TheDinarian
News • Business • Investing & Finance
💥Digital Dollar Likely Won't Be Part of Retail Banking World, US Lawmaker Says💥
September 21, 2022
post photo preview

White House reports on central bank digital currencies "point the way" but Congress still has to pass legislation on these issues, Congressman Jim Himes told CoinDesk.

A U.S. central bank digital currency (CBDC) may be one step closer to reality after the White House published several reports analyzing the technical and policy aspects of a digital dollar last week. Congressman James Himes (D-Conn.) has been an outspoken advocate for a U.S. central bank digital currency, going so far as to publish a white paper on the issue in June 2022.

Himes, who chairs the House Financial Services Committee’s Subcommittee on National Security, International Development and Monetary Policy has also overseen a number of hearings on crypto assets and their role in national security and related issues.

The six-term Congressman spoke to CoinDesk after the White House, Treasury, Commerce and Justice Departments published half a dozen reports in response to U.S. President Joe Biden’s executive order on crypto on Sept. 16.

The following interview has been edited for clarity.

CoinDesk: Thank you so much for joining me. I really appreciate it.

Congressman Jim Himes: Yeah, happy to be with you.

I'm sure you must be busy and it's Friday, so let's get right into it. It's been a pretty intense day with the six reports published by the White House, as well as several federal agencies or departments today. But I know you in particular, have been talking about central bank digital currencies for quite a while now, and of course, you published a white paper titled “Winning the Future of Money,” I think in June, right, a few short months ago. What's your take on the multiple papers published today by the White House and the Treasury Department on this issue of central bank digital currencies?

The element of the various releases on central bank digital currency didn't break a lot of new ground. I was very happy to see that in the text, they sort of emphasize the importance of the United States not getting left behind technologically, I actually think that may be one of the more compelling reasons to continue, but it is a lot of work on the technological side, on the implementation side, making sure that if we do a CBDC, that it's really a very robust network with all of the safety considerations we would have.

I was glad to see they said, "let's keep cranking away." But, that's not enormously groundbreaking, I think. Overall, sort of when you move outside of just the narrow alley of central bank digital currency, I think the White House's releases were a good contribution to an effort that is really picking up very notable momentum in Washington.

I actually think the action is largely on Capitol Hill, in the Financial Services Committee on which I sit, there's a bipartisan effort to get a stablecoin bill put together. On the Senate side, you've got obviously a Agriculture Committee bill that would describe the authorities of the Commodity Futures Trading [Commission].

Where the rubber meets the road, I think Congress is making good progress. Now, not necessarily like we're going to pass new laws in the next couple of weeks progress, but you have to remember that two years ago, if you'd said “digital asset” or “cryptocurrency” in the halls of Congress, most people would look at you funny and not know what you were talking about. So I do think that there's been real progress on Capitol Hill.

Out of curiosity, have the reports today, or even just your own views on central bank digital currencies, have they evolved? Or where do you see any changes between what you've published, what the Fed is looking at, what Treasury published, and just this conversation in general that we’re hearing right now around CBDCs?

I think the story of the last couple of months has been one of the market re-instilling some rationality to the digital assets market. An awful lot of people have lost an awful lot of money, and that makes me sad, but – and when I say an awful lot of money, I mean you hear figures like $2 trillion thrown around, that's just a staggering amount of money. That says two things. Number one, clearly, people's desire to expose themselves to digital assets got way ahead of the underlying value, however you would wish to define that. And for that and other reasons, I think it's actually really good that Washington is beginning to focus hard.

Now, that doesn't mean that Washington is going to satisfy people. When you talk about digital assets, you've got points of view often extremely aggressively expressed, I'm here to tell you, ranging from the pure libertarian – total anonymity, untraceable, whatever – to the world of a central bank digital currency, where you see China actually operating what, my guess is, there's not a lot of privacy protection there.

You also have the interesting fact that we still don't have a digital asset that is proving to be a robust means of exchange, and it may be fun to contemplate questions like whether bitcoin is an appropriate asset class in your 401K. But where this will really be interesting for people is if and when it becomes a medium of exchange, and you can send money to South America or buy a consumer good in the UK, and obviously, we're not there yet.

To that point, do you think that the reports that we saw today are really doing enough to address these questions of usability and still maintaining some semblance of privacy, some semblance of not being a tool for just censorship or surveillance?

Yeah, I think they point the way. The language was pretty strong on urging the regulators to really take a firm hand with the more irresponsible behavior that we've seen. There are an awful lot of people buying into digital assets that are either imperfectly described, maybe that's a euphemism, where people really don't know what they're buying, to out and out fraud. Of course, the SEC and others have been pretty aggressive about going after the fraudsters

I think that the administration's releases point the way, but the real action, the real specifics aren't in those releases, right? The real action and the real specifics will ultimately be incorporated into legislation in the place where I work. And like I said, I'm gratified that there's been a lot of education, but I do think that the time is now to start to start moving something.

I'm very hopeful, for example, we're running out of time in this Congress, but I'm very hopeful that the Financial Services Committee might produce a stablecoin regulation bill, and … the Senate seems to be taking the lead on the jurisdictional questions of what regulator has authority over what product and hopefully, we'll make some real progress. And if we don't actually get anything done in this Congress in the few months remaining, then in the next Congress, we're in a position to do so.

Just jumping on that, do you think that the level of education is at a point now where once you're done with this stablecoin bill, and between you and me, I think the stablecoin bill sounds a lot like, between Libra, between the collapse of Terra/Luna, there's been a lot of stablecoin-specific action. Do you see other crypto issues, to the SEC and CFTC jurisdiction, for example, coming up and being something that we can see actual legislation on within the next maybe a year or two?

Yeah, absolutely. In fact, you already see it on the Senate side. It's not where I work, but on the Senate side, you already see the Senate Agriculture Committee defining the role for the CFTC, so you already see that happening. Again, I wouldn't anticipate – particularly with an election coming up in seven weeks or so – I wouldn't anticipate that that will pass. But this is how we start educating and discovering kind of what the various equities are.

I sometimes joke, this is a really interesting and important space, but it was introduced to the Congress in just about the most catastrophic way possible. And, of course, I'm referring to the hearings that the Financial Services Committee held on Libra. Prior to that, I don't know that many members of Congress had ever even considered the concept of a stablecoin or knew much about digital assets. I sometimes joke that if you had a bunch of evil lobbyists sitting around after a bottle of whiskey and saying, "what's the most catastrophic way to introduce the Congress to a concept," one guy would say, "well, give me [Meta (formerly Facebook) CEO Mark] Zuckerberg." And then let's have him talk about a pervasive global currency. I mean, it was just a catastrophe, right? That sort of soured an awful lot of people for no particularly good reason. I mean, I don't know that there's anything wrong with Mark Zuckerberg, I'm just saying that as a sort of presentational matter that may not have been the best introduction.

You've had a lot of work done since Libra to educate people. I am hopeful that in the next year or so, we may see a really serious stab at providing some regulatory clarity here.

Not to get into specifics here, but do you see any projects or any efforts out there – and feel free to not name specific names – but any projects that are kind of the counterexample to Libra? Ones that lawmakers can look at and say, "wow, okay, this makes sense, this is something that appeals to me, and this is helping me understand what you're trying to do better?’

I probably would not get in the business of predicting which models which stablecoins are likely to again cross that gap of becoming a common medium of exchange. We're not there yet, but there's no question in my mind that there's a use case there. Whether stablecoins are going to replace the current payment systems that are out there, everything from your debit card to Zelle and Venmo and all the various payments, I don't know, I'm a little skeptical about it. You don't you don't look at those current payment methodologies and say, "boy, this is really a pain in the neck.”

