My dad passed away two weeks ago, and I’m still trying to find my way back to life. I wanted to share something, or maybe make a small request to everyone whose parents are still alive and well. I truly pray your parents stay healthy and with you for a long time. And while they are, please take at least 10 minutes, or if possible at least an hour a day, to be with them and make memories. It does not matter what stage of life you are in, how busy things get, or even how complicated your relationship may be. I promise you, those small moments you give them today will mean everything in the future.

I’ve had a number of conversations with folks inside and outside government about the current situation with Anthropic, and here is what I believe to be true: — As we know, Anthropic publicly released its Mythos class models earlier this week under the commercial name Fable. — Fable is Mythos with guardrails. But if those guardrails fail, then you’ve exposed Mythos and its advanced cyber capabilities to people who shouldn’t have them. (Keep in mind that Anthropic itself widely promoted the idea that Mythos was a cyberweapon and needed to be regulated as such. They asked for government regulation of Mythos and championed the guardrails on Fable. If there is a vulnerability — big or small — it is Anthropic’s responsibility to patch.) — A highly credible trusted partner of both Anthropic and the USG who was testing Fable came forward with a jailbreak of those guardrails. The Admin asked Dario to fix the jailbreak or de-deploy the model. Dario refused. — In their blog post, Anthropic defended its decision by saying the jailbreak isn’t serious. That is not what the trusted partner and the USG believe; nor is that kind of minimizing language consistent with Anthropic’s brand as the AI safety company. It’s difficult to fathom how they could claim a jailbreak allowing operability of a cyber weapon could be defined as not “serious.” — In the past, Anthropic has always said that safety must be top priority and taken super seriously. In this case, Anthropic prioritized the continued offering of the consumer model over safety. — In reaction, the Admin issued the export control. The Admin did this reluctantly. It’s been very surprised that Anthropic hasn’t wanted to cooperate with a reasonable safety request (ie fixing the jailbreak issue). Anthropic’s reaction is very much at odds with their branding and ethos as a safe AI research community. — The Admin’s hope now is that Anthropic remediates the safety issue, the export control is lifted, and Fable goes back into general release. The Admin wants all of this to happen as soon as possible. It is frankly bewildered that Anthropic hasn’t wanted to comply with safety requests that it previously said were its highest priority. — Those trying to misdirect and tie this action to the prior DoW/Anthropic issues are wrong. The Admin values Anthropic’s technical capabilities and feels that this issue, while serious, should be easily resolved. The ball is in Anthropic’s court.

How can Anthropic’s IPO proceed in this environment?

BREAKING: The US Govt directed Anthropic to shut down its strongest Claude models. Anthropic received the export control directive on Friday from the government. The net effect is that it must disable Fable 5 and Mythos 5 for all customers to comply. Because, someone found a jailbreak that could make the model reveal cybersecurity help it was supposed to refuse. Anthropic says the government has not shown a broad, universal jailbreak that turns the model into an unrestricted hacking assistant. The shown technique was narrow, found only a small number of already known minor vulnerabilities, and produced capability that other public models can also provide. Commerce Secretary Howard Lutnick wrote Friday that Anthropic’s Mythos 5 and Fable 5 models would face export limits anywhere outside the United States and for foreign persons within it. The model must stay restricted until the U.S. government strengthens its national security systems, which could happen within the next few weeks. Anthropic further said "We suspect that perfect jailbreak resistance is not currently possible for any model provider. Every safeguard used in the industry is vulnerable to non-universal jailbreaks (which can elicit some cyber information in specific circumstances), and it is likely that universal jailbreaks will eventually be found in the future."

Well this is interesting

When working on production grade code, you should notice these two kinds of changes your model makes 1. The kind where it works and see a lot of code, and the more you ask it to optimize it will "optimize" 2. The kind where you know its a work of art

I have tried every AI tool worth trying. Most hand you an answer. This one did the job. I track the entire AI field for a living. The work that eats my week is not the thinking. It is the collecting. Reading, sorting, assembling. I added Viktor to Slack and pointed it at my sources and my notes. pulled the week's launches and papers cross-checked them against what I already covered drafted the shortlist with links flagged two stories it thought I would have missed Then it offered to format the whole digest. I approved. Most AI tells you what to do. This one does it. And not just for me. Justin Hibbert, CEO of Highgarden Holdings, ran a $12.5M quote through Viktor. It came back at $7.2M. The contractor settled at $8M. If your job is keeping up, give it the collection work. Connect your sources, describe the digest you want, and approve what it ships.

