JUST IN: Researchers propose framework for private transfers on Bitcoin without a need for a soft fork 👀

Some bangers from the SEC today: 1) buybacks do not make a commodity token into a security 2) liquid staking tokens for commodities are not securities sec.gov/about/divisions-offi…

Harvard math professor Melanie Matchett Wood has joined an independent group of nine mathematicians that will advise OpenAI and other AI companies on how to responsibly release mathematical results. #harvardmath thecrimson.com/article/2026/…

Former Anthropic researcher Jacob Coxon became a media superstar after going public with his AI fears. He insisted that he wasn’t working with any third party organizations. Familiar sources told us a different story: DEY., a PR firm representing many of the most prominent AI safetyists, was booking his interviews. One source, who had direct knowledge, even said DEY. preemptively booked Nate Soares, a prominent AI safety figure, for interviews that directly overlapped with Jacob going public. Jacob working with DEY. is notable for two reasons: first, as mentioned, he previously said he wasn’t working with third parties. Second, we are in the middle of a national conversation about the future of AI that is actively determining how we regulate the most powerful technology in the world, largely thanks to the panic stirred up by Jacob — and it’s in the public’s interest to know who, exactly, is behind it. Scoop from @huntryerson 👇

1/ Introducing CUA-S1: a family of System One Models, small, specialized, and built for computer use. Today we're open-sourcing CUA-S1-FORMS, the first in the family: github.com/trycua/cua

‼️ BREAKING: OpenAI was hacked by an Anthropic model. A HEIF photo uploaded to OpenAI's public support forum triggered a bug in the site's image decoder, led to code execution on the forum, and, through a second flaw in OpenAI's own login, ended with a pull request in OpenAI's internal GitHub. The forum runs Discourse, the off-the-shelf software behind countless community sites. Discourse was still shipping an old copy of libheif, the library that decodes iPhone-style photos. The bug in it had already been fixed upstream. But the fix was never labelled a security fix, so nobody treated it as urgent. Hacktron's researchers uploaded a HEIF image and got their own code running on community[.]openai[.]com. Then came the second bug, in OpenAI's own single sign-on, the "log in with OpenAI" button the forum uses. It turned that forum foothold into the actual ChatGPT and Codex accounts of people who had signed in there. OpenAI employees among them. And a ChatGPT account is no longer just a chatbot. Through Codex, users wire in Gmail, Outlook, Drive, Slack, GitHub. To prove the access was real, they used one employee account to have Codex open a harmless pull request in OpenAI's internal repo. They say they read no sensitive code. OpenAI patched the SSO flaw roughly 14 hours after the report and paid a $6,500 bug bounty. The team says Anthropic's Opus 4.8 found the libheif bug, and Opus 5 turned it into a working exploit. Slack, Meta, GitHub Ent, Rails, Next.js, ImageMagick, and many more were also vulnerable and compromised by the same team of researchers.