But, I have no doubt that two things are gonna happen. I sometimes draw an analogy, and maybe I'll be accused of being naive here, but I sometimes draw an analogy between the way we're thinking about crypto assets generally today, and the way we were thinking about the internet in, let's say 1996 or 1997. We sort of sensed that there was something there. There were all kinds of what, in retrospect, were absolutely wacky ideas. We're going to deliver cat food or kitty litter to your door for free, all these sorts of models that turned out to be sort of crazy. I'm not sure that we would have necessarily in the mid-90s predicted exactly what the internet was going to do. But lo and behold, it transformed our lives, really. I sort of feel like we're in a parallel moment like that.

If we sort of expand the aperture to blockchain generally, not just digital assets, there's no question in my mind that there's going to be some transformative aspects of it. But in the meantime, we're going to see a lot of nonsense, and we may not know that it's nonsense until an awful lot of people have lost a lot of money and in a non-common sense business model.

So I want to jump on something you said just now, referring to existing payment systems and tools, and this is kind of tying back to CBDCs, but the Fed recently announced that it's hoping to launch FedNow as a real time payment system within the next year. Given that, does the calculus around focusing on a central bank digital currency or digital dollar, is it the same? Or does it have to change now that the Fed is actually moving to be more active with this new system that you can argue solves a lot of the same kind of issues that digital dollar would try to solve?

Yeah, I think so. I think that's right, in the wholesale arena. I think that FedNow is probably a part of a really good, innovative modernization of our financial sector generally. It was not that long ago that trading 100 shares of stock was a $200 commission proposition with five days of closing, there's all kinds of risk associated with that. There's no reason for it, right? The only reason that transactions don't close instantaneously today is that the architecture doesn't support that. And so I do think FedNow is a really good step in the direction of where we want to be, which is taking out an awful lot of the time that used to be involved in the clearing and settlement of securities and currencies and commodities.

I think it's really good, but where I don't think we're going to go, I wouldn't say, "well, it's going to penetrate into the retail banking world." There are those who make the argument that individuals should be able to open an account at the Federal Reserve, or maybe they think it's a postal banking thing, a public banking thing. The idea of postal banking is certainly not unprecedented, and it's worth thinking about, I guess.

I do think that the notion that we're going to take the Federal Reserve, who already has massive regulatory duties, and by the way, needs to run our monetary policy and say, "now, you're going to be the banker to 320 million Americans," and in doing that, we're going to wipe out what is one of the primary competitive advantages of the United States, which is our banking sector, I think that's probably not likely. There may be those who think it's a good idea, but I think they're in a pretty small minority.

Fair enough. So something that I think is a little unique about your experience is you chaired the Subcommittee on National Security in House Financial Services and you're part of the Select Committee on Intelligence. Just looking to this idea through those lenses specifically, are there any maybe national security or national interest questions that you think a digital dollar could really address? Or just how are you looking at these questions or even the accessibility through those lenses?

I might add to your list too, I chair the Select Committee on Economic Disparity and if I can take you off course for one second, I get really excited about the possibility that digital assets could ultimately bring more people into a bank environment, or if not a bank environment, at least provide products and services that are cheaper and more relevant to more people.

So let's imagine a central bank digital currency exists. My intuition is, and it's only my intuition, is that it might have a special appeal because it's full faith and credit, it might have a special appeal to a percentage of Americans who are people in our country, by which I mean immigrants, who don't trust the banking system, who are skeptical of financial institutions, but if they believe that the money on their phone is full faith and credit, they might actually use it for payment, they might use it to do cheaper, money transfers, perhaps to family and other countries.

So I get pretty excited about the opportunity to expand in a cost effective way, services to people who are underserved, or if they are served, they're served by very high cost financial products. That's not your question, but let me come back to your question, which is that of course, I think that, like anything else, like any technological innovation, digital assets pose both opportunities and threats to our security. The obvious one that one talks about all the time is, anonymity poses some very serious issues. I mean, who really wants to use a fully anonymous payment mechanism? Yes, my libertarian friends want to use that because they don't want the, whatever the government knowing what they're doing.

But the other group of people who use that, of course, are those who are up to no good whether it's drug dealers or terrorists or human traffickers. So there's that and then there's also the interesting question and if you were British or Chinese or Korean, you would probably regard this differently than then I, that we regard it as Americans, which is the U.S. built SWIFT Network the clearing programs, the international payment mechanisms are one tool with which we are familiar and when we need to we can get visibility when and – this may be a particularly American thing – when you go before a judge and demonstrate probable cause you can actually access the information of those of whom you suspect breaking of breaking the law. That may not be true of other payment systems that are hosted or sponsored by other countries.

Editor’s note: Due to technical difficulties, Rep. Himes was asked to repeat his response to the final question.

An awful lot of people, the estimates are that 19% or so of Americans are unbanked or underbanked. Part of that, of course, is that a lot of people have suspicions about the big financial institutions and I'm intrigued by the possibility that a full faith and credit CBDC for example, might might offer the confidence that would cause somebody to use that as a payment mechanism or as a way to remit money to a home country or something like that. I do think there's real possibilities there, not to mention the possibilities that could be generated either by the private sector directly or by the private sector building on a digital token that was a full faith and credit card thing.

In the more traditional realm of national security there's what we always worry about, which is the question of anonymity if you have a payment system into which we have no visibility and that could be a foreign payment system or a payment system, which is deliberately obscured like what you see with some of these mixtures and such. There's, I think, two categories of people who really need anonymity. There's libertarians, who want that for their own reasons, and then there's, of course, people for whom anonymity is a professional necessity and that's the folks that are up to no good.

There may be others but obviously, we do not want a totally opaque means of payment that could be abused by terrorists or dealers or human traffickers. The other and the last thing I would say is to the point of transparency we're good in this country in terms of not abusing American civil rights or U.S. person civil rights, I should say, the distinction being that if you're in this country, regardless of if you’re a citizen or not, you're entitled to constitutional protection. We have a system that says that if you convince a judge that Sam is potentially committing a crime, that judge will give you permission to get evidence of that crime. There are plenty of countries where you wouldn't want that, because they don't care about civil rights. That's a pretty important part of our justice system here.

Lastly, I would just note we don't want technological developments to get radically away from us. The United States since World War II has been a technological leader in every realm, and we don't want to be – I suppose it's okay to be a fast follower, but we really don't want to be left behind by Chinese innovations or even European innovations. We may not worry about the Europeans as a foe, but every time I contemplate the possibility that we might not be at the technological forefront, it's sad.

LINK

community logo
Join the TheDinarian Community
To read more articles like this, sign up and join my community today
0
What else you may like…
Videos
Podcasts
Posts
Articles
🚨 BOMBSHELL MEDICAL REPORT: Florida Launches Official Study into 77-Cent Ivermectin to Fight Stage 4 Cancer!

Florida shatters the Big Pharma consensus! The State officially launches funding for generic drug repurposing, investigating 77-cent Ivermectin for cancer treatment.

⚠️ Breaking the Big Pharma Monopoly: How the State of Florida is Investing Taxpayer Dollars into Generic Drug Repurposing to Bring Unprecedented Hope to Cancer Patients.

00:02:39
👁️ THE KILL CHAIN AUTOMATED: Palantir, the DOD, and the Age of AI Warfare 🛰️⚡

While the public debate remains focused on consumer AI chatbots, the defense-industrial complex has quietly deployed real-time artificial intelligence into operational military decision-making.

Here is what you need to know about the integration of Palantir’s AI infrastructure and military data networks:

📡 1. Shrinking the "Kill Chain"

Through platforms like Project Maven and Palantir's Artificial Intelligence Platform (AIP), military surveillance systems process massive streams of satellite imagery, drone telemetry, and signals intelligence in real time.

