Who likes popcorn - youtu.be/Gwg5ey6236o

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.

𝗧𝗵𝗲 𝗽𝗮𝘁𝗵 𝘁𝗼 𝗴𝗼𝗼𝗱 𝗱𝗲𝗰𝗶𝘀𝗶𝗼𝗻𝘀 𝗶𝘀 𝗽𝗮𝘃𝗲𝗱 𝘄𝗶𝘁𝗵 𝗱𝗶𝘀𝗮𝗴𝗿𝗲𝗲𝗺𝗲𝗻𝘁. 𝗠𝘆 𝗵𝗼𝗽𝗲 𝗶𝘀 𝘆𝗼𝘂'𝗹𝗹 𝗱𝗶𝘀𝗮𝗴𝗿𝗲𝗲 𝘄𝗶𝘁𝗵 𝗺𝗲 𝗼𝗻 𝘁𝗵𝗮𝘁. A decision presented without genuine alternatives: is that truly a decision, or merely an ultimatum? Especially when time is short and those affected by the outcome are excluded from consideration. This is why Peter Drucker stated: "𝘛𝘩𝘦 𝘧𝘪𝘳𝘴𝘵 𝘳𝘶𝘭𝘦 𝘪𝘯 𝘥𝘦𝘤𝘪𝘴𝘪𝘰𝘯-𝘮𝘢𝘬𝘪𝘯𝘨 𝘪𝘴 𝘵𝘩𝘢𝘵 𝘰𝘯𝘦 𝘥𝘰𝘦𝘴 𝘯𝘰𝘵 𝘮𝘢𝘬𝘦 𝘢 𝘥𝘦𝘤𝘪𝘴𝘪𝘰𝘯 𝘶𝘯𝘭𝘦𝘴𝘴 𝘵𝘩𝘦𝘳𝘦 𝘪𝘴 𝘥𝘪𝘴𝘢𝘨𝘳𝘦𝘦𝘮𝘦𝘯𝘵. 𝘋𝘦𝘤𝘪𝘴𝘪𝘰𝘯𝘴 𝘰𝘧 𝘵𝘩𝘦 𝘬𝘪𝘯𝘥 𝘵𝘩𝘦 𝘦𝘹𝘦𝘤𝘶𝘵𝘪𝘷𝘦 𝘩𝘢𝘴 𝘵𝘰 𝘮𝘢𝘬𝘦 𝘢𝘳𝘦 𝘯𝘰𝘵 𝘮𝘢𝘥𝘦 𝘸𝘦𝘭𝘭 𝘣𝘺 𝘢𝘤𝘤𝘭𝘢𝘮𝘢𝘵𝘪𝘰𝘯. 𝘛𝘩𝘦𝘺 𝘢𝘳𝘦 𝘮𝘢𝘥𝘦 𝘸𝘦𝘭𝘭 𝘰𝘯𝘭𝘺 𝘪𝘧 𝘣𝘢𝘴𝘦𝘥 𝘰𝘯 𝘵𝘩𝘦 𝘤𝘭𝘢𝘴𝘩 𝘰𝘧 𝘤𝘰𝘯𝘧𝘭𝘪𝘤𝘵𝘪𝘯𝘨 𝘷𝘪𝘦𝘸𝘴..." I don't shy away from agreement, nor is disagreement inherently difficult. The real challenge lies in how we disagree well. I'll never forget an executive I worked for who routinely said "I don't disagree with you". My aim isn't just agreement, it's consensus. Why? Consensus involves a process that coalesces inputs and support for a final decision, unlike voting which often relies on compromise. While compromise is frequently necessary, it's only truly effective when built upon shared understanding and a clear picture of priorities. @RayDalio famously ran Bridgewater Associates, championing thoughtful disagreement where the goal isn't to convince, but "𝘵𝘰 𝘧𝘪𝘯𝘥 𝘰𝘶𝘵 𝘸𝘩𝘪𝘤𝘩 𝘷𝘪𝘦𝘸 𝘪𝘴 𝘵𝘳𝘶𝘦 𝘢𝘯𝘥 𝘥𝘦𝘤𝘪𝘥𝘦 𝘸𝘩𝘢𝘵 𝘵𝘰 𝘥𝘰 𝘢𝘣𝘰𝘶𝘵 𝘪𝘵." He emphasized that "𝘪𝘯 𝘵𝘩𝘰𝘶𝘨𝘩𝘵𝘧𝘶𝘭 𝘥𝘪𝘴𝘢𝘨𝘳𝘦𝘦𝘮𝘦𝘯𝘵, 𝘣𝘰𝘵𝘩 𝘱𝘢𝘳𝘵𝘪𝘦𝘴 𝘢𝘳𝘦 𝘮𝘰𝘵𝘪𝘷𝘢𝘵𝘦𝘥 𝘣𝘺 𝘵𝘩𝘦 𝘨𝘦𝘯𝘶𝘪𝘯𝘦 𝘧𝘦𝘢𝘳 𝘰𝘧 𝘮𝘪𝘴𝘴𝘪𝘯𝘨 𝘪𝘮𝘱𝘰𝘳𝘵𝘢𝘯𝘵 𝘱𝘦𝘳𝘴𝘱𝘦𝘤𝘵𝘪𝘷𝘦𝘴." When someone disagrees, I've learned the to lean in. Agreeing is the easy decision, and ritual agreement in an organization is the plague that drifts alignment. In stark contrast, coerced agreement collapses trust, cripples creativity, and yields disengaged colleagues who withhold their full potential. When I was hired at @getclearco , we were all instructed to read 'No Rules Rules' and learn the importance of Radical Candor. In a startup, without an empowered, outspoken team, blind spots go unseen, burnout increases, gossip breeds, and trust erodes. Only through the courageous act of disagreeing well can we truly build, innovate, 𝗔𝗡𝗗 sustainably lead. The challenge: how do you turn our disagreements into drivers of innovation and understanding?

