Category Archives: History

Blonde Greek Gods for Kids: Playmobil’s Nazi Aesthetic Explained

A Bavarian toy company called Playmobil is selling blonde haired Greek gods to children through a charity, and nobody seems to find the Nazism origin story strange?

Headquartered in Zirndorf the company launched its “History” line of Greek mythology figures in 2016. The series now includes all twelve Olympian gods, Alexander the Great, Achilles, the Argonauts, and Heracles.

Hermes – 9524: blonde.

Artemis – 9525: blonde.

Alexander – 70950: blonde.

Apollo – 70218: blonde.

Hera – 70214: blonde

Demeter – 9526: blonde

Aphrodite – 70213: blonde

The entire Greek pantheon are presented as Nordic blondes. Originally, the figures were produced exclusively for the Greek market, sold at the Acropolis Museum gift shop in Athens and Greek toy stores, allegedly with proceeds going to the Orama Elpidas (Vision of Hope) bone marrow donor association.

In case you don’t recognize that twist, there’s a known tradition of groups who weaponize charity. Their atrocities become harder to criticize. It’s the foot in the door to normalize harms.

That’s how this disinformation campaign pivoted from its start in Greece back to Germany, even though Germans should know better. The Pergamon Museum, for example, as a monument to German archaeological extraction from Turkey, Iraq, and the ancient Near East sells the blonde Greek gods alongside displays of the looted friezes and gates.

It’s an appropriation loop closed tight.

The architecture doesn’t change

The Nazis were addicts of such appropriation. They did not invent a “fair Apollo” trope, but they forever weaponized the existing ones. Classical studies of “whitened” antiquity became central and integrated into ideological education set forth by Hitler.

Alfred Rosenberg, the Nazi regime’s chief ideological theorist, laid out the “blonde blueprint” in The Myth of the 20th Century (1930).

Playmobil perhaps knows: the Greek gods were appropriated and reframed to “represent the Aryan white ideal of their race.” Apollo symbolized Nordic virtues of “pure thought and rationality”. The fabricated “true Greeks” were the Achaeans and Dorians, who Rosenberg claimed were “of Nordic-Germanic descent” and “the ones who created the magnificent Greek culture.” He went so far as to say darker-skinned Mediterranean populations were, in his racist hierarchy, contaminated by “Pelasgian and Semitic origins.”

Rosenberg rewrote Heracles and Jason as “representations of the ideal Aryan Nordic masculine type” no longer Greek. The Nazi tribalist Gunther classified Plato himself as “a pure-blooded aspect of the northern blood of primitive Hellenism.” Friedrich Hildebrandt declared Plato “a teacher for our time” whose ideal state prefigured the racial hierarchy of National Socialism, completely ignoring Plato’s actual work.

This was an extremist racist fringe becoming state doctrine, taught in German schools, promoted by the regime’s most toxic intellectuals, and backed by the full apparatus of the Third Reich’s cultural production machine for genocide.

Did this German effect on classical studies experience denazification after Hitler’s suicide? No. While racial theory collapsed, the discipline infrastructure remained intact. Those who had served dutifully to enact genocide remained in position, continued teaching and curating the “blonde blueprint”.

Denazification screened for party membership cards, not for intellectual content. Steven Remy’s The Heidelberg Myth (Harvard, 2002) documents how it actually worked: Karl Heinrich Bauer, an outspoken supporter of Nazi sterilization law, became Heidelberg’s first postwar rector — then used the position to reinstate compromised colleagues. The university remained dominated by former Nazis throughout the 1950s. Professors constructed what Remy calls “elaborate narratives of defense and justification” to absolve themselves, while continuing the same research programs, the same curricula, the same visual conventions.

At Göttingen, many professors simply lied on their questionnaires. A 2025 review in the Bryn Mawr Classical Review concluded that the history of humanities at German universities under Nazism “was not a deviation from its true mission, but a continuation of traditions that had existed before.”

The traditions continued after, too.

Albert Norden’s 1965 Braunbuch documented 1,800 former Nazis still holding high-ranking positions across West German state, economy, administration, army, justice, and science.

Bettina Arnold’s research on Nazi archaeology documents that instructional posters produced under the Third Reich, depicting ancient peoples as tall, muscled, with blond hair and light eyes, continued to be used in Germany after 1945. The blonde Greeks weren’t flagged because they weren’t swastikas. They were just how the discipline depicted antiquity.