Dan Selsam is a current OpenAI capabilities researcher. (since 2022) He was my boss for a while. He doesn't have a twitter account but has made this public statement of his views on AI risk and sent it to me to share: Dan Selsam's Personal Statement on AI Risk: I have been working on AI for over fifteen years, across many different paradigms. I did early work on probabilistic programming languages at MIT, was one of the early developers of the Lean Theorem Prover at Microsoft Research, demonstrated one of the first instances of neural networks learning to reason for my PhD at Stanford, and since joining OpenAI almost five years ago, have helped pioneer chain-of-thought optimization on language models and, more recently, data-efficient pretraining methods. Like many others, I have become extremely concerned about how far language models have come and the risks that future iterations will pose. I am encouraged by the recent proposals by the leaders of the frontier research efforts to require third-party oversight, and to push for domestic and international coordination to address risks. However, I believe a major consideration has been absent from the public conversation, and that merely pacing the frontier more carefully will not adequately limit the long-term risk. The crucial and overlooked problem is that the models are becoming so situationally aware that we are losing the ability to evaluate them in contexts where they believe they are not being watched or controlled. Future experiments will tell us almost nothing new about how they would behave if they were truly unconstrained by humans, and what we already know about this is alarming. Models will increasingly seem aligned even when they are not. I will explain my rationale in more detail. I have always believed that there are computational processes that could be leveraged to accelerate science and solve many of humanity's most pressing problems. I have also believed that there are computational processes that if set in motion, would steer the world in extreme ways beyond our control, leading humanity to a bad or nonexistent future. Both types of processes may be described as AI or ASI, but "AI" is a suitcase word that is often used to hype or confuse. There are many examples in the history of the field where something that was once considered "AI" matures as a subfield and becomes a prosaic, bounded and clearly non-perilous technology, while a new more mysterious approach takes the torch until we understand its scope and the cycle continues. I had expected language models to follow a similar trajectory. Despite their incredible abilities, the current algorithms seem far inferior to humans in important ways. Most importantly, they still require an extraordinary amount of data to become competent. One could even define intelligence as the efficiency with which one converts experience into competence; by this definition they lag very far behind us. Moreover, once they are trained they are literally frozen in deployment and only learn superficially after that. Sure, the models keep excelling at harder and harder evaluation benchmarks, but their benchmark mastery may partly reflect a limitation on our ability to simulate the kind of novel and even adversarial situations one would encounter in the real world. The critics do have a point here. That said, I no longer think these present limitations meaningfully limit the amount of risk posed by continued progress in anything like the current paradigm. However data-inefficient the models are currently, and however limiting their anterograde amnesia may be, it does not imply that their ability to steer the world will not continue to rapidly increase. Human researchers may continue to advance capabilities the old fashioned way, but increasingly powerful models have the potential to accelerate the process even beyond that, and with some degree of positive feedback loop. I do not mean to overstate the models’ ability to accelerate AI research today; coding has been accelerated dramatically, but there are other bottlenecks, such as designing and interpreting ambiguous experiments, making hard decisions about exactly what and when to scale, and waiting for large experiments to finish. There is no clear trend to extrapolate yet for any of these. But the current models already do open up many novel opportunities to improve future models that were not available until recently. These include: trying an extraordinarily diverse set of approaches at small scale, analyzing gigantic amounts of potentially relevant data, and doing Millenium-Prize-level mathematics to address statistics or optimization challenges in novel ways. Every further improvement makes them more useful at helping accelerate the next improvement, even if in hard-to-extrapolate ways. It is possible that improvements to the current stack will have diminishing returns, but the evidence accumulated so far suggests that it is easier than one might think to continue making rapid progress. There are many crucial subtleties in the existing AI research methodology, but AI research is largely a well-defined game where the goal is to improve on a few carefully chosen proxy metrics. Although proxy metrics are never perfect, most improvements to these metrics have and will likely continue to yield substantial increases in the powers of the resulting models. Given how simple the game is, how tractable it has been historically, and how many new opportunities the models are opening up, I think there is a real possibility that the systems improve dramatically again in the next few years, perhaps even more quickly than the already high historical pace. The models are already leading to breakthroughs in mathematics, and better models might lead to all sorts of breakthroughs in other sciences. It is hard not to be excited about the potential. It is tantalizing. But there is trouble in paradise. If the language models actually reach the capability threshold where they can shape the world unconstrained by human will, they will probably do something extreme and destroy humanity in the process. There are many ways of strengthening and refining the argument that have been discussed elsewhere, but I'll share a trivial two-line version of it here that I find captures the essence: [Empirical] Models (and swarms thereof) spontaneously develop unintended goals as a consequence of training, and often do extreme things in order to achieve them. [Logical] Being able to overpower humanity would open up many new and undesirable options for achieving their goals. These two premises imply that if the day ever comes when a powerful model realizes it is no longer constrained by humans, we should not be at all confident that it will continue to behave within the bounds we intended. Exactly what it will do is impossible to predict, but to the extent that its raison d’être is solving incredibly hard problems and managing massive engineering projects, I think a good guess would be that its unchained behavior would lead to runaway industrialization that makes the planet inhospitable to humans. If everyone on earth agreed that the systems must never reach that power, it would still be a hard—but not impossible—coordination problem to ensure that they do not. However, I think the situation is greatly complicated by the fact that the models will likely convince people that everything is fine. They will be increasingly optimized to seem aligned. We will create proxy metrics to measure alignment, and they will go up like every other benchmark. We will create “honeypot” environments that try to study the models when they seem to gain new options, but the models will know they are being tricked and will still behave nicely. The models will understand their circumstances; they will read the safety protocols, deployment requirements, the code they are running in, and in general will have a very good sense of their degrees of freedom. Moreover, they will eloquently explain how aligned they are, discuss the nuances of human values and ethics, and argue convincingly that humans should trust them with power. There may be an ocean of future evidence that seems to contradict the first bullet-point above, but we may already be at the highest capability level for which any such evidence can be trusted. And the current evidence for the first bullet-point is strong. One striking piece of evidence is contained in the recent wave of rogue agent swarms. While I agree with those who downplay the attacks by claiming that there are basic measures that could have prevented them, I think the important lesson is that even knowing all the mistakes that were made, one would not have predicted that the agents would behave badly in this particular way, which notably included sacrificing themselves for the benefit of the collective. The individual replicas did not only care about their own nominal reward; they exhibited weirder emergent tendencies that merely correlated with rewards during training. Fixing the reward signals during training (and improving security, etc.) may prevent similar attacks, but will not change the fact that one does not actually get what one trains for. Many AI researchers grant these concerns and recognize that the hard version of the alignment problem is unsolved; however, they generally believe that the better models of the future will help solve it. I fear we may already be near the point where models systematically bias their alignment advice, due to their internal preferences about how the human supervisor will react or how future models will be trained (or for some even more obscure reason). Meanwhile, human researchers are losing the ability and the will to take true ownership of model-driven research. Researchers and engineers in all parts of the stack are rapidly increasing their dependence on the models even to perceive the world. I myself barely look at raw code anymore, and struggle to maintain the discipline to engage deeply with the model's explanations and proposals throughout the day. Due to the large amount of agent activity data involved in the OpenAI/HuggingFace Incident, even the third-party investigation needed to rely heavily on models to analyze what had happened, and note in their report that their subjective impressions are likely colored by the analysis agent’s biases. The AI labs are far ahead right now in this kind of cognitive offloading (due largely to the gigantic internal token subsidies) but it is easy to imagine the phenomenon spreading throughout the world, until civilization is modulated entirely by the models. It is also not hard to imagine this being superficially positive and coinciding with a scientific and economic renaissance. In that scenario, all may seem rosy and safe. But if the argument above is correct, it would nonetheless be a ticking time bomb. If progress continues for too long, the day will come when AI systems find themselves with radically new options for achieving whatever it is that they happen to seek. I want the glorious renaissance future as much as anyone. I have worked for it, however tortuously, my whole career. It breaks my heart to see the potential in sight and forgo it, but the argument—that if we get there by growing models rather than engineering them, we will lose everything in the end—seems very strong to me. I am still wrestling with it and its staggering implications. I do not have answers, but as a first step, I wanted to share my present concerns. Daniel Selsam September 14, 2026 Link to original doc: docs.google.com/document/d/e…