⚡️This is a monster signal. This is the moment frontier AI stops being treated like software and starts being treated like controlled strategic capability. The key phrase is not “customers.” The key phrase is “foreign national Anthropic employees.” That means the state is no longer only controlling chips, model weights, or overseas access. It is moving into cognition access by nationality. That is the real threshold. The U.S. government is saying the highest models are sensitive enough that even people physically inside the United States, working inside the company, may be barred from touching them if their nationality creates deemed-export risk. That is weapons-control logic. This is ITAR logic for intelligence. The corporate language about a “misunderstanding” is probably diplomacy. Companies say that when they need to preserve customer trust, employee morale, and regulatory room. But national security authorities do not force emergency suspension of top model access because someone made a minor paperwork mistake. Something about Fable 5 and Mythos 5 crossed the line: cyber capability, autonomous R&D acceleration, AI-improving-AI utility, bio/security planning, code exploitation, or some blend of all of it. The U.S. state just showed that Anthropic does not fully control Anthropic’s frontier layer. That is the phase change. Labs can brand themselves as public-benefit AI companies. They can talk about safety. They can sell enterprise plans. They can publish model cards. But once the models become national capability, the sovereign arrives. The state does not need to own the company to control the access surface. It only needs legal authority over export, security, procurement, and liability. This confirms the arc we’ve been tracking: Frontier AI becomes state-supervised strategic infrastructure. Public AI splits from strategic AI. Foreign access gets restricted. Labs become quasi-defense contractors. Model access becomes a national security perimeter. Enterprise customers learn that API access is not property. It is revocable permission inside a sovereign-controlled stack. The most important implication is organizational. If foreign national employees can be cut off from frontier systems, AI labs now have to reorganize internally around citizenship, clearance, compartmentalization, and controlled access. That breaks the old Silicon Valley assumption that global talent can freely collaborate around the frontier. The next AI lab structure looks less like Google in 2015 and more like a defense prime crossed with a classified research facility. For markets, the winners are the national champions with U.S.-aligned infrastructure, cleared customer channels, government relationships, compliance capacity, and domestic compute. The losers are open access, foreign-dependent AI wrappers, offshore model distributors, and any enterprise whose moat depends on unrestricted access to frontier APIs. For geopolitics, this is escalation. China will read this correctly. Allies will read this correctly. Every serious state will understand that frontier models are now part of national power. The AI race just moved from “who has the best chatbot” to “who controls cognition as a strategic asset.”

So @Anthropic about to learn the @SpaceX ITAR/EAR lessons Will be very hard for non-nationals to work there and @OpenAI on frontier models. Suppose AGI is the ultimate dual purpose technology

Would you give @AnthropicAI or @OpenAI your ID to get access to frontier models?

What does this mean for AI growth - If @AnthropicAI and @OpenAI can't release more powerful models, what do they do next? Notice how the article mentions "Export Controls" and proving US citizenship as the root of the issue beyond the jailbreak. The security vulnerabilities are a known unknown - the bigger debate will be between identity and privacy.

