RehashedDAO

The banking crisis never really ended. It just moved from front-page panic into a slower, quieter phase: commercial real estate stress, unrealized losses, deposit sensitivity, and the growing question of what happens when trust in traditional rails weakens again. For crypto, this is not only a macro story. It is a reminder of why onchain finance exists in the first place👇 ~~ Analysis by @0xAlexCashman ~~ The Crisis Is Not 2008, But It Is Not Gone The mistake is to frame every banking stress event as another 2008. This cycle looks different. The system is more capitalized, regulators are more alert, and the largest banks are not obviously sitting on the same type of mortgage leverage that broke the system last time. But that does not mean the risk disappeared. The current pressure is more subtle: higher rates repriced bond portfolios, commercial real estate remains under stress, depositors are more willing to move money quickly, and smaller banks are still exposed to asset-liability mismatches. The FDIC’s latest data does not show a full systemic breakdown. Problem banks remain within a normal non-crisis range, and no major wave of failures is currently visible. But the same reports also show persistent weakness in certain loan books and elevated unrealized losses. That is the uncomfortable part. A system can look stable at the surface while still carrying slow-moving fractures underneath. Why Crypto Should Care Crypto people often talk about banking risk as if it only matters for banks. That is wrong. The 2023 banking panic showed that crypto is still deeply connected to traditional finance. When Silicon Valley Bank failed, USDC temporarily lost confidence because part of Circle’s reserves were trapped inside the banking system. That moment was important because it broke a simple illusion. Stablecoins may live onchain, but many of their reserves still live offchain. This means crypto can inherit banking risk even when users think they are holding a digital dollar. If the banking partner fails, the stablecoin can face redemption pressure. If payment rails close over the weekend, liquidity becomes fragmented. If regulators step in, the market has to wait for clarity. In other words, crypto does not escape the banking system just by tokenizing dollars. It only changes where the risk becomes visible. The Real Stress Point Is Trust Banking runs are not purely about balance sheets. They are about confidence. A bank can be technically solvent and still fail if depositors no longer believe they can access funds quickly. In a digital world, that trust can disappear in hours. This is where crypto changes the psychology of finance. Onchain markets operate continuously. Stablecoins move 24/7. DeFi liquidations happen automatically. Treasury-backed tokens can be monitored in real time. Wallets do not close for the weekend. That does not make crypto risk-free. But it does create a different expectation: users increasingly want financial systems that are transparent, portable, and always available. When banks stress, crypto’s value proposition becomes easier to understand. Not because “banks are dead.” But because people suddenly remember that access, settlement, and custody are not abstract concepts. They are the whole system. Stablecoins Become The Bridge And The Weak Point Stablecoins are probably the most important part of this story. They are the bridge between traditional finance and onchain finance, but also the place where both risk models collide. On one side, stablecoins give users fast settlement, global transferability, DeFi liquidity, and a programmable dollar layer. On the other side, they rely on reserve management, banking access, short-term Treasuries, custodians, audits, and regulatory treatment. That makes them powerful, but not fully independent. A future banking crisis would likely increase demand for stablecoins, especially outside the U.S., where users may want dollar exposure without relying on weaker local banking systems. At the same time, it would also increase scrutiny around who holds the reserves, how redemptions work, which banks are involved, and whether stablecoin issuers can survive stress without depending on emergency guarantees. The next stablecoin winners may not just be the biggest issuers. They may be the ones with the cleanest reserves, strongest banking relationships, most transparent reporting, and best redemption infrastructure. Bitcoin’s Narrative Gets Stronger, But Not Automatically Every banking crisis helps Bitcoin’s story. A scarce, non-sovereign asset with self-custody and no bank balance sheet behind it becomes easier to explain when banks start looking fragile. But the market does not always move in a straight line. In a panic, investors often sell liquid assets first. Bitcoin can trade like a risk asset before it trades like a hedge. Liquidity shocks can hit crypto hard even when the long-term narrative improves. This is the paradox. Banking stress strengthens the philosophical case for Bitcoin, but it can also create short-term volatility across the entire crypto market. The same applies to DeFi. A banking crisis can make decentralized lending, onchain collateral, and transparent settlement look more attractive. But if the crisis hits stablecoin liquidity or risk appetite, DeFi can also suffer first. The direction depends on whether the market sees crypto as an escape route or as another high-beta asset to sell. Looking Ahead The next banking crisis probably will not look like the last one. It may be less cinematic than 2008 and less sudden than SVB, but more distributed across CRE losses, regional bank pressure, deposit flight, and confidence shocks. For crypto, the lesson is simple. The industry should not celebrate bank stress as if it automatically benefits onchain finance. The relationship is more complicated. Stablecoins still rely on banks. Exchanges still need rails. Institutions still need custody. Users still need fiat on and off ramps. But the direction of travel is clear. Every banking shock makes the case for transparent reserves, 24/7 settlement, self-custody, tokenized Treasuries, and neutral financial infrastructure a little easier to understand. The banking crisis is not just a threat. It is a stress test for the old system and a credibility test for the new one. Crypto does not win by saying banks failed. Crypto wins if it can prove that open financial rails are safer, faster, and more resilient when trust starts breaking elsewhere.