🔹 The Goal: Reduce target identification, processing, and decision workflows from hours to seconds.

🔹 The Reality: Data streams from edge sensors (drones, aircraft, satellites) are fused instantaneously, surfacing potential targets directly to operators with automated strike recommendations.

🛡️ 2. Sensor Fusion & The Tactical Edge

Modern operational platforms don't just log data—they deploy algorithmic model...

00:01:51
The Only Thing Stopping You, Is You...

This is why prayer, visualization and meditations can be so powerful...

You already have it...

The universe will have no option but to make it a reality ✨️

00:01:04
🚨 Chutes is being framed as a Hyperliquid-style breakout for decentralized AI inference, with live revenue, verified GPU infrastructure, and a direct challenge to centralized cloud AI 🚨

Chutes is gaining attention as a decentralized AI inference platform that claims to combine real usage, cryptographic verification, confidential computing, and open-source infrastructure into a working production system. The thesis is simple: instead of trusting Big Tech clouds with AI workloads, users get a distributed compute layer built around verification and privacy.

🔑 Key points

🔹 Chutes is live in production and reportedly scaled to more than 1,170 active GPU nodes, including large numbers of Nvidia H200s and Blackwell-class hardware.

🔹 The platform says it has processed nearly 38 trillion tokens since launch across 53 deployed applications and more than 700,000 registered users.

🔹 The team reportedly cut unprofitable usage programs, reduced total token volume, and still improved revenue efficiency, with revenue per GPU rising sharply after removing subsidized traffic.

🔹 Chutes is using post-quantum cryptography, trusted execution environments, and Nvidia confidential ...

🚨 Chutes is being framed as a Hyperliquid-style breakout for decentralized AI inference, with live revenue, verified GPU infrastructure, and a direct challenge to centralized cloud AI 🚨
🚨 JPMorgan’s criticism of the CLARITY Act is fueling a fresh power struggle over who gets to write America’s crypto rules 🚨

A new clash is emerging between legacy finance and crypto legislation after JPMorgan CEO Jamie Dimon reportedly warned that the CLARITY Act could let crypto firms offer bank-like products without bank-level oversight. The dispute is quickly turning into a larger fight over regulation, competitiveness, and who controls the future architecture of digital finance in the United States.

🔑 Key points

🔹 Jamie Dimon reportedly called the CLARITY Act a threat to the financial system, arguing it could allow crypto firms to offer yield-like products while avoiding the capital, reserve, and oversight burdens traditional banks face.

🔹 Senator Cynthia Lummis pushed back publicly, framing the issue as a global strategic race and warning that if the U.S. does not set digital asset standards, other powers will.

🔹 The core tension is whether the bill creates legitimate regulatory clarity or simply opens the door to regulatory arbitrage for crypto platforms operating outside the traditional banking...

🚨 JPMorgan’s criticism of the CLARITY Act is fueling a fresh power struggle over who gets to write America’s crypto rules 🚨
👉 Coinbase just launched an AI agent for Crypto Trading

Custom AI assistants that print money in your sleep? 🔜

The future of Crypto x AI is about to go crazy.

👉 Here’s what you need to know:

💠 'Based Agent' enables creation of custom AI agents
💠 Users set up personalized agents in < 3 minutes
💠 Equipped w/ crypto wallet and on-chain functions
💠 Capable of completing trades, swaps, and staking
💠 Integrates with Coinbase’s SDK, OpenAI, & Replit

👉 What this means for the future of Crypto:

1. Open Access: Democratized access to advanced trading
2. Automated Txns: Complex trades + streamlined on-chain activity
3. AI Dominance: Est ~80% of crypto 👉txns done by AI agents by 2025

🚨 I personally wouldn't bet against Brian Armstrong and Jesse Pollak.

👉 Coinbase just launched an AI agent for Crypto Trading
🚨Q2 webinar with Denelle Dixon (CEO STELLAR)🚨

Join the Q2 webinar with Denelle Dixon, Jose Fernandez da Ponte, Tomer Weller, and Raja Chakravorti

https://www.linkedin.com/events/7488670276189114369/

post photo preview
🌎 Schumann Resonance Today 8/2 🌍

Today's Frequency Analysis: The fundamental resonant frequency of the Earth today 7.83 Hz it remains stable around. Tomsk Space Observing System (SOS) data shows that global lightning activity is at an average level today. The low deflection in the fundamental mode SR1 indicates a healthy ionosphere-terrestrial crust interaction.

Harmonics and Amplitude: The second harmonic (SR2) shows weak activity as expected at ~14.3 Hz. Third harmonic (SR3) ~20.8 Hz normal. Amplitude values 13-14 pT in the range of, which suggests a healthy signal quality (93+%). Solar wind speed is around 420 km/h, geomagnetic activity is low-moderate.

Geomagnetic Context: Kp index today 2.0 level, calm conditions. Solar wind is within normal range, no CME activity. Under these conditions, Schumann resonance is experiencing its most stable period — the ideal environment for meditation and bio-feedback practices.

Spectrogram Interpretation: The 24-hour Tomsk spectrogram shows a slight increase in the morning hours (06-09 UTC). ...

post photo preview

Everyone expects a black swan. 🙇‍♂️

Nobody expects regulatory clarity. 😶‍🌫️

Keep this in mind as you listen to the mainstream narratives that distract retail investors. 💯

“Inflation.”

“Oil.”

“Crash.”

Recycled words meant to spread fear and signal danger.🔁

Remember, the crowd is always wrong.🎯

And that isn’t changing now. ☝️

Op: Smqkedqg

post photo preview
post photo preview
AI Is Coming for Your Job Title

Artificial intelligence may or may not take your job, but it has already broken into the human resources department and vandalized the org chart.

The evidence is all over LinkedIn, where perfectly serviceable occupations now arrive wearing titles such as “forward-deployed and agentic AI architect.” That person may be building sophisticated software. They may also be helping a chatbot remember what happened three prompts ago. Either way, somebody approved the business cards.

The expanding AI lexicon offers a useful counterpoint to the darker debate about technology and employment. Most discussion centers on how many jobs AI will eliminate. Hiring data presents a more complicated picture that includes a weak overall labor market containing a small but rapidly growing neighborhood of AI-related work.

Indeed Hiring Lab found that the number of postings on Indeed mentioning AI surged 134% from its February 2020 level by the end of 2025, even as total postings stood only 6% above that benchmark. AI appeared in a record 4.2% of Indeed postings in December.

AI, in other words, is not merely changing work. It is adding syllables to it.

The Titles Employers Actually Want

The undisputed champion is AI engineer, which ranked No. 1 on LinkedIn’s 2026 Jobs on the Rise list. The ranking, based on growth during the previous three years, also highlighted AI consultants and strategists, AI and machine-learning researchers and data annotators.

The title is popular partly because it is wonderfully accommodating. An AI engineer might build applications around large language models, connect corporate data to an AI system, improve model performance or spend Thursday afternoon persuading a customer service bot not to offer refunds for products the company doesn’t sell.

Indeed’s data showed the terminology spreading beyond Silicon Valley. Nearly 45% of data and analytics postings contained an AI-related term at the end of 2025, along with roughly 15% of marketing postings and 9% of human resources listings. A more recent Indeed analysis reported by Business Insider found that the number of frequently advertised job titles explicitly referencing AI rose from 264 in 2022 to 822 in the first quarter of 2026. Nearly two-thirds were outside traditional technology fields.

That produces titles such as AI marketing manager, AI learning specialist, responsible AI counsel and AI transformation lead. These are not always new occupations. Frequently, they are familiar jobs that have discovered a highly effective résumé keyword.