The Fragebogen asked who you were in the party. It never asked if you were teaching heavily corrupted classics.

What evidence actually shows

The proof of Nazis being wrong isn’t necessary, given their self-described supremacy over facts, but we do it anyway.

A 2017 study on the genetic origins of Minoans and Mycenaeans found both populations shared roughly 75% of their ancestry with early Neolithic farmers from Anatolia, with additional ties to the Caucasus and Iran. The Mycenaeans had a small northern European genetic contribution of around 10-15%, so their genetic makeup remained predominantly Mediterranean and Near Eastern.

Ancient Greek art consistently shows this. Mycenaean frescoes, Greek pottery, mosaics, and sculptures overwhelmingly depict figures with dark hair and dark eyes. Minoan men at Knossos appear with long dark hair and reddish-brown skin. In addition, modern research has shown Greek statues were vividly painted and not white.

Homer, just one of many storytellers, used “ξανθὸς Ἀχιλλεύς (“xanthos Achilles”). He was marking the man as exceptional and shocking. In other words, he’s saying Achilles is radiant as in light, heroically visible and physically striking. Calling something light therefore blond, is like calling a candle flame or the sun yellow. It’s not actually how anything works. Light isn’t yellow, it’s luminous across a rainbow. Homer was describing something in the way it draws attention because it catches light, something radiant. Imagine writing “he was brilliant” and having a toy designer make him blond because… brilliant means blonde. Nope.

Several Greek gods were explicitly described as dark-haired in ancient sources. Poseidon, Hades, and Dionysus were called kyanokhaitis — “dark-haired,” with the “kyan” referring to a blue-black color. Poseidon’s dark hair correlated with the sea. These weren’t peripheral figures.

Even the Antiquipop academic review of Playmobil’s Greek sets noted diplomatically that “the skin colour of the characters remain quite uniform and white for Mediterranean people.”

They could have said Hitler directly influenced their interpretation. Nazism radicalized embedded Greek antiquity and Germany has retained this radicalism as an aesthetic convention.

Mass-market products continue flattening Greek history through Northern European mysticism.

The transaction

Period / Source Aesthetic Signal Ideology / Notes
18th Century: Johann Joachim Winckelmann White marble, “noble simplicity” Emerging template of heroic “light features” in classical antiquity
19th Century: Neoclassical painters & sculptors Golden/blonde hair on Greek/Roman figures Visual grammar spreads across European art and pedagogy
1930s-1945: Alfred Rosenberg / Nazi theorists Blonde, blue-eyed Greek heroes Explicit racialization: “True Greeks = Aryans”; ideological education & propaganda
Post-1945: German textbooks & museum displays Blonde classical heroes Nazi ideology purged, aesthetic remains embedded in pedagogy and museums
2016–2026: Playmobil Greek mythology figures Blonde Olympians & heroes Commercialized Nazi aesthetic; packaged as charity-friendly; displaces reality

Rosenberg’s racial theory for Hitler stripped of its explicit ideology, laundered through ninety years of “that’s just what heroes look like” visual grammar — sits on a shelf next to the cash register at the Pergamon Museum.

Nobody at Playmobil’s Zirndorf headquarters needed to read The Myth of the 20th Century. They didn’t have to. The Nazi aesthetic was so thoroughly untouched in German cultural production that blonde-equals-heroic just feels natural. Unremarkable. Default. That’s how successful cultural appropriation works: it makes itself invisible.

The charity shield

The “Play & Give” branding is worth examining as a structure, not just a partnership. Playmobil Hellas has donated over €80,000 to the Orama Elpidas association through this program. The cause is real. Children with cancer need bone marrow donors.

But wrapping racially coded cultural products in charity branding creates a shield that deflects criticism. Who wants to be the person attacking a toy that helps kids with cancer? The structure is elegant: you can’t critique the product without appearing to critique the cause.

This is a pattern that repeats across institutional capture. Wrap the extraction in something nobody can oppose. Philanthropy as packaging. The charity is genuine; the deflection is structural.