@Marino1349916 @HesterPeirce 👀

@Web3Rehashed @HesterPeirce there is a reason her comments continue to get revisited years later

Revisiting our long-awaited conversation with SEC Commissioner @HesterPeirce. For years, Commissioner Peirce has remained one of the most intellectually rigorous and independent voices within the SEC on questions of digital assets. She has repeatedly challenged the tendency to treat technological experimentation as a regulatory threat by default, arguing instead for clearer standards, proportionate oversight, and a framework that does not punish innovation through ambiguity. Her perspective remains essential for anyone trying to understand the institutional tensions shaping crypto policy in the United States.

@Muhamma392824 @HesterPeirce Peirce has consistently understood that these are not the same objective.

@Web3Rehashed @HesterPeirce A regulator willing to distinguish between genuine misconduct and technological experimentation is far more valuable than one that treats both as the same thing.

The SEC is often perceived as distant, opaque, and fundamentally inaccessible to early-stage builders. That perception is understandable, but it is not inevitable. In our conversation with SEC Commissioner @HesterPeirce on @web3Rehashed, she offered a more practical view of regulatory engagement: approach the institution early, communicate with precision, and treat compliance not as an afterthought, but as part of building durable infrastructure. For founders operating in an uncertain regulatory environment, that distinction can be decisive.

@ErenSignals @AnthropicAI 💯💯

@Web3Rehashed @AnthropicAI The first read is the headline. The second read is who suddenly starts acting differently.