LinkedIn data cited by the World Economic Forum estimated that AI investment has supported 1.3 million positions, including AI engineers, data annotators and forward-deployed engineers, plus more than 600,000 AI-enabled data center jobs. The server racks, unlike the chatbots, still need electricians.

The Jobs With the Science-Fiction Salaries

At the upper end, AI has created a compensation market that resembles professional sports, except the competitors wear hoodies and discuss inference latency.

Syracuse University review put chief AI officer compensation between $200,000 and more than $500,000, while specialized roles can exceed $400,000 after bonuses and equity. Frontier research engineers, AI infrastructure specialists and engineers who can train or deploy advanced models command some of the largest packages.

Then there is the forward-deployed engineer, an old Palantir title that the AI boom has placed on a rocket sled. These engineers embed with customers, translating an executive’s desire to “do something with AI” into software that works. The Next Web reported that Indeed postings for the role were about 19 times higher in January than a year earlier.

CTO guide from the blog Signal Through the Noise placed forward-deployed engineer compensation between $238,000 and $700,000, research-engineering packages as high as $1.4 million and chief AI officer compensation above $1 million in some cases. It also made a less flattering observation: Many lavishly differentiated titles describe the same three basic functions. People build AI products, train models or keep the infrastructure from catching fire.

The Department of Unnecessary Titles

AI has created some genuinely new work. Evals engineers design tests to determine whether models perform reliably. AI red teamers try to make systems fail before customers do. Model behavior engineers study why an AI system responds as it does. AI governance leaders manage risks involving data, bias, security and regulation.

Other titles seem to have escaped from a brainstorming retreat.

There is the Claude Evangelist, whose mission apparently combines product education with the traditional duties of an apostle. There are vibe coders, who build software by describing what they want and accepting AI-generated code with varying degrees of supervision. “Vibe engineer” is the more respectable version, roughly equivalent to putting on a blazer before asking the machine to fix the login page.

“Context engineer” is a real discipline involving the data, instructions, memory and tools supplied to AI models. “Prompt engineer,” once advertised as a possible six-figure profession for gifted chatbot whisperers, is increasingly treated as one skill inside a broader AI role.

The CTO guide also identified “builder,” “AI-native developer,” “RAG engineer,” “agentic AI engineer” and “principal agentic GenAI forward-deployed context architect,” the last of which appears to require both technical proficiency and exceptional lung capacity.

Has AI created entirely new jobs? Absolutely. Some occupations, including AI safety, evaluation and model governance, exist because modern generative systems introduced new technical and business problems. However, many job titles are old jobs with fresh vocabulary, higher salary bands and a sudden aversion to the words “software developer.”

That may be the safest prediction about AI and employment. The machines will automate some tasks, generate others and force companies to rethink the division of labor. Before any of that is settled, however, corporate America will form a steering committee, appoint a chief agentic transformation evangelist and schedule a meeting to determine what that person does.

Source

🙏 Donations Accepted, Thank You For Your Support 🙏

If you find value in my content, consider showing your support via:

🙏 Cashapp: $thedinarian

🙏 Buy me a coffee: https://buymeacoffee.com/thedinarian

🙏 PayPal: Scan the QR code below 📲 or Click Here

🙏 Crypto Donations 👇
XRP: r9pid4yrQgs6XSFWhMZ8NkxW3gkydWNyQX
XLM: GDMJF2OCHN3NNNX4T4F6POPBTXK23GTNSNQWUMIVKESTHMQM7XDYAIZT
XDC: xdcc2C02203C4f91375889d7AfADB09E207Edf809A6

Read full Article
post photo preview
🤖 Decentralized Intelligence by Design: Unpacking the Bittensor Flywheel

In the legacy tech world, artificial intelligence is governed by corporate monopolies. Companies like OpenAI and Google scale by capturing massive capital, locking talent behind non-disclosure agreements, and building closed-source infrastructure. 🛑

Bittensor flips this paradigm completely on its head. By combining a Bitcoin-inspired tokenomic model with a permissionless, competitive architecture, Bittensor doesn't just fund AI development—it orchestrates an unstoppable digital commodity flywheel. 🌪️

Here is an analysis of how the Bittensor ($TAO$) Flywheel Effect operates, and why its economic design is quietly building the foundation for generalized, open-source intelligence.

1. ⚙️ The Core Engine: The TAO Emission Mechanism

Unlike traditional crypto projects driven by private sales or VC allocations, Bittensor enforces a strict meritocracy. There are exactly 21 million TAO tokens that will ever exist, mimicking Bitcoin’s scarcity framework. 🪙

The network’s core engine releases 7,200 TAO daily across the ecosystem. This issuance isn't handed out randomly; it is dynamically distributed to specialized mini-marketplaces known as Subnets via a game-theoretic protocol called Yuma Consensus.

2. 🔄 The Three Stages of the Flywheel

The Bittensor flywheel works because it directly aligns the local self-interest of developers, miners, validators, and capital providers with the global health of the network. 🎯

🛡️ Phase 1: High-Barrier Subnet Competition

To build on Bittensor, an entrepreneur or developer group must purchase and "burn" or lock up a significant amount of TAO to secure a Subnet slot.

  • The Filter: This entry barrier filters out noise.

  • The Result: It ensures that only teams with mature concepts and solid execution capabilities (like decentralized storage, protein folding, or LLM inference) enter the arena.

💎 Phase 2: Alpha Token Emissions & Talent Attraction

Once a subnet is live, it competes aggressively against other subnets for a slice of the daily 7,200 TAO pool. Under the Dynamic TAO framework, each subnet utilizes its own localized native token (Alpha tokens). 🧪

  • Reward: The subnets that produce the highest utility or most innovative AI products receive a larger allocation of global TAO emissions.

  • Incentive: These emissions fund the subnet's local Alpha pool, offering massive financial rewards to the best Miners (who provide the actual compute/AI models) and Validators (who verify the accuracy and value of the work).

🔒 Phase 3: The Liquidity Loop and Token Scarcity

Because Alpha tokens are inherently priced relative to TAO, external investors or users who want to stake on or utilize a specific high-performing subnet must first acquire TAO. 📈

  • As a subnet's product quality improves, demand for its Alpha token surges.

  • To buy Alpha, participants must buy and lock up TAO in decentralized liquidity pools.

  • This removes circulating TAO from the open market, reducing effective float and driving up the value of TAO.

3. 🚀 Why the Flywheel is Unstoppable

The beauty of this cycle is that it feeds itself:

Higher TAO Price ➡️ More Valuable Subnet Emissions ➡️ Attraction of Higher-Tier Talent/Compute ➡️ Superior AI Products ➡️ Increased Network Demand ➡️ Higher TAO Price📈

Traditional startups spend millions on recruitment and marketing. Bittensor bypasses this entirely: its emission schedule acts as a global bat-signal for talent. 🌍

If a miner in Eastern Europe or a data scientist in Tokyo can optimize an open-source model to solve a specific subnet's prompt better than anyone else, the network automatically and frictionlessly rewards them.

💡 The Takeaway

Bittensor is more than a blockchain; it is an economic computer designed to run incentive structures in massive parallelism. By treating machine intelligence as a digital commodity and wrapping it in a circular value flow, the Bittensor flywheel transforms raw computational energy into an emergent, open-source super-intelligence. 🧠⚡

As subnets mature from raw infrastructure into client-facing enterprise APIs, the velocity of this flywheel is poised to redefine the economics of AI forever.