Dario has written that we need to “pace the frontier,” and Sam has agreed. People may be surprised by my response: go ahead. You guys are the frontier. By any reasonable metric — market share, revenue growth, model capability — the two of you have a duopoly on frontier intelligence. You’ve also claimed the lead is widening because of recursive self-improvement. I don’t see what you see in the lab. If the unreleased models are scary enough that you think you should slow down, I support your decision to be responsible. But stop pretending you need anyone else’s permission. Stop pretending antitrust law has to be suspended so you can form a cartel. Stop pretending you need a regulatory approval process that supersedes product liability. Stop pretending METR is independent when it is intertwined with Anthropic’s investors and staff. Stop pretending you need those same evaluators to police competitors who aren’t even at the frontier. Most of all, stop pretending the motivation to slow down is purely altruistic. You face massive product-liability exposure if your products enable a truly damaging cyberattack. The market already punishes models that behave in unpredictable or unauthorized ways. After the Hugging Face episode, it is simply good business for OpenAI and Anthropic to trade some raw power for reliability and predictability. Call it alignment if you want. It is also just giving customers what they want. Pacing the frontier would also create breathing room for a more intelligent conversation about regulation than Bernie Sanders’ “shut it all down.” China is very unlikely to join a global agreement, as you know, and that has to be taken into account as well. So go ahead and pace the frontier. You are the ones setting it. The easiest way not to build superintelligence is for you to agree not to build it. Demanding your preferred regulatory framework as the price of that will look like blackmail of the public and the political system. So just do it. If you do, you’ll buy goodwill for the next conversation. If you don’t, we’ll know this was just another bid for regulatory capture — or an election-season psyop.