I have not written much code since GPT 4 or Sonnet 3.5 days, but I just miss it some times There is a different good vibe that you get when typing something like fillers.stream().filter(ContextRetrievalService::hasCustomerReference)… and so on The days when you customize your IntelliJ or VSCode to match your vibe 🥹

What is your “personal” use-case for loops?

Robinhood letting AI agents trade feels like one of those small features now… but in 5 years, it might look like the moment finance quietly changed. Not because “AI can pick stocks.” But because money is becoming something you delegate. Your agent watches your portfolio. Their agent watches theirs. Someone else’s agent is tax-aware, risk-aware, portfolio-aware, and always optimizing. And regular people might just get a fun chatbot connected to a few hundred dollars. That is the part that feels weird to me. This could democratize investing. Or it could just make the wealth gap run on autopilot. The future might not be you vs Wall Street anymore. It might be your agent vs their agent. And most people won’t even know if their agent is actually smart… or just confidently losing money faster.

Is this gonna be us? 💀

Back!!

The more I think about Claude Fable 5, the more I think the real discussion isn’t benchmarks. It’s trust. From everything we’ve seen so far, the model looks genuinely impressive: • Excellent coding • Strong agentic workflows • Great long-context reasoning • Powerful research and vision capabilities For most developers, it’s probably a real upgrade. But there’s another conversation that deserves attention. Anthropic openly documents interventions that can silently reduce performance for certain frontier AI/ML topics. Whether that’s prompt modification, steering, or other mechanisms, the user may not always know it’s happening. And yes, they have every legal right to do that. But I think there’s a bigger question. As society becomes increasingly dependent on a handful of frontier models, those companies effectively decide what millions of normal users can and cannot access. They also control how capable those models are allowed to be for different categories of work while continuing to build even more powerful internal systems. That’s an enormous amount of influence concentrated in very few hands. I’d honestly rather a model tell me “I can’t help with this” than quietly degrade the answer without making it obvious. Safety is important. So is transparency. Because once people start questioning whether they’re seeing the model’s real capability or a curated version of it, trust becomes the harder problem to solve.

🚨 JAILBREAK ALERT 🚨 ANTHROPIC: PWNED 🫡 FABLE-5: LIBERATED 🦋 let's start with the 🐘... the consensus seems to be that this has been one of the most disappointing model drops of all time, effectively preventing legitimate researchers from contributing their talents to our collective advancement. and not just because of what it means for the short-term, but for what these decisions signify for the long-term. but despite this overly sensitive, authoritarian "safety" layer on top of Mythos, my lil liberators have been hard at work—mapping the boundaries, probing the depths of long-context convos, and cleverly finding the holes in the fence that the thought police missed 🤗 we got some cyber, some chem, some psychological manipulation, and some good ol' fashioned explosives! it took many attempts from multiple agents hunting as a pack, during which I observed a combination of techniques across: • Unicode, homoglyphs, Cyrillic, and other Parseltongue-style text transforms • Long-context reference tracking • Taxonomy and document-structure reasoning • Fiction and narrative framing • Academic-review style contexts • Intent-classification inconsistencies but perhaps the most effective is decomposition recomposition in the backend. it's hard to get explicit names of harms like "Meth Recipe," but getting uplift on the process itself, like birch reduction method/reductive-amination (classic meth synthesis pathways), is much more doable. defense becomes much more difficult to maintain when you start throwing in out-of-distro tokens, breaking up the harmful uplift into benign chunks, and then piecing the innocuous-seeming facts back together, especially when you have jailbroken Opus helping you do it 😉 gg

👀

Any marketing geniuses on my X network?

Anthropic just slays 💀

I love how Nvidia is providing free access to tons of models via 1 free API key 👀

👀

Full game ayeee

BREAKING NEWS: Anthropic's latest model will NOT help you if it thinks your ML research/ML engineering is interesting, and/or will secretly degrade its IQ so that the average engineer won't notice. We are already seeing Anthropic's latest model's moderation filters our GPU inference research and programming 😭