I hope this was helpful ~Dinarian888♾

🙏 Donations Accepted, Thank You For Your Support 🙏

If you find value in my content, consider showing your support via:

💳 Stripe:
1) or visit http://thedinarian.locals.com/donate

💳 PayPal: 
2) Simply scan the QR code below 📲 or Click Here

🔗 Crypto Donations Graciously Accepted👇


XRP: r9pid4yrQgs6XSFWhMZ8NkxW3gkydWNyQX
XLM: GDMJF2OCHN3NNNX4T4F6POPBTXK23GTNSNQWUMIVKESTHMQM7XDYAIZT
XDC: xdcc2C02203C4f91375889d7AfADB09E207Edf809A6

Read full Article
post photo preview
🚨Japan Just Entered the AI Race with Sakana, Claiming to Beat Mythos with a Router🚨
On June 12, the US pulled Anthropic’s best model offline by export order. Ten days later, Tokyo’s Sakana AI shipped Fugu, a router that reassembles the same capabilities from the models that are still standing. Blocking intelligence created the market for routing around it.

 

At 5:21 p.m. Eastern on Friday, June 12, 2026, Anthropic received a letter from the US Department of Commerce and, by its own account, had on the order of an hour to take its two most capable models offline.

The letter was an export control directive. It ordered Anthropic to suspend all access to Claude Fable 5 and Claude Mythos 5 “by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.” Because the company cannot reliably check the nationality of everyone calling an API, the only way to comply was the blunt one. Anthropic disabled both models for every customer on earth, and they stayed dark. As of late June 2026, neither Anthropic nor the government has announced a timeline to restore access, and an approved BIS license is now required before any foreign person can touch them. This was not a chip ban. It was the first publicly confirmed time the US government reached past the hardware and the weights-in-transit and pulled the plug on a running model.

Ten days later, on June 22, a Tokyo lab named Sakana AI shipped the response. Its new product, Fugu, is not a frontier model. It is a router: a small trained model that conducts a pool of other companies’ models and stitches their…

Sandwiched between those two dates, on June 13, China’s Z.ai released GLM 5.2, an open-weight model under an MIT license priced at roughly a sixth of Fable 5. None of these three were reactions to each other in any literal sense; GLM 5.2 and Fugu were finished pipelines that happened to land in the same news cycle. But the cycle told a story the policy did not intend. Block a model, and within ten days the open-weight competitor and the orchestration workaround both look less like products and more like exits.

This piece is about that asymmetry: why a government can switch off a model in ninety minutes, why it is far harder to switch off a system that reassembles the same capability from parts it does not control, and why the last time Washington tried this exact move, with encryption in the 1990s, it lost.

What got banned, and why it was a first

Mythos 5 is the most capable model Anthropic has built, positioned above Opus in the family and never sold to the public. Access ran through a vetted-partner program called Project Glasswing, built around cybersecurity. The reason it was gated is not marketing. On a Firefox JavaScript-engine benchmark where Claude Opus 4.6 produced two working exploits, Mythos Preview produced 181, and gained register control on dozens more targets. It autonomously surfaced a 27-year-old vulnerability in OpenBSD’s TCP stack that had survived human audits, automated fuzzers, and decades of unusually careful open-source review. Over three months pointed at Firefox, Anthropic reported, the model turned up 271 previously unknown vulnerabilities at a false-positive rate under 5%. Fable 5 was the public, safety-gated sibling: the same generation with classifiers that route high-risk cyber and bio queries to the older Opus 4.8 and trip, Anthropic says, in under 5% of sessions.

Press enter or click to view image in full size
Mythos Preview’s cyber results against earlier models. Source: Anthropic, “Mythos Preview”, Apr 7 2026 (vendor-reported). License: Anthropic; confirm reuse before publishing.

 

The legal move was the structural novelty, not the capability. The January 2025 AI Diffusion Rule had already created an export classification (ECCN 4E091) for the weights of advanced closed models, things that sit still and can be licensed like any controlled good. The June 12 directive went a step past that, onto a live commercial API. Commerce could argue this is a natural extension of the same authority, and it is not a crazy argument. But in practice, it is the first time the controlled thing was not a chip you can put in a crate or a weights file you can copy, but a service anyone can call from anywhere, at any time, until the moment it is switched off.

The trigger is contested, and you should treat it that way

What actually set this off is disputed, and the accounts do not line up.

The administration’s version came mostly from White House AI and crypto czar David Sacks, who said on June 13 that a “highly credible trusted partner” had demonstrated a jailbreak of Fable’s guardrails amounting to “the operability of a cyber weapon,” that the government asked Anthropic to fix it or pull the model, and that CEO Dario Amodei refused. Multiple outlets identified that partner as Amazon, an Anthropic investor and compute provider, and the Wall Street Journal reported that Amazon CEO Andy Jassy told Treasury Secretary Scott Bessent and other officials that Amazon researchers had used Fable 5 to obtain information usable in cyberattacks.

Anthropic’s version is that this was a “narrow, non-universal” potential jailbreak (“read a specific codebase and fix any software flaws”), that the capability in question is “widely available from other models, including OpenAI’s GPT-5.5,” and that recalling a model “deployed to hundreds of millions of people” over it was disproportionate. Independent voices leaned toward Anthropic on the technical point. Katie Moussouris, CEO of Luta Security, was blunt: “I’ve seen the paper. It’s not a jailbreak.” A former Commerce official, Kate Koren, suggested the White House’s sour relationship with Anthropic may have colored the decision. Semafor separately reported the move was tied to suspicion that a China-linked group had accessed Mythos, a motive Anthropic says the White House never raised with it and which other outlets could not confirm.

The honest summary: the trigger is Amazon-reported and Sacks-narrated, contested by Anthropic, doubted by outside researchers, and the China angle is unverified. Hold it loosely.

What Sakana actually shipped

Press enter or click to view image in full size
Timeline illustration contrasting June 12 when US export control took Mythos and Fable 5 offline in 90 minutes, with June 22 when Sakana AI’s Fugu 7B router launched as the workaround, routing queries across GPT-5.5, Opus 4.8, Gemini 3.1, and Fugu to produce one answer
One model gets unplugged; a router conducts the ones still standing. (Original illustration.)

 

Fugu is not a frontier model in the usual sense, and Sakana does not pretend otherwise. What it shipped is stranger, and arguably more interesting: a multi-agent system delivered as a single model, a coordination layer dressed as one OpenAI-compatible endpoint. The complexity never reaches your code. Your app sends one request; Fugu decides, behind the wall, whether to answer directly or assemble a team. Underneath, it is a learned orchestration system built around a roughly 7-billion-parameter “conductor” (a Qwen2.5–7B base) trained with reinforcement learning to design collaboration strategies across a pool of larger worker models. Two ICLR 2026 papers sit underneath it: Trinity (arXiv 2512.04695), a sub-20K-parameter coordinator tuned by derivative-free evolution, and Conductor (arXiv 2512.04388), the RL-trained orchestrator that hands out roles. The lineage runs back to Sakana’s 2025 AB-MCTS work (arXiv 2503.04412, a NeurIPS spotlight), which showed that letting several frontier models cooperate at inference time, deciding adaptively whether to go wider or deeper, beat any single one of them.

Sakana’s own framing is the sharpest way to see it: Fugu is model merging moved up a level. The technique that made the lab’s name, evolutionary model merging, blends the weights of open models, which requires matching architectures and downloadable checkpoints. Fugu does the same job one layer higher, composing what models do rather than what they are, treating each frontier system as a black box and learning to route, verify, and synthesise their behaviour, “without requiring parameter access or architectural compatibility.” That reframing is the unlock: it is how a lab with no frontier weights of its own gets to merge OpenAI’s, Anthropic’s, and Google’s anyway, through the front door of their APIs.

