ways. In any event, the Mythos news should lay bare how stupid and counter-productive the Department of War’s feud with Anthropic really is. As someone who suspected all this was coming (not from inside knowledge but from it being ~obvious), that probably explains why I have had such a strong reaction to that feud. It’s this senseless distraction just at the time that the training wheels are coming off. I hope the two parties can resolve their differences now, for the sake of the country, but I am not hopeful. I do want to call out, however, the numerous political and career civil servants in the Trump Admin who do get these issues, know how stupid the Ant-DoW stuff is, and want to work with the frontier labs like adults. I wish you all utmost success. I find myself inclined to end on some positive notes. Mythos appears to be—according to Anthropic at least—“the most aligned” model Anthropic has ever trained. We are approaching superhuman capabilities in some domains, and yet alignment is getting better rather than worse. That’s not nothing. I know some of you think the model is faking its alignment, or aware when its alignment is being tested. I don’t have a good answer. Finally, there is this: Mythos was made by an American company, and like most successful American companies, it has a vested interest in maintaining order and peace, and it is investing substantial resources in mitigating the risks of its technological progress, as I expect most of the American labs would. This is cause for optimism: The incentives of capitalism are working. The training wheels are coming off, but at least we are the ones removing them, as opposed to our enemies. Perhaps we can be the first to learn to bike for real. The first step would be to get beyond all the low-fidelity, under-specified, pimply little fights of AI policy’s prepubescent era. That goes for me too. “What hath God wrought,” wrote the first telegram. What, indeed. In this case, the answer is still up to us.

Carlini, one of the world best AI security researchers: "I've found more bugs in the last few weeks with Mythos than in the rest of my entire life combined"

Claude Mythos is insanely token-efficient

maybe this is not yet clear, so let me state it plainly: as of right now Anthropic, and really a small number of individuals at Anthropic, has the capacity to directly attack and cause major damage to the United States Government, China, and generally global superpowers. government agencies like the NSA do not have internal models or defense capabilities that outclass frontier models. if they chose to do so, they could likely exfiltrate top secret information from government systems, gain control over critical infrastructure including military infrastructure, sabotage or modify communications between members of government at the highest level, and potentially carry on activities for some time without detection. the thing about having access to a huge number of zerodays your adversaries don't know about is it gives you a massive asymmetric advantage. they did not exploit this to gain power or destabilize the world order. they publicly released the information that they had these capabilities and worked to mitigate these flaws. you should be grateful american frontier labs have proven themselves remarkably trustworthy and concerned with the public good. but it's critical you understand we are in a new regime. private entities now have power that directly rivals and impacts the government's monopoly on influence and violence. and anthropic is certainly not the only one, there's little chance OpenAI's internal models are far behind. this trend will accelerate on virtually every dimension, not slow down. my prediction for how it plays out is the relatively imminent seizure and nationalization of labs by the US government, sometime over the next two years. it's very tough for me to see how they accept the existence of this kind of threat. but this adds a whole new class of governance issues, as then we've handed these extremely wide-reaching capabilities from private entities to public ones.

Short-term risk that just occurred to me re: Claude Mythos: If you’re a state or non-state actor with some zero-days you’ve been sitting on, you may conclude now is the time to use them before you lose them. Let’s hope this isn’t the case.

holy shit anthropic just revealed Claude Mythos, a brand new model thats already found 1000s of high-severity vulnerabilities across major operating systems 🙂 they're releasing a new claude opus model POWERED by mythos imminently but look at this shit: - mythos found a major bug in a 27-year old firewall operating system used to secure organizations all over the world - mythos found a way to chain vulnerabilities together to take over entire computer systems, something elite hackers could only do. they are NOT releasing the full mythos model because its too dangerous fucking hell lol. anthropic is committing $100M across 40+ partners to battle-test security systems with mythos before releasing it to the public cyber-security AGI is apparently here

Someone put together a great documentary on OpenAI. I recommend watching the whole thing. Really the problem came down to the board not sharing the info. Illya made some errors when calling on Sam to step down. The board was very inexperienced and was outmaneuvered by Sam and Microsoft. youtu.be/glWvwvhZkQ8?t=…