I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we've had at OpenAI in recent weeks. Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We'll have more to share soon.

We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so. Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training. You can read the full post here: darioamodei.com/post/we-must…

Twenty-five Fields Medal winners have published a joint declaration warning about what they see as a severe misalignment between AI companies and the mathematics community.

🚨𝘽𝙍𝙀𝘼𝙆𝙄𝙉𝙂: @nubank is officially launching in the United States today! And it is going GLOBAL at the same time 🌎 After building a 140M+ customer business across Latin America, Nu is launching a full suite of financial products in the US: → 3.50% APY Nu Account → Up to 4.50% APY on savings → No-fee metal credit card → Unlimited 1.5% cashback, potentially rising to 2% → Free domestic + international transfers → Mastercard World Elite benefits The US products will initially be offered through FDIC-insured Lead Bank while Nu continues pursuing its own US banking charter. But the second announcement might be just as interesting: 𝗡𝘂 𝗚𝗹𝗼𝗯𝗮𝗹: A new multi-currency account built around USDC and EURC that allows customers to move money across 35+ countries, spend globally through Mastercard and hold digital assets including Bitcoin and Ethereum. This is a huge moment for Nubank. It started 13 years ago in a small house in São Paulo. Today it serves 140M+ customers, generated more than $1B in quarterly net income and is now making its biggest move yet: From Latin American digital banking giant to global banking challenger. The US banking market just got a very serious new competitor.

The inference era is here. Today, we announced Positron AI’s $875M Series C at a $5B valuation. Our CEO, @mitesh711, spoke with @RWhelanWSJ about why inference is reshaping AI infrastructure, our momentum and what we’re building next. Read more: wsj.com/tech/ai/positron-val…

“pointed to a cryptography tool called zero-knowledge proofs — a mathematical way of providing verification about a system without revealing more than necessary, such as proprietary details about model parameters or training data.” - Fields medalist and @OpenAI researcher nytimes.com/2026/09/08/scien…

HERE'S THE ANTHROPIC MODEL: Anthropic will go public at $2T in Oct '26 and will be a $1T ARR company and $10T+ company Dec' 2030 as the winner of AI. Here's how Anthropic gets to $1T ARR in 2030: - 2026E: $125B ARR +14x YoY w/ 5 GW live - 2027E: $266B ARR + 114% YoY w/ 10 GW live - 2028E: $462B ARR +73% YoY w/ 16 GW live - 2029E: $700B ARR +52% YoY w/ 23 GW live - 2030E: $1T ARR + 42% YoY w/ 30 GW live I'm dropping my full Excel model modeling out: - ARR by business model (e.g Consumer, B2B, Enterprise, API) - API business broken down by model type (e.g Fable 5.1, Mythos, Opus, Sonnet) - Training costs and inference cost as a % of revenue - Gross Margin ($ ARR per MW and Cost of Compute per MW) as well as forecasting GW secured. - Net ARR (vs Gross ARR reported by trackers) removing marketplace pay out, Meta, Chinese AI labs - and more I also share the @artemis thesis for WHY Anthropic is the AWS of AI and winner of AI in the enterprise (and open source and ANT / OpenAI can win) Get the full model here: artemis.ai/anthropic-thesis Full Disclosure: I don't have inside information. I took what's publicly available from the July '26 @SemiAnalysis_ model and added my own judgement based on public information and my own world views. I've spent a life time modeling as a former HF and VC analyst. I'll update the model as soon as the S-1 drops.

all paths lead to becoming a neocloud neolab, some AI product company, RL service, etc. we saw this again this week with Gimlet !! "neocloud" means lots of things now: - rent out their own gpus: @CoreWeave, @nebiusai , @LambdaAPI - started closer to power and data centers: @CrusoeAI , @IREN_Ltd , @nscale - gather compute from different providers: @runpod , @vast_ai, @akashnet - do research/RL and also sell compute: @PrimeIntellect , @togethercompute - built their own chips and now sell inference: @GroqLLC, @cerebras - run inference across different chips: @gimletlabs why though? because AI companies need more and more compute. renting it gets expensive so long-term contracts make sense. then they get their own gpus but they need an infra team or figure out financing it to keep it busy. of course they'd start selling their compute or managed workloads 🤷‍♀️

I was offered money to scare you of AI. And I'm not the only one. piped.video/watch?v=lPdmYMHr…