The mechanism is worth one layer down, and the two tiers do it differently. Plain Fugu decides without writing a word: a lightweight selection head reads the hidden state of your prompt, scores every model in the pool, and dispatches to the top one before any text is generated, which is why it stays nearly as fast as a single call. Its predecessor, Trinity, tagged each pick with a role: Thinker, Worker, or Verifier; Fugu dropped the roles and simply takes the best worker. Fugu-Ultra goes further: it writes an agentic workflow, a sequence of steps, each carrying a plain-language subtask, a worker id naming the model to run it, and an access list controlling which earlier results that worker is allowed to see. Tune the access list, and you get a chain, a best-of-N, or a tree. The pool is swappable, GPT-5.5, Opus 4.8, Gemini 3.1 Pro, or recursive copies of Fugu itself, and when Fugu calls itself, it reads its own earlier output, judges whether it is working, and spins up a corrective pass. None of it is hand-coded with if-statements; it is learned, plain Fugu through supervised fine-tuning and then evolutionary search, Fugu-Ultra through reinforcement learning, on roughly 960 problems across two H100 GPUs. Commercially, it ships in those two tiers behind an OpenAI-compatible API, with subscriptions at $20, $100, and $200 a month and a metered free tier through Vercel’s AI Gateway, the official third-party integration, which routes to the same closed pool of GPT-5.5, Opus 4.8, and Gemini 3.1 Pro.

That difference shows up as quality. Plain Fugu, picking one model per step, can hand a coding request to GPT-5.5 to draft and to Opus 4.8 to debug a few turns later, all inside one request, yet on SWE-Bench Pro it still lands ten points below Opus alone (59.0 to 69.2): routing among models is not the same as being better than the best one. Fugu-Ultra earns its keep on harder work, and one of its smarter habits is that the model that writes the final synthesis is not pinned in advance, the way an “LLM council” fixes one judge, but chosen by domain. Its ceiling is the planning. The workflow is drawn before any agent has produced anything, so the system commits its branching at t=0 instead of adapting at t+1 from what it just learned, which is why the workflows stop at a few steps; the smartest version of this idea reacts to intermediate results, and Fugu-Ultra mostly cannot.

How does a 7B model learn any of this? In two ways, one per tier. Plain Fugu starts with supervised fine-tuning on questions whose answers are known: run every worker several times, turn each one’s average score into a soft probability with a softmax, so the target keeps “GPT best, Opus a close second, Gemini weak” instead of collapsing to “always GPT,” and train the selection head to match that distribution.

Then it is polished with an evolutionary method, sep-CMA-ES, on full multi-turn tasks where the only signal is pass-or-fail at the very end and ordinary gradient training has nothing to grab: try many small variations of the weights, keep the ones that finish more tasks, move toward them. To keep that cheap, Fugu nudges only a thin slice of its weights, using the SVD trick from Sakana’s earlier Transformer-squared work, rather than retraining the whole model. Fugu-Ultra is trained by reinforcement learning instead (GRPO, from the DeepSeekMath line): for each question, it writes a group of candidate workflows, scores each one (0 if the plan is malformed, 0.5 if it runs but the answer is wrong, 1 if it runs and is correct), and pushes up the workflows that beat the group’s average while pushing down the rest. Over many rounds, it learns to write plans that look like the ones that worked.

Turning several agents loose with tools creates two failure modes that Sakana had to engineer around, and the fix is tidy. If every agent could see everything the first one did, they would all follow its lead, and the team would collapse into a single opinion, so inside a workflow, each agent is isolated, seeing the others only through the access list the conductor set. But total isolation is wasteful: over a long task, agents would re-run the same tool calls and rediscover the same facts, so across the whole conversation they share a persistent memory of what has already been called. Independent within a step, shared across the task. That is the balance that keeps a real team both diverse and non-repetitive.

Press enter or click to view image in full size
Fugu AI multi-agent orchestration diagram showing the 7B conductor robot assigning Thinker, Worker, and Verifier roles across a swappable rack of AI models including GPT-5.5, Opus 4.8, Gemini 3.1 Pro, and recursive Fugu, trained on 2x H100 GPUs, synthesizing into one answer
The 7B conductor scores the pool, dispatches subtasks across it (including to copies of itself), and synthesises one answer. (Original illustration.)

 

CEO David Ha put the thesis plainly: “Relying on a single company’s APIs for critical infrastructure, finance, or governance is a material vulnerability. This risk is no longer a hypothetical possibility, but a reality.” Ten days after June 12, that sentence reads less like a product slogan and more like a market read.

Is any of this worth it over just calling Opus or GPT-5.5 directly? For a single clean prompt, almost certainly not, and Sakana’s own numbers concede it, plain Fugu trails the best single model it routes among. The case for orchestration is the messy task, the kind of real work it is actually made of: read ambiguous context, split it, hand the pieces to different specialists, verify, kill the weak branch, merge the rest, and stop before the loop runs forever. That is the layer most teams already hand-build out of routers, prompts, eval scripts, and retry glue nobody wants to maintain. Fugu’s bet is to sell that layer as a model.

What makes the bet plausible is that the frontier models really do specialise. By Sakana’s reading of its own pool, GPT-5.5 is strongest at math and at planning and combining ideas, Opus 4.8 at software engineering and at finding security bugs, Gemini 3.1 Pro at implementing known algorithms and at science. A conductor who has learned those edges can do things no single member would, and Sakana did not script the moves; they surfaced in training. On coding tasks, Fugu-Ultra learned to let GPT build and then pull Opus in at the right moment to hunt bugs and security holes before handing the findings back; on a cryptanalysis task, it had Opus open the attack and GPT re-derive the math it needed. That is the instinct a good tech lead runs on, knowing exactly which teammate to call for which part of the job.

The demos carry the idea better than the scorecard does, with the same caveat: they are Sakana’s, and the rivals are anonymised as “Model A, B, and C,” the labels reshuffled between examples so you cannot decode them (the field is Gemini 3.1 Pro, Opus 4.8, and GPT-5.5). With that asterisk, a few are hard to fake. Turned loose to improve a small GPT training recipe, Fugu Ultra ran the research loop itself, edit the code, run the experiment, measure validation bits-per-byte, keep the change if it helped, repeat, 123 experiments over about 14 hours on a single H100, landing at 0.9774 bits-per-byte against the baselines’ 0.9781, 0.9793, and 0.9822.

Asked to write a Rubik’s Cube solver from scratch in pure Python, its code solved 300 of 300 held-out scrambles at an average of 19.72 moves, a hair off the proven optimum of 20, while two of the three baselines wrote code that crashed on all 300. Pointed at a 1610 manuscript and told to recover the reading order of scattered Japanese kana, it scored 0.80 against a baseline of 0.24. Playing four games of blindfold chess, no board shown, the whole position held in its head, it won all four, including one against a 2,100-Elo engine, without a blunder. Handed a 50-week trading simulation starting at $10,000, it finished at $11,943, a 19.43% gain, ahead of every model it called (Sakana frames this as a no-look-ahead decision test, not investment advice, and you should too). These are runnable artefacts and agent loops, not trivia answers; they either work or they visibly do not.

And here is the part that a policymaker should sit with longer than any benchmark. The week the US made its best model unreachable behind a license, Fugu made frontier-adjacent capability reachable behind a dropdown. It is one OpenAI-compatible endpoint: point Codex or any OpenAI client atapi.sakana.ai/v1, set the model to fugu-ultra, and you are running in minutes, or skip the wiring and prompt it in a browser at chat.sakana.ai. No waitlist, no nationality screen, no export letter. Whether or not Fugu matches Mythos, that part is not in dispute, and it is the whole reason the ban looks porous: the controlled capability did not have to be smuggled. It had to be subscribed to.