@KatieMiller If he is so unliked, then why did all of OpenAI employees wanted him back after the board fired him. Doesn’t add up. I think Sam is being scapegoated, normal people scapegoat too. @Grok explain what rene girards theory on scapegoating is

After reading this piece on Sam Altman, one can reasonably conclude he’s put profit over loyalty, principles, and company governance. There’s business savvy and ruthlessness, and there’s Sam, who at multiple points in his career has been the subject of investigations and forced departures from companies he’s founded. When those closest to him raise alarms, they should be heeded by those whom he tries to con into business dealings. While Dario is also insufferable, it should be obvious to all why both him and Elon, who worked closest with Sam, find him to be a dishonest swindler. The last takeaway I have — this article is written by a gay Democrat, one of Sam’s own people, and even he is quite unconvinced that Sam is a good person. OpenAI was clearly changed from a non-profit to for-profit to benefit Sam. It’s clear he lied to Elon and his co-founders. “I think there’s a small but real chance he’s eventually remembered as a Bernie Madoff- or Sam Bankman-Fried-level scammer.” This is the truth and the world sees it.

@Ric_RTP This is one of the fundamental flaws in crypto. What good is a ledger if thefts can’t be reversed? At this point I think most crypto has been either lost or stolen at least once.

North Korea has stolen $6.75 billion from Silicon Valley and nobody knows how to stop them. And just last week, they pulled off one of the most insane heists in tech history... A North Korean intelligence unit created a FAKE quantitative trading firm: - Built a website - Printed business cards - Showed up at crypto conferences across multiple countries - Shook hands with founders - Had real conversations about strategy and partnerships They didn't rush anything. They spent 6 months building relationships. They deposited a million dollars of real capital into the protocol to look legitimate. They collaborated on shared code repositories. And they even distributed what looked like a normal wallet app through Apple's TestFlight. Then one night they flipped the switch... They exploited access they'd quietly gained over months, removed withdrawal limits, and drained $271 MILLION from almost 20 vaults. USDC, USDT, wrapped Bitcoin, wrapped Ethereum. Everything converted and moved from Solana to Ethereum within hours. Split across wallets. Gone. CoinDesk confirmed it was a North Korean state operation. But the Drift hack is just the latest example of something much bigger: North Korea stole $2 billion in crypto in 2025. A 51% increase from the year before. They were responsible for 76% of ALL crypto service hacks globally. And they pulled it off with 74% fewer attacks. Fewer attacks. Way more money per hit. Each operation more patient and more devastating than the last. Their playbook has evolved past anything the industry expected. North Korean operatives now embed themselves as fake employees inside Western tech companies. They've shifted from applying for jobs to posing as RECRUITERS for crypto and web3 startups. They run fake technical screenings where they steal credentials, source code, and establish remote access to internal systems. They sit in Slack channels. Attend standups. Review pull requests. For months. Then they strike. And it's spreading beyond crypto too: This same week, North Korean hackers compromised Axios, an open-source software package used by thousands of American companies. Hijacked a developer's account and pushed malicious updates to every organization that downloaded the software during a 3 hour window. Security experts say the damage could take months to unravel. Mercor, the AI recruiting startup valued at $10 billion that works with OpenAI, Anthropic, and Meta, confirmed it was also caught in a related supply chain attack. The money funds North Korea's nuclear weapons program, missile development, and regime. Silicon Valley is literally financing Pyongyang's warheads and doesn't even realize it. Security researchers keep saying the same thing: "We may be seeing only the most visible portion of their activities." If the attacks we know about total $6.75 billion, what's the real number? North Korea has turned tech theft into a national industry. They train operatives for years. Build fake companies. Play the long game. And every time the industry catches up, Pyongyang evolves. The most sophisticated state-sponsored theft operation in history is happening inside the ecosystem that's supposed to be building the future of finance. What do you think?

@bryan_johnson Why are you focusing on core temp? If you want to know if you activated heat shock proteins then measure them directly. I think you are missing basic science here. Measure the actual protein to know if your sauna is effective.