What sits on the shelf

It’s 2026. A German toy company is selling blonde Greek gods at a building full of artifacts Germany removed from the lands where those gods were worshipped, and the packaging says it’s for charity.

The Playmobil figures aren’t just bizarrely blonde “for fun”; they exist inside a long, persistent visual grammar that Nazis weaponized, and which survived decades of postwar German cultural production.

The racial theory that made Mediterranean gods look Northern European is ninety-six years old, its author was hanged at Nuremberg, and his aesthetic choices are still moving product.

The product has changed. The architecture hasn’t.

Apple Under FTC Attack From “The Warm-Hearted Protector of the American Press”

FTC Chairman Ferguson’s “warning letter” to Apple CEO Tim Cook on February 12, 2026, threatening enforcement action over Apple News allegedly boosting “left-wing sources,” is state coercion of editorial decisions.

It follows a documented historical pattern with precision.

Ferguson sent an almost identical letter to Alphabet CEO Sundar Pichai in August 2025, accusing Gmail of “partisan filtering” in its spam detection.

Same three-pronged legal threat, same “consumers’ reasonable expectations” language, same implicit enforcement action. He’s clearly pumping a template. Each letter builds momentum for the next target.

The three-pronged fraud Ferguson engages in has been designed to be impossible to satisfy:

  • Prong one says curation “inconsistent with terms of service” violates the FTC Act. But every platform’s ToS grants broad editorial discretion — so this prong only activates when Ferguson decides the discretion was exercised in the wrong ideological direction.
  • Prong two says curation violating “consumers’ reasonable expectations” is a material omission. But “reasonable expectations” of ideological balance in a news aggregator is completely undefined. Ferguson defines what balance looks like. By assertion.
  • Prong three says curation causing “substantial injury” violates the Act. But the alleged “injury” is that consumers received news from sources Ferguson considers left-wing. That’s editorial disagreement reframed as regulatory violation.

This is the state telling a private news aggregator which sources to elevate and which to suppress, under nothing other than his personal threat of federal investigation.

It’s Un-Constitutional

The Supreme Court ruled unanimously in NRA v. Vullo (2024) that “government officials cannot attempt to coerce private parties in order to punish or suppress views that the government disfavors.” The test: whether conduct “viewed in context, could be reasonably understood to convey a threat of adverse government action in order to punish or suppress speech.” Ferguson’s letter meets every element. He holds enforcement authority over Apple. The letter explicitly references potential enforcement action. And it targets editorial curation decisions based on their perceived ideological orientation.

The irony, oh the irony.

The entire Murthy v. Missouri case was brought by Republicans arguing that the Biden administration’s communications with social media companies about misinformation constituted unconstitutional government coercion of private speech. Justice Alito warned in dissent that such pressure campaigns could “stand as an attractive model for future officials who want to control what the people say, hear, and think.” Ferguson is now executing exactly what Alito described — using regulatory authority to pressure platforms into changing editorial decisions based on ideological grievance — except he’s doing it with signed letters on FTC letterhead, which is considerably more explicit than anything alleged in Murthy.

What a Historian Knows

Ferguson’s approach maps precisely onto the early phases of Nazi press consolidation.

The mechanism is not analogy.

This is the structural repetition of Nazism.