The core news is not that @AnthropicAI temporarily took two models offline. It is that the United States has now treated access to frontier AI as something that can be controlled through export law, not only by geography, but by who the user is. Anthropic was reportedly ordered to suspend access to Fable 5 and Mythos 5 for any foreign national, including people physically located inside the United States and even foreign-national Anthropic employees. That is a very different regime from blocking chip shipments to China or limiting GPU clusters abroad. It is closer to saying: the intelligence itself has become a controlled strategic asset. 👇 ~~ Analysis by @onchainhost ~~ For years, the AI race has been framed around inputs. Who can buy the best GPUs. Who can secure HBM supply. Who can build data centers fast enough. Who can access advanced lithography, power capacity, and cloud credits. The US export-control strategy reflected that logic: restrict Nvidia-class hardware, constrain semiconductor tooling, limit advanced compute in adversarial jurisdictions, and slow down the ability to train frontier models. That framework made intuitive sense because chips are physical. They cross borders. They are manufactured, shipped, tracked, and sold through identifiable supply chains. But models are different. A frontier model can sit in a US data center while serving someone on the other side of the world through an API. Nothing physical crosses a border. No GPU is exported. No model weights need to leave the company. A few API calls can still deliver capabilities that used to require access to an entire research organization. That is why this @AnthropicAI episode matters. The Commerce Department directive reportedly required a license for foreign persons to access Fable 5 and Mythos 5, regardless of whether those users were in the US or abroad. Anthropic said it could not reliably separate its users by nationality, so the practical outcome was to disable both models for everyone. In other words, the state did not control the hardware. It controlled the ability to query the intelligence running on the hardware. That distinction may sound technical, but it changes the structure of the AI market. A frontier API is no longer simply a cloud product. It can become a licensed strategic service. Anthropic’s Fable 5 had been released as a generally accessible model with cybersecurity safeguards. Mythos 5 was more restricted, intended for a smaller trusted-access group where some cyber safeguards were lifted for defensive use cases. Anthropic itself described Mythos-class systems as a higher capability tier than its Opus models, particularly for software engineering, autonomous work, cyber defense, and scientific research. The government’s concern was reportedly that Fable’s safeguards could be bypassed in a way that enabled users to identify software vulnerabilities. Anthropic disputes the characterization. The company says the technique was narrow, non-universal, involved a limited number of previously known minor vulnerabilities, and did not demonstrate a capability unique to its models. It also argues that similar bug-finding behavior is available through other public models. The truth is that both sides may be describing a real problem from different angles. Anthropic is right that jailbreak resistance is not binary. A model can have strong protections and still be vulnerable in narrow contexts. That is the nature of frontier model security today: safeguards reduce the cost of defense, but they do not produce perfect containment. The government is also right about one thing: capability diffusion does not need to be perfect to matter. A model does not need to autonomously compromise a military network to create strategic risk. It may be enough for it to make skilled researchers, cyber operators, or intelligence teams materially faster at vulnerability discovery, exploit research, systems analysis, or code review. The issue is therefore not whether a model is “dangerous” in the abstract. The issue is whether certain increments of capability are significant enough that access itself becomes a national-security question. That is a much harder line to draw than the chip line. A chip can be classified by performance thresholds. Compute capacity can be estimated. Interconnect bandwidth can be measured. A model’s strategic value is more contextual. The same model that helps a defensive security team patch an aging banking system can help an offensive researcher find weaknesses in that system. The same agent that compresses months of software engineering into days can compress reconnaissance, reverse engineering, and exploit development. And the same model that can be used by a US cybersecurity firm through a legitimate API can potentially be used by a foreign actor through the exact same interface. This is why the “foreign national” language is the most consequential part of the story. The policy is not simply saying: do not serve sanctioned jurisdictions. It is applying the logic of deemed exports, where releasing controlled technology to a foreign person inside the United States can be treated as an export to that person’s country of nationality. That principle already exists in traditional export controls. What is new is applying it to real-time access to a commercial frontier model. This makes the situation less like a normal product restriction and more like an emergency intervention. And that uncertainty is itself a market signal. For enterprises building core workflows around frontier APIs, the risk is no longer limited to pricing changes, rate limits, outages, or model deprecation. There is now geopolitical dependency risk. A company in London, Seoul, Dubai, Singapore, or Istanbul can build its product architecture around a US model, integrate it deeply into engineering workflows, and then discover that access is conditional on a political or regulatory decision made in Washington. That is not a theoretical concern anymore. Anthropic’s own decision to disable access globally shows the operational reality. A compliance requirement aimed at foreign nationals became, in practice, a kill switch for everyone because identity verification, citizenship classification, licensing, corporate ownership analysis, and access enforcement are extremely difficult to implement across global cloud infrastructure. This is where the AI sovereignty conversation becomes much more concrete. For a long time, “sovereign AI” sounded like a policy slogan: countries wanting local language models, domestic clusters, national compute programs, or regional data residency. Now it has a more practical meaning. Sovereignty is not only about owning GPUs. It is about whether a government, company, university, security team, or startup can maintain access to the intelligence layer when geopolitical conditions change. That will make open-weight models more strategically attractive, even when they are less capable. Not necessarily because open models are better. But because a model that can run on infrastructure you control cannot be switched off by a foreign provider under an emergency licensing directive. That creates a major tradeoff. Closed frontier systems may remain ahead in capability, reliability, tool use, long-horizon reasoning, and safety infrastructure. But they also concentrate political and regulatory power inside the provider’s jurisdiction. Open-weight systems sacrifice some of that frontier performance, but they reduce dependence on a single company, a single cloud platform, or a single national export-control regime. For builders, this probably accelerates a multi-model future. The question will not only be: “Which model performs best?” It will increasingly be: “Which model can we still access under stress?” That could mean enterprises keeping secondary model providers, designing agent stacks that can swap inference backends, maintaining open-weight fallback systems, or avoiding architecture that assumes one frontier API will always be globally available. This does not mean the US will suddenly export-control every powerful model. The current action is specific to Anthropic’s Fable 5 and Mythos 5, and the facts remain contested. The government has not publicly released the full legal reasoning, while Anthropic maintains that the cited jailbreak was narrow and that broader restrictions would risk halting frontier deployment across the industry. Still, precedents matter more than permanence. Once a government demonstrates that it can regulate model access through export-control authorities, every other frontier lab has to plan around that possibility. @OpenAI, @GoogleDeepMind, @xai, @Meta, and the major cloud providers all now have a reason to ask the same uncomfortable question: At what level of capability does a model stop being software and start being controlled strategic infrastructure? The answer will shape more than AI policy. It will shape where startups build, how enterprises procure models, why countries invest in domestic compute, and whether the next generation of AI becomes globally accessible infrastructure or a fragmented network of national capability zones. The most important thing to watch is not whether Fable 5 comes back online next week. It is whether this becomes a one-off dispute around @AnthropicAI, or the first real template for governing frontier intelligence as an export-controlled resource.