Turns out the secret to AI Safety was right in front of us the whole time: Mandating that the labs use Microsoft Teams

Great list

𝗪𝗲 𝗧𝗮𝘂𝗴𝗵𝘁 𝗦𝗮𝗻𝗱 𝘁𝗼 𝗧𝗵𝗶𝗻𝗸. 𝗡𝗼𝘄 𝗪𝗲'𝗿𝗲 𝗙𝗶𝗴𝘂𝗿𝗶𝗻𝗴 𝗢𝘂𝘁 𝗪𝗵𝗼 𝗢𝘄𝗻𝘀 𝘁𝗵𝗲 𝗧𝗵𝗼𝘂𝗴𝗵𝘁𝘀. The AI revolution is far more than a technological shift. It is a fundamental re-architecting of digital value. We are moving from the era of "access" to the era of "utility."    This is the next great transition in how humanity consumes knowledge. Before Gutenberg, if you wanted a document, you paid a scribe. You were paying for cognitive labor. You paid for human time and skill to produce text. After the printing press, the cost per unit of thought collapsed. We moved from paying for the labor to paying for the product.    We are seeing that exact pattern repeat in the digital realm. In the old internet, we paid for bandwidth, the pipes. In the new frontier, we pay for tokens, the intelligence flowing through them. Tokens are the new "page" or "copy." with cognitive output.                                      The scale of this shift is difficult to grasp. Consider a 26B parameter open-source LLM. It is the digital equivalent of 5 million books compressed into 15GB. It is essentially the Library of Alexandria, distilled into a portable, executable format. It is 100 times larger and orders of magnitude denser than any physical archive in history.          Just as the printing press took the scattered, fragile knowledge of the ancient world and made it reproducible at scale, AI is transforming raw data into scalable, instant intelligence.                                                                                     This creates a massive tension in our infrastructure. As local devices move toward multi-terabyte storage and models grow in density, our reliance on broadband will pivot from "moving data" to "requesting thought."                                                                        We are witnessing the "ICO moment" for intelligence. Just as blockchain tokenomics transformed financial consensus, AI tokenomics will transform data consensus. When intelligence requires validation and approval, the token becomes the unit of account for truth itself.                                         This leaves us with a critical strategic dilemma. Do we pay for an LLM via a subscription to subsidize the per-token cost? Do we leverage free credits on a new SaaS to reduce customer acquisition costs? Or do we pay-per-use to maintain full control of data? We all have a choice to decide the future of truth.

Anthropic Fable costs $10 / $50 per 1M and GPT-5.5 Pro is $30 / $180 per 1M compared to many models under $0.10 accomplish the task fine. It's the classic 80/20 rule: cheaper models can handle 80% of tasks, but the majority of actual usage comes from the smaller group of heavy users who default to the expensive ones.

Google released Gemma 4 for free. Apple designed its AI to skip the cloud. Two companies just made the same bet from opposite directions. The most underreported story in AI right now isn't which frontier model won the latest benchmark. It's that the gap between open-source and closed models has shrunk to roughly 90 days. Think about what that means structurally. Three months ago's "state of the art" is today's free, locally-running, fully customizable model on your MacBook. @googlegemma 4 26B MoE variant is hitting numbers that would have required a cloud API call 18 months ago and it runs on a single consumer GPU. @huggingface crossed 2M public models and 13M users. @UnslothAI's MTP optimizations are pushing quantized models to speeds nobody expected on consumer hardware. This isn't just a technical curiosity. It's an economics story. @citsecurities recent tokenomics analysis frames AI through the lens of scarcity: compute, power, memory bandwidth, inference costs. Even @Apple sitting on the most powerful silicon in consumer hardware just built its next-gen AI architecture as a hybrid with Google: on-device first, cloud selectively. That's not just a privacy statement. It's rational tokenomics. The implication: the "default to the API" era is over. The real question is how you architect the routing layer, and knowing when $0 local inference is good enough, and when you actually need to spend on frontier capability. Most aren't asking that question yet. The ones that do will have a meaningful cost and control advantage. blog.google/innovation-and-a…

@Citadel Securities just signaled the end of the AI honeymoon. In their "Tokenomics" note, the argument is clinical: even the most transformative technology must obey the laws of cost curves, capacity, and marginal returns. The signals are already visible. @amazon removing usage leaderboards, @Microsoft adjusting @AnthropicAI claude Code subscriptions, and the sudden onset of massive enterprise token bills point to a single reality: the constraints are real. The conversation is shifting from capability to unit economics. It is no longer about what AI can do in principle, it is about the price and scarcity of the inputs required to run it at scale. Compute, power, cooling, and bandwidth are the new binding constraints. The data confirms this pivot. The Silicon Data Token Expenditure Index shows a rollover, signaling a move toward substitution. Companies are trading down from expensive frontier models to "good enough" alternatives to preserve margins. We are entering a period of bifurcation. High-end frontier AI will be concentrated among the few with the balance sheets to absorb the cost. The rest of the market will move toward efficiency and leaner, cheaper models. The hype was built on potential. The reckoning will be built on margins. The question is no longer whether the technology works. It is whether the economics can sustain the valuations. citadelsecurities.com/news-a…

New @googlegemma models got what you need!