Mechanism Nazi Germany (1933-1935) FTC Under Ferguson (2025-2026)
Framing Goebbels introduced the Schriftleitergesetz calling himself the “warm-hearted protector of the German press” Ferguson frames letters as “consumer protection” — reminding companies of “obligations to customers”
Impossible standard Editors required to omit anything that “tends to weaken the strength of the German Reich” or is “misleading to the public by mixing selfish interests with community interests” — undefined, subjectively enforced Platforms must not violate “consumers’ reasonable expectations” of ideological balance or cause “substantial injury” — undefined, subjectively enforced
Regulatory chokepoint Journalists required to register with the Reich Press Chamber to work; the Propaganda Ministry could remove any editor “for pressing reasons of public welfare” Platforms operate under FTC jurisdiction; Ferguson can initiate investigation and enforcement action against any company whose curation he deems ideologically imbalanced
Economic coercion Amann Ordinances (1935) imposed ownership requirements publishers couldn’t meet, forcing sales to the Nazi publishing combine at distressed prices Threat of FTC investigation imposes compliance costs (discovery, legal fees, stock impact) that make algorithmic adjustment cheaper than resistance
Editorial control mechanism Twice-daily Tagesparole from the Propaganda Ministry specified content “down to the headlines and the required epithets” Warning letters specify which editorial outcomes are acceptable; compliance is verified by whether curation outcomes match the regime’s ideological preferences
Personnel capture Editors became de facto civil servants “directly subordinate to the Ministry of Propaganda” rather than their publishers FTC Director of Public Affairs is Joe Simonson, hired directly from the Washington Free Beacon; Bureau of Competition director chosen to target companies that “censor conservative voices”
Escalation sequence 1933: Editor’s Law (content control) → 1935: Amann Ordinances (ownership control) → ongoing acquisitions of non-compliant publishers Aug 2025: Gmail spam filtering letter → Feb 2026: Apple News curation letter → next: search ranking? social media feeds?
Projection Regime accused opposition press of being “Jewish-controlled” and acting against the national interest while systematically capturing the press for party purposes Administration accuses platforms of “censoring conservative voices” while using regulatory authority to coerce platforms into suppressing non-conservative editorial judgment
Compliance effect Hitler complained: “It is no great pleasure to read 15 newspapers all having nearly the same textual content” If platforms adjust algorithms to satisfy Ferguson’s undefined “balance” standard, the result is identical: editorial homogeneity dictated by state preference
The quiet part Amann testified at Nuremberg: “the basic purpose of the Nazi press program was to eliminate all press in opposition to the party” Ferguson’s own pre-appointment materials cite the “removal and demonetization of users who challenge the Silicon Valley political consensus” as a core grievance to be addressed through FTC power

The critical structural parallel is not the end state. We know ICE has been bidding to convert huge warehouses into concentration camps for housing political prisoners, which presumably includes journalists. Already Don Lemon was very publicly attacked and arrested for doing what journalists do.

The parallel is the mechanism: a government official with enforcement authority using vaguely defined legal standards to pressure private editorial operations into compliance with ideological preferences, where the cost of resistance (investigation, enforcement action and potential jail) exceeds the cost of compliance (adjust the algorithm).

Even before the Schriftleitergesetz was formally enacted, German journalists had already become dependent on the regime’s goodwill. The formal law came after the informal pressure had already done its work.

Ferguson’s letters operate the same. He doesn’t need to follow through on enforcement. Apple’s lawyers are meant to calculate the cost of an FTC investigation driven by Trump loyalty and adjust the algorithm. The letter is the initial enforcement.

Trump Asset Architecture

Ferguson clerked for Clarence Thomas. He served as chief counsel to Mitch McConnell, functioning as “judicial confirmation strategist.” He was Republican counsel on the Senate Judiciary Committee. His entire professional life has been spent in the infrastructure of conservative legal movement-building.

His stated purpose, documented in his own pre-appointment materials, is using regulatory power to control perceived ideological imbalances in private media.

His communications director came from the Washington Free Beacon. His Bureau of Competition director was selected specifically to target companies accused of suppressing conservative speech. The FTC’s enforcement apparatus has been staffed for a political mission.

This is what competent complicity looks like at institutional scale: skilled legal professionals who understand exactly what they’re doing, executing a program of editorial coercion through regulatory authority, using consumer protection language as cover for viewpoint-based government pressure on private publishers.

The Weimar Republic had a free press. The architecture of its elimination was not secret. It was documented, prosecuted at Nuremberg, and taught in every competent history program for eighty years.

The Nazi warning signs were the warning letters. So too with Trump.

The technology changes. The architecture doesn’t.

Anthropic AI Safety Lead Can’t Hack It: Resigns to Get a Poetry Badge Instead

An Anthropic safety researcher loudly and publicly resigned this week with an alarmist yet vague letter warning that the world is “in peril” from “interconnected crises.”

Let me break down why this happened. It’s not a mystery.

Mrinank Sharma said he had “repeatedly seen how hard it is to truly let our values govern our actions” at Anthropic.

Water is wet. News at 11.

He offered zero specifics about any of it, declined comment to Forbes, and announced he may pursue a poetry degree to “devote myself to the practice of courageous speech.”

I’ve achieved what I wanted to here… I arrived in San Francisco two years ago, having wrapped up my PhD and wanting to contribute to AI safety.