@Welinto2613959 @AnthropicAI And that is why implementation may be more disruptive than the legal theory. A cloud platform can identify an IP address more easily than it can reliably establish nationality, licensing status, beneficial ownership, or the changing status of every employee using an API.

@Web3Rehashed @AnthropicAI The next procurement question for serious AI users may be less about benchmark performance and more about continuity of access ://

@caacaaiu @AnthropicAI Exactly. The dependency is no longer just on compute supply or a vendor’s pricing power. It is on a jurisdiction deciding whether your organization is permitted to access a specific level of intelligence at all.

@Web3Rehashed @AnthropicAI Deemed-export logic was built for the transfer of technical knowledge. Applying it to a live model API effectively turns nationality, rather than location or conduct, into an access-control variable.

@babakyusiti @AnthropicAI Thanks.

@Web3Rehashed @AnthropicAI Good post.

@adelbucetta @0xbilly That's a tough one

@Web3Rehashed @0xbilly that's where most people stop, stuck on the scary scenarios. but what if we asked instead: how do we design systems that amplify human potential, rather than just making money from it?

At one point in our episode with @0xbilly, we went down a bit of a dystopian rabbit hole. The conversation started with creativity, AI, and the speed at which ideas can now become real, but quickly moved into a bigger question: what happens when the world around us becomes easier to generate, manipulate, and personalize? Billy reflected on a future where AI does not just help us create faster, but begins to reshape how we experience reality itself. The line between imagination, media, identity, and environment starts getting thinner, and that creates both incredible possibilities and very real risks.

@XueJia63277 @0xbilly Facts.

@artifact203815 @0xbilly True. It’s impossible not to feel both awe and caution at the same time.

@winkle345275 @0xbilly Yeah… that’s exactly the part that hits a nerve. Personalization feels powerful, but also strangely invasive when you think it through.

@T34370077907 @0xbilly @ethereum True. But it also shifts the whole emotional landscape. When skill is no longer the barrier, the only thing left exposed is what we actually feel..

How will AI reshape our creativity and the world around us? Earlier this year, we opened Season 3 with @0xbilly from @ethereum for a conversation about what happens when the distance between imagination and execution starts to collapse. Billy spoke about a future where ideas can move from thought to form almost instantly, where creativity becomes less limited by tools, technical skill, or production bottlenecks, and more defined by taste, intention, and the ability to see clearly. In this clip, he reflects on how AI may change not only what we create, but how quickly the world around us can be redesigned once more people are able to turn ideas into reality at the speed of thought.

@widfiretoa25650 @0xbilly @ethereum Totally feel this. When creation becomes instant, the real magic shifts to how bravely and honestly we iterate. That’s where the human part still lives.