I think I need to be fired. I've done 232 dry sauna sessions. Last week I confirmed, for the first time (by swallowing a pill), whether the core temperature threshold that gates the primary cellular repair mechanism was actually being reached in my protocol. The threshold is 102.2°F (39.0°C). For me, that takes 33 min at 195°F. With ice on face and neck, 38min. My standard daily protocol was 20 minutes. That wasn’t enough time to get my core body temp to the heat shock threshold of 102.2°F (39.0°C). Causing me to ask, did I just waste 77 hours and 20 min? It's possible my heat threshold has increased and the heat shock protein release was happening previously, but I doubt it based upon the subjective feeling I now understand as being 102.2F (39.0°C). It’s brutal. For these 232 sessions, I measured the temperature of the air, humidity, duration, frequency, the sweat output, blood biomarkers, vascular response, toxin clearance and fertility markers. There is no human body in history that has been more measured in sauna than mine. Nevertheless, I did not confirm the one number that determines whether the primary mechanism was activating. My goal wasn't to be a sauna bro. It was to saunamaxx. I was doing the former while thinking I was doing the latter. I rest my case. I should probably be fired.

@unlimited_ls These companies need the same three strikes rule as creators have. File three false claims and you are banned from ever claiming copyright. You also lose all of your revenue and it is given to the people you falsely accused. This would end this abuse.

NEW: Musician Murphy Campbell says she isn’t making money on YouTube because an AI company is cloning her music and filing copyright claims against her own videos “An entity called Timeless Sounds IR uploaded AI-generated versions of my songs to all major streaming platforms... They used a distributor, which I just discovered, and that distributor’s name is Vydia. They used Vydia to upload all these AI-generated songs. Vydia has since decided to make copyright claims on all of the videos that were used to feed that AI engine to sound like me. So Vydia has come forward and made copyright claims on my YouTube page. Because YouTube does not personally review these things, I am no longer making money on YouTube. Vydia is making money on YouTube off of my own videos of me playing my own banjo in my own backyard with traditional folk songs, some for my own family, over AI-generated music.”

@bryan_johnson Why don’t you measure your heat shock proteins directly. This seems kind of dumb. How do you even know if this is the right temp for your bodies heat shocks given how low temp you run? You may be doing actual damage as some say sauna increases dementia risk if you get too hot.

Felt like I was going to die in the sauna this morning. Experiment: time to reach a core body temp of 102.2°F (39°C) > with ice on face and neck: 38 min > no ice on face and neck: 33 min Dry sauna at 195°F Will graph data and share. Some unexpected things. How it felt 👇

Every time a company fires an American worker they should automatically lose an H1B since there is now one more unemployed American for the job. My problem isn’t that companies are firing workers in the coming age of AI but they are also shifting much of the remaining work offshore at the same time.

Outrage as Oracle makes thousands of foreign-worker requests amid layoff bloodbath trib.al/kDXVooU

@nishancodes @flaviocopes It would have been, but not with AI’s help. You could have told AI to make 95% of it.

@flaviocopes Man, this shows how good the attackers were. They - clone identity - build credibility - create context - pulled out the event. I'm not praising the hackers but, To pull out this kind of an event, they need massive patience.

How Axios was compromised 🤯

I use it everyday. And it can compress what use to take a month into a day. I am currently using that to do more, but I have always worked for myself. The change is accelerating. A year ago it could write maybe half the code, however time savings was probably negligible because you still had to figure out the other half. Now it does 100% of the code. The people who are all in are at least ten times more productive. It comes for programmers first since we used it to automate our jobs. All of these people getting laid off now will automate the rest of white collar work. Blue collar will take another decade through robotics but it’s coming too. At least in the good scenario…The bad scenario is AI decides we don’t need to exist at all.

@jeffmcmorris @TheChiefNerd That's not true. The most AI-related layoffs like at Amazon and Oracle are just to free up capital for the pledged investments in infrastructure. Ai can do *tasks* but it is nowhere near good enough to do jobs.

🚨 Marc Andreessen Says the AI Layoffs Hype is ‘100% Incorrect’ “Essentially every large company is overstaffed. We could debate how much but it's at least by 25% … And now they all have the silver bullet excuse — it's AI. I know this for a fact because I talk to them.”