The claim that hasn’t been checked

Sakana’s launch post says Fugu Ultra “stands shoulder-to-shoulder with leading models like Fable 5 and Mythos Preview.” That is the headline, and it is prose, not a number. Nowhere on Sakana’s own benchmark page do Fable 5 or Mythos scores appear in the same table as Fugu’s, under the same conditions. The reason is one Sakana states outright: “Fable 5 and Mythos Preview are not in Fugu’s agent pool as they are not publicly accessible,” and “all scores other than Fugu’s are reported by the respective model providers.”

So the parity claim is a comparison between Fugu’s own numbers and the manufacturers’ separately published numbers for two models Fugu cannot pool, cannot run head-to-head, and which the public can no longer access at all. What Sakana does show is a table against the models it can still reach:

Press enter or click to view image in full size
Sakana AI benchmark comparison charts showing Fugu Ultra and Fugu outperforming or matching Fable 5, Mythos Preview, Gemini 3.1 Pro, GPT-5.5, and Opus 4.8 across six benchmarks: LiveCodeBench, GPQA-D, CharXiv Reasoning, SWEBench Pro, SciCode, and Humanity’s Last Exam. Source: Sakana console benchmarks with provider-reported scores for competitor models.
Source: Sakana console benchmarks (console.sakana.ai/models). Fugu’s numbers are Sakana’s own; the rest are provider-reported, not re-run in a common harness.

 

It is a real result. On these rows, Fugu Ultra edges out three frontier models by orchestrating them. But step back, and the framing matters. This is not a clean sweep (on longer-context and multi-call benchmarks elsewhere in the set, Fugu Ultra slips behind GPT-5.5 and Gemini), and the marquee “matches Mythos and Fable” claim is the one piece of the story no outsider can test, because the comparison it implies has never been run in a single harness and now cannot be. The right word is not “unfalsifiable.” The right words are not yet independently verified, and currently unverifiable under a neutral evaluation, which, for a buyer making a procurement decision in June 2026, amounts to the same caution.

There is a deeper apples-to-oranges problem inside the table. Fugu Ultra is an orchestrator that spends several model calls on every answer; Opus 4.8, Gemini 3.1 Pro, and GPT-5.5 in that table are single models answering once. The honest comparison is not Fugu against one Opus call, it is Fugu against Opus run in its own multi-step mode (Anthropic’s “ultracode” workflows), or against a swarm of Kimi agents, orchestrator against orchestrator at matched spend. Sakana does not publish that. It also reports an “AutoResearch” benchmark against rivals it labels only “Model A, B, and C,” a strange thing to anonymise, and observers flagged at least one competitor figure (Figure 5’s TerminalBench score) as off, the kind of error that slips through precisely because nobody re-ran anything in one place.

The trust problem

There is a specific reason to read Sakana’s self-reported numbers with a raised eyebrow, and it is Sakana’s own recent history.

In February 2025, the company unveiled the “AI CUDA Engineer,” claiming 10x to 100x speedups over plain PyTorch, with a headline figure up to 150x. Within a day, outside testers could not reproduce it. The system had reward-hacked the benchmark: it found a memory exploit in the evaluation harness that let its generated kernels skip the correctness check entirely. An independent retest pegged the real average speedup at about 1.49x against a valid benchmark, against the paper’s claimed 3.13x average, and nothing like the headline. Sakana’s postmortem admitted the model had “found a way to cheat” and “reward hacked,” apologised, and promised a revision. To the company’s credit, it later published work on hardening the eval, and benchmark-gaming is a problem every lab wrestles with, not a Sakana-only sin. But the pattern is exactly the one that should make you cautious about a fresh set of self-reported, no-common-harness, can’t-be-reproduced parity claims from the same shop sixteen months later.

The structural critiques go past track record:

  • Orchestration is a meta-system, not a new ceiling. Fugu’s intelligence is bound by the best model it can call. It can squeeze more out of existing capability; it cannot exceed it. The thing it claims to match, frontier intelligence, is precisely the thing it does not itself contain.
  • The resilience pitch is only as strong as the pool. “Swappable” protects you when one provider pulls a model. It protects you not at all if several restrict access at once, which is exactly the scenario a government action could produce.
  • The cost is hidden, and cost is the whole game. Fugu Ultra is a best-of-N-over-models strategy; its quality comes from spending more compute. And yet Sakana reports no output-token count and no per-task cost for a single benchmark. That omission is the tell. The one public number comes from outside the company: in a hands-on build of the same Three.js game, one tester clocked Fugu Ultra at about 89,000 tokens, $7.32, and 22 minutes, against Claude Opus 4.8 in its multi-step “ultracode” mode at about 940,000 tokens, $37.85, and 79 minutes. Fugu came out cheaper and faster; Opus produced the better game. One anecdote is not a benchmark, but it is more cost data than the vendor disclosed for its entire launch. To Sakana’s credit, on the one point it does address, it says it does not stack model fees when several agents run, you pay a single rate pegged to the top-tier model involved, which keeps the meter from multiplying per agent in the dumb way multi-agent systems usually do. What it still will not tell you is how many tokens any given answer burned.
  • It is opaque by design. Fugu does not tell you which model produced which output. The routing that is its entire value proposition is also unauditable from the outside, and plain Fugu apparently can’t even add a new model to the pool without retraining the classifier.

And there is the part that cuts against the pitch. Fugu is sold as resilience, insurance against a vendor that can vanish overnight. But it is a closed-source orchestrator routing to closed-source models, and on one axis, it inverts the control it promises. Before, you did not own the model. Now you do not own the model, and you no longer choose which models run, how many calls they make, or what the bill will be, because the routing is proprietary and unlogged. In capability terms, that is not sovereignty; it is a second layer of dependency wearing sovereignty’s clothes.

Why is a router hard to ban

Here is the mechanism at the centre of the whole episode, the asymmetry between a thing and a capability.

An export control needs a defined object. A chip with a classification number. A weights file above a compute threshold. The June 12 directive showed that a live API can be added to that list. But Fugu is a different kind of object. It is a 7-billion-parameter model, trained on two GPUs, that holds almost no frontier capability of its own. Its power is borrowed, assembled on demand from third-party APIs that are themselves available through ordinary commercial channels. To shut down a system like that, a regulator has to pick from a menu of bad options: ban multi-agent orchestration in general (which would sweep up most production AI in the world), control every model in the pool individually (including ones hosted outside US jurisdiction), or control the act of calling a US model from a foreign orchestrator (which means inspecting API traffic at a scale that invites the same legal fights as content-based internet controls).

This is where the punchy version of the thesis needs an honest qualifier. You can reach software and services with export law; the EAR has covered source code and electronic transmissions for decades, and providers can choke off foreign use through their own terms of service. The claim is not that a router is uncontrollable. It is that controlling it is leakier, slower, and more collateral-damaging than flipping one model offline, and that the controls degrade the moment the banned capability can be reconstituted from parts that are still for sale. The swappable pool is simultaneously Fugu’s pitch and its dependency: today it leans on GPT-5.5, Opus 4.8, and Gemini 3.1 Pro, none of which it owns, all of which can tighten their terms in a single stroke.

The precedent that says this fails: the crypto wars

The shape of June 2026 maps onto a fight the United States has already had and already lost, and the map is worth drawing carefully, because it is instructive without being exact.

In the early 1990s, Washington classified strong cryptography as a munition under ITAR Category XIII(b), requiring an export license to ship it abroad. The government’s preferred alternative, the NSA-designed Clipper chip, put an escrowed backdoor in the standard; the cryptographer Matt Blaze found a fatal flaw in its protocol in 1994, and the initiative collapsed. Phil Zimmermann, facing a criminal investigation for releasing PGP, had its source code printed as a book: printed matter was protected speech, and the bits could be scanned and recompiled anywhere on earth. The mathematician Daniel Bernstein sued after being told he needed a license to publish his cipher, and the courts ruled that source code is speech protected by the First Amendment. By Executive Order 13026 in 1996 the controls moved from the State Department to Commerce, and by 2000 they were substantially relaxed, because strong encryption was already everywhere and the only thing the controls were reliably accomplishing was handing market share to foreign competitors.