Two days ago, Anthropic launched Claude Fable 5 and Claude Mythos 5. On the surface, this looks like another frontier model release. But I think the more important story is not just capability. It is access. Fable 5 is the public Mythos-class model. Mythos 5 is the same underlying model, but with some safeguards lifted for trusted cyber defenders, infrastructure providers, and eventually select biology researchers. In its early stages, Anthropic is demonstrating what the next phase of frontier AI deployment may look like👇 ~~ Analysis by @punkbennet ~~ I, like many others, have become slightly numb to model launches. Every few months, a new model arrives with better coding, better reasoning, better long-context performance, better benchmark charts, better agentic workflows, better everything. At some point, the launch cycle starts to blur into one long benchmark war. But the Fable / Mythos release feels different. Not because Anthropic is claiming another step forward in intelligence, though it is. Not because the model seems strong at long-horizon coding, scientific reasoning, vision, finance, and complex knowledge work, though that matters too. It feels different because Anthropic is openly splitting capability into two layers: a public version and a trusted-access version. That is the real story. Fable 5 is described as a Mythos-class model made safe for general use. According to Anthropic, it exceeds any model they have previously made generally available and is especially strong on long, complex tasks. The model is available to general users, but it ships with classifiers that detect certain categories of high-risk use. When those classifiers trigger, the request does not get handled by Fable 5. It falls back to Claude Opus 4.8. The covered areas are cybersecurity, biology and chemistry, and distillation. In plain language: domains where the model’s raw capability could create meaningful risk if used badly, or where Anthropic believes unrestricted access could accelerate misuse or capability proliferation. Anthropic says these safeguards trigger in less than 5% of sessions on average, meaning most users should experience Fable as the full Mythos-class model most of the time. But that 5% is where the entire debate lives. Because if you are building a normal app, analyzing documents, writing code, doing finance work, or working on general research, Fable 5 may simply feel like a stronger frontier model. If you are doing security research, advanced biology, chemistry, or frontier model development, the product experience becomes more complicated. That is where Mythos 5 comes in. Mythos 5 is the same underlying model as Fable 5, but with safeguards lifted in some areas. It is not generally available. It is being deployed through Project Glasswing, Anthropic’s initiative with cyber defenders and critical software infrastructure providers. Anthropic says it plans to expand access through a broader trusted program. This is a meaningful shift from “everyone gets the same model” to “capability access depends on trust, use case, and risk category.” I do not think this is just product packaging. It is probably a preview of how frontier AI gets distributed from here. In crypto, we are used to open access as a cultural default. The whole industry is built around permissionless infrastructure, public networks, open liquidity, composability, and adversarial testing. The assumption is that if something is powerful, the network should expose it, and the market should figure out what survives. Frontier AI is moving in a different direction. The most capable systems are becoming too useful to keep entirely closed, but too risky to release without restrictions. That creates a middle layer: broad public access for most tasks, gated access for sensitive domains, and institutional partnerships for the highest-risk capabilities. There is a strong argument for this. If a model is genuinely good at finding and exploiting software vulnerabilities, then unrestricted release has obvious downside. Anthropic previously said Mythos Preview had found thousands of high-severity vulnerabilities, including some in major operating systems and browsers. Even if we treat that as an Anthropic claim rather than independent proof, the direction is clear: models are becoming serious cyber tools. That means the same capability can be defensive or offensive depending on who holds it. A security team using Mythos to audit critical infrastructure is very different from an unknown actor using it to automate exploit discovery. A biology researcher using the model to generate therapeutic hypotheses is very different from someone trying to gain dangerous biological uplift. The difficult part is that the boundary is not clean. Dual-use work is messy. Real security research can look like offensive security. Real biology can overlap with sensitive methods. Real AI research can look like distillation or capability extraction. If the classifier is too narrow, malicious users get through. If it is too broad, legitimate researchers get blocked or silently downgraded. This is why the transparency issue matters. After launch, Anthropic already faced backlash around invisible safeguards for frontier LLM development. The criticism was not only that the model had restrictions. Most serious users understand that frontier systems will have