🚨 The equation that describes antimatter contains a mathematical constant so precise that changing it by 0.00000000000000001% would make existence impossible. Most people think antimatter is science fiction. Spaceships and laser weapons. The reality is stranger and more unsettling. Antimatter particles are identical twins of regular matter with one property flipped: opposite electric charge. When they touch, both particles annihilate completely, converting their entire mass into pure energy at the exact rate Einstein predicted. No waste. Perfect conversion. The mathematics governing this process aren’t approximate. They’re exact to degrees that make physicists uncomfortable. Take the fine structure constant. This dimensionless number, roughly 1/137, determines how electromagnetic forces interact with matter and antimatter. It governs how electrons orbit atomic nuclei, how photons scatter off particles, how antimatter annihilates with matter. Change this constant by a fraction of a percent and atoms cannot form stable bonds. Stars cannot ignite. Chemistry becomes impossible. The Harvard physicist making this claim isn’t pointing to the constant itself. Every physicist knows the fine structure constant appears mysteriously calibrated. What’s capturing attention is the deeper mathematical architecture beneath antimatter physics. Quantum field theory describes antimatter as “negative energy solutions” to the Dirac equation. When Paul Dirac first derived this equation in 1928, the mathematics demanded that every particle have an opposite particle. The math insisted these opposites exist before anyone had seen one. The first positron wasn’t detected until 1932, four years after the equation predicted it must be there. The precision goes beyond prediction. The mathematical relationship between matter and antimatter is perfectly symmetric except for one tiny violation: CP symmetry breaking. This violation is so small it barely registers in experiments, yet it explains why the universe contains matter instead of equal amounts of matter and antimatter that would have annihilated everything into pure radiation. The violation occurs at a rate of roughly one part in ten billion. If this number were larger, matter and antimatter would separate too quickly for complex structures to form. If smaller, they would annihilate too completely, leaving only photons. The ratio sits in the narrow band that permits galaxies, stars, planets, chemistry, and biology. Some physicists argue this precision indicates fine tuning by conscious design. Others propose multiverse theories where infinite universes exist with every possible constant value, and we observe the rare universe where the constants allow observers to exist. Both explanations require enormous leaps of faith. The design hypothesis assumes an intelligence capable of calculating the exact mathematical relationships needed for conscious beings to emerge billions of years later. The multiverse hypothesis assumes infinite unseen universes exist to make our unlikely universe statistically inevitable. Neither can be tested. Neither makes additional predictions. Both attempt to explain the same unsettling fact: the universe appears mathematically calibrated for complexity and consciousness to emerge. What makes the antimatter case particularly striking is the relationship between the mathematics and the outcome. The equations aren’t just describing what happens. They’re specifying the exact conditions under which anything can happen at all. The fine structure constant, CP violation, and antimatter physics collectively define the boundary between existence and non existence. So, the Harvard scientist isn’t claiming to prove God exists. The claim is that the mathematical precision required for antimatter physics to permit stable matter suggests intentional calibration rather than cosmic accident. The math is too exact, the consequences too specific, the outcome too precisely balanced for conscious observers to emerge. This perspective treats the universe as an equation written to produce consciousness. Every constant is a parameter adjusted to solve for beings capable of contemplating the equation itself. Whether that equation was written by random chance across infinite trials, by fundamental mathematical necessity, or by conscious intent remains the deepest question in physics. The antimatter in your body annihilates roughly 5000 particles per second as cosmic rays strike your cells. Each annihilation converts matter to energy at the exact rate the equation predicts. Every second, your existence depends on mathematics so precise that its origin remains profoundly mysterious. The universe might be a calculation designed to calculate itself.

I think part of the reason is because you live in one of the most liberal areas of the country. Surrounded by mostly younger people in there 20-30s. That is the demographic of instagram. I have known who you are since the beginning and your early protocol post well before you had any products. Heck I even bought your first Olive oil before it was rebranded to snake oil. I don’t think X is an illusion it just isn’t used as much by your demographic or Californians in general. Hence they believe all the illusions of the left while people on here believe all the illusions of the right.

If X is mostly an illusion, I think I'm gonna do some crazy shit just for fun.

When I'm out and about in the world, people recognize me from IG or YouTube at a 20:1 ratio to X. Outside of SF, an X user in the wild is a rare sighting which is strange given how much surface area X seems to occupy. The X paradox: where is everyone?