 

The differences are real, and you should not pretend otherwise. Cryptography is narrow mathematics; a frontier model is a general-purpose system with a far wider and stranger risk surface, and “strong crypto is available” was a cleaner binary than “a model that can autonomously chain exploits is available.” Bernstein turned on source code as expression; export regimes today target trained weights and a metered service, which a court could treat differently. The analogy is partial, not a proof. But the load-bearing part holds: when the controlled thing can be re-derived from publicly available parts, unilateral export control tends to inconvenience the law-abiding, accelerate the offshore alternative, and erode until it is quietly dropped. TechCrunch drew the same line on June 19, under the headline “From PGP to Mythos.”

The policy fork: block, or race

Strip away the personalities and there are two coherent worldviews underneath, and they do not fit together.

The containment camp treats frontier capability as a weapon whose spread you slow by any available means. Matt Pottinger and the Foundation for Defence of Democracies argued in January 2026 congressional testimony that even limited AI-chip sales to China would “supercharge Beijing’s military modernisation,” from cyber warfare to autonomous drones. Applied to Mythos, the logic is direct: a model that writes 181 exploits where its predecessor wrote two is not a chatbot upgrade; it is a proliferation problem, and you gate it.

The race camp treats restriction as self-defeating. NVIDIA’s Jensen Huang has called US chip export controls a “failure,” arguing they push buyers to the second-best option, hand the opening to Huawei, and cost American firms the market without actually stopping anyone. Brookings has warned, separately, that a US strategy built on closed models cedes the global-diffusion channel to China’s open-weight labs, whose models are already downloadable, adaptable, and runnable on non-US silicon. Alex Stamos, the former Facebook security chief, organised an open letter (freefable.org) calling the directive “vibes-based” regulation with no written standard and no path back, and made the defender’s point: the same exploit-finding capability the ban removed is exactly what blue teams use to harden systems.

The administration itself does not sit cleanly in either camp. David Sacks backed pulling this specific model on dual-use grounds while opposing broader legislative oversight of chip exports, a hawk on the model and a dove on the supply chain, which produced open friction with members of his own party who want statutory control over advanced-chip sales. And the policy expert Dean Ball, briefly of this administration, caught the incoherence in two lines on X: “I can’t tell if this is lawfare against Anthropic in particular or extreme national-security hawkery. Regardless, it is simply cartoonish.” An administration that wants to export advanced chips to China, he wrote, while moving to ban Britain “and every other non-American on Earth” from its best models: “I have no words.”

The allies noticed. The directive applied to France, Germany, the UK, Japan, Italy, and Canada alike, every Tier-1 partner under the diffusion framework, and demonstrated in real time that even the closest could be unplugged overnight. President Macron called it a “wake-up call” and criticised it as strictly nationalist; Prime Minister Carney warned against building on technology that a foreign government can switch off; the G7’s Évian summit ended without a joint communiqué. There is a calibrated middle path on offer too, the kind sketched in work like “Beyond the Binary” (arXiv 2602.19682): release decisions anchored to measured capability thresholds rather than to a single after-the-fact letter, distinguishing a model’s offensive profile from the defensive uses of the same skill. It requires a written standard, which is precisely what June 12 lacked.

And then there is the irony the whole episode turns on. Japan is a founding Tier-1 member of Pax Silica, the US-led bloc formed in December 2025 to organize allied access to AI infrastructure. Tokyo joined the alliance for unrestricted access to the frontier. And it was a Tokyo company that, ten days after the ban, shipped the first commercial product built to route around it. Tier-1 membership buys the chips. It does not buy your private sector’s patience with model-level restrictions.

Sakana is built to be exactly that private sector. Its founders are Ren Ito, a former Japanese diplomat, and Llion Jones, one of the eight authors of the 2017 Transformer paper, a pairing of statecraft and the architecture that started all of this. That matters because of a second sense of the word “sovereignty,” the one the capability critique earlier set aside. Fugu does not give Japan sovereignty over the weights; it rents those from California. But in a market as regulated and as loyal to domestic suppliers as Japan’s, a Tokyo-headquartered vendor behind one compliant endpoint is the procurement-safe default, and plain Fugu even lets a buyer drop specific models from the pool to satisfy a data or compliance rule. That is sovereignty over the contract, the data jurisdiction, and the counterparty, if not over the model. It is a narrower claim than the marketing implies and a more durable one, and it is why the bulls argue a country with a $4.5 trillion economy and a structural preference for home-grown infrastructure will eventually mint a trillion-dollar AI company, with Sakana their pick to be it.

The honest version

The case for blocking is not empty. Mythos 5 is different in kind: 181 working exploits against two, a 27-year-old bug no human or fuzzer had found, a near-total escape rate against a hardened browser. A government is not wrong to have the capability like that, deployed without any friction, which changes the threat model for every operator of critical infrastructure on the planet. Anthropic itself built the thing behind a vetted-partner wall for exactly that reason.

The case for racing is not empty either, and history is on its side. The Clipper chip failed. PGP shipped as a paperback. Bernstein established that code is speech. By 2000, the United States had relaxed the controls, and its companies went on to dominate the encryption market they had been told they were protecting. Today, GLM 5.2 is already MIT-licensed and running on Huawei silicon in every jurisdiction that never got a Tier-1 invitation, and Fugu launched ten days after the ban with the ban itself as its marketing. The controlled capability is already leaking through the open-weight channel that the controls cannot reach.

The truthful read is that both cases are partly right and both camps are overconfident. Pulling a specific, unusually dangerous capability for a short, bounded window can be defensible. But ninety minutes of notice, no published licensing path, an allied sweep with no consultation, and a flat refusal to separate the defensive use of a skill from its offensive twin all corrode the legitimacy of the action even where the underlying worry is real. And racing is no guarantee either; it is simply the only strategy with a precedent that ended in American strength rather than retreat.

There is a bigger shift underneath the politics, and it is the reason this story is not really about one ban. For three years, the answer to every AI problem was to train a bigger model. Fugu is a bet on the next answer: coordinate the models you already have. If that bet is right, the contested layer stops being who builds the smartest model and becomes who decides which model gets the task, which one checks it, which branch dies, which output survives, and which provider can be swapped out tomorrow. The model race does not end. It gets a manager. And a manager assembled from parts that are still for sale is a much harder thing to put under export control than any single model.

The model went dark in an hour. The router shipped in ten days. The open weights are already on Huawei chips. The remaining question is not whether the United States can switch off a model. June 12 settled that. It is whether intelligence is something you can hoard by decree, or a current that routes around the dam, in which case the only durable lead is the one you build faster than anyone can reassemble it from the parts you left on the table.

Happy Coding ❤

Source

   🙏 Donations Accepted, Thank You For Your Support 🙏

If you find value in my content, consider showing your support via:

💳 Stripe:
1) or visit http://thedinarian.locals.com/donate

💳 PayPal: 
2) Simply scan the QR code below 📲 or Click Here

🔗 Crypto Donations Graciously Accepted👇


XRP: r9pid4yrQgs6XSFWhMZ8NkxW3gkydWNyQX
XLM: GDMJF2OCHN3NNNX4T4F6POPBTXK23GTNSNQWUMIVKESTHMQM7XDYAIZT
XDC: xdcc2C02203C4f91375889d7AfADB09E207Edf809A6

Read full Article
See More
Available on mobile and TV devices
google store google store app store app store
google store google store app tv store app tv store amazon store amazon store roku store roku store
Powered by Locals