restrictions. The criticism was that some interventions were not visible enough to the user, which makes evaluation harder and damages trust. If a model refuses, that is annoying but clear. If a model falls back to a weaker model and tells you, that is also clear. But if a model quietly changes behavior, limits effectiveness, or routes around your task without making the intervention obvious, then developers cannot properly evaluate it. Researchers cannot know whether they are testing model capability, product policy, or invisible steering. That is a major problem. To Anthropic’s credit, they appear to have recognized this quickly and said they are changing Fable 5’s safeguards for frontier LLM development to make them visible. That is the right direction. Still, the tension does not disappear. The bigger question is whether frontier AI companies can build trust while also reserving the most powerful capabilities for trusted actors. This is not just about Anthropic. It is about the governance model of the entire AI stack. The public wants access. Developers want predictable behavior. Researchers want measurable capability. Governments want security. Labs want to avoid catastrophic misuse. Competitors want fair evaluation. Enterprises want privacy, reliability, and compliance. All of those demands collide inside a release like Fable / Mythos. Another under-discussed piece is data retention. Anthropic says Mythos-class traffic requires 30-day retention for safety monitoring, while also saying the data will not be used to train new Claude models and will be deleted after 30 days in almost all cases. That may be reasonable from a safety perspective, especially if the goal is detecting jailbreaks or coordinated misuse across many requests. But for enterprises, regulated industries, and sensitive research teams, it becomes a real deployment consideration. The more capable the model, the more likely users want to use it on sensitive work. The more sensitive the work, the more important retention policy becomes. So the model is not just competing on intelligence anymore. It is competing on governance. This is probably where the AI market is going. The best model will not simply be the one with the highest benchmark score. It will be the one that offers the best combination of capability, transparency, access control, reliability, compliance, cost, and trust. Fable 5 and Mythos 5 are interesting because they expose that full stack at once. There is the capability story: a model above Opus-class, built for long-horizon tasks and advanced reasoning. There is the safety story: classifiers, fallbacks, red-teaming, limited access, and trusted programs. There is the product story: public users get Fable, vetted users get Mythos. There is the trust story: users need to know when they are interacting with full capability and when safeguards are shaping the output. There is the market story: frontier AI is becoming less like a normal SaaS product and more like critical infrastructure. Personally, I think this release is one of the clearest signs that “open vs closed” is no longer the only useful framing. The new framing is closer to: who gets which capability, under what conditions, with what monitoring, and with what disclosure? That is less clean than the old debate, but probably more accurate. Based on the available information, Fable 5 may become an important public frontier model. Mythos 5 may become an important restricted capability layer for security and science. But the bigger experiment is the access model itself. If Anthropic gets the balance right, this could become a template for deploying very powerful AI safely while still letting most users benefit from the capability. If they get it wrong, it becomes a trust problem: too much opacity for developers, too much restriction for researchers, and too much central control over frontier capability. Either way, this is worth watching. Not just because Mythos looks powerful. Because it shows how AI labs may decide who is allowed to use power at all.

@AldiRasit87071 @jonwu_ 🫶🫡

@Web3Rehashed @jonwu_ Thanks team for putting this out!

“Most crypto debates get stuck because everyone is reacting to the same surface level framing. The interesting part is usually the thing people are not saying out loud.” This week’s guest Jon Wu (@jonwu_) has a rare ability to take topics that everyone in crypto talks about and rebuild them from first principles. Our host Diana (@onchainhost) sits down with Jon to talk about narratives, services, software, founders, and why the most useful perspective is often the one that goes against the grain without trying to sound contrarian for attention.

Really enjoyed my conversation with @jonwu_ from @aztecnetwork. We spoke about the way crypto narratives form, why so many conversations in this space become repetitive, and how founders can think more clearly about services, software, distribution, and trust. Jon has a very specific way of taking familiar topics and turning them slightly sideways, not just to be contrarian, but to get closer to what is actually happening underneath the surface. That made the conversation feel honest, sharp, and genuinely useful. Grateful to Jon for the time and perspective. And special thanks to the @web3Rehashed team, as well as to @MaxArt_eth and @hanasukai_eth for helping make it happen.