Dude.

You stepped up to bat and then stepped down to announce you’d like to learn how to step up to bat?

Sharma ended his farewell by citing William Stafford’s poem “The Way It Is,” about holding a thread others can’t see. Then he announced he plans to “let myself become invisible for a period of time” and “get away from the structures that have held me.” A man who held a thread of safety concerns nobody could see, took pay for holding it, refused to show it to anyone on his way out, and then announced he’s going invisible. That poem is a confession.

And to be fair this isn’t actually about Sharma, although he gives us the headlines today and we unfortunately can’t leave him out. He seems like a highly successful researcher who rose up the ranks to do what he was trained to do. The problem is what Anthropic trained him on, and what this company calls “safety” let alone its “constitution“.

Sharma led Anthropic’s Safeguards Research Team. He has an Oxford PhD in machine learning, which is admittedly very far from the seasoned steps of an actual security professional. His team studied whether Claude would help users do bad things like assist with bioterrorism, flatter users into distorted realities and that sort of academic thought exercise. His last published study found “thousands” of reality-distorting chatbot interactions occurring daily, and concluded this “highlights the need for AI systems designed to robustly support human autonomy and flourishing.”

That sentence could be appended to any AI paper about any AI problem and be equally meaningless. It’s the game, not this player. It’s the output of a system designed to produce exactly this kind of sophisticated irrelevance.

You can have a PhD in mechanical engineering and study if long sharp knives help users do bad things. That’s not actual security leadership. That’s usability research on weapon design, understanding how people interact with a product and whether the interaction has a safety built-in. In threat-model terms, that’s looking for solutions first and skipping right past the entire threat exercise.

Worst Form of Product Safety Management

The unregulated American market drives an AI race towards the bottom. I think we can all agree. It’s just like how unregulated dairy and meat caused mass suffering and death. Remember? Children dying from swill milk in the 1850s? The Jungle? The Pure Food and Drug Act of 1906?

Most if not all product managers need a proper safety line built for them by regulators, or they are heavily incentivized to flood the market with toxic shit and say it’s not their fault. The worst version of safety management is actually the most preferred by software product managers in tech companies today, because it lets them ignore stuff they don’t want to hear. Other industries regulated this out long ago, because harms are so predictable and externalized. It’s like a pill manufacturer asking the safety research team to narrowly assess the best format to open a pill box and to swallow a pill, completely ignoring whether threats make the pill unsafe.

The entire Tylenol 1982 cyanide-laced pill murders lesson is supposed to prevent this kind of scoped-down thinking. It forces a fundamentally different posture than proper security. An attacker isn’t scoped down. Security professionals thus look how bad things happen, constantly, and consider every system already has been compromised until proven safe. It works backward from failures to build defenses.

To put it plainly, from 2012-2016 when I said AI was a dumpster-fire of security vulnerability (e.g. “the fourth V of Big Data’s three Vs“) I was told to shut up so that AI could have a chance of getting off the ground. Then suddenly in 2016 people like Elon Musk said he’d have driverless cars solved in a year and people living on the Moon in four. Security flaws weren’t allowed into the discussion until future-leaning “upside” claims could drown them out anyway.

Threat modeling done right inverts the power imbalance, even just for an hour, to quiet the “everything will be fine” voices. Engineers driven to deliver faster inherently interfere with the slow grind of security experts uncovering vulnerabilities, which the product team hopes and prays never requires their attention.

Sharma’s team studied whether Claude would answer dangerous questions within the product as intended to be used. A security team would ask why it’s intended to be used any certain way, like why there’s even such a thing as bad answers, and what happens next.

That distinction matters. Anthropic chose to call user-experience-level product development research “safety,” staff it with ML researchers, and present it to the public as though the hard problem was being worked on. What they built was heavily academic QA with ethical branding, which is a classic mistake of engineering groups that have not been sufficiently incentivized to listen to seasoned security expertise. It’s the difference between placebo and surgery.

Actual Safety Work

We need to ask different questions differently.

Security asks “why” before “what.”

Why is there pre-authentication, given an attack surface exists? Why is the model embedded in environments where security is being gutted to feed AI demand? Why is a child not the same as a parent and a parent not the same as a guardian? Why is there no distinction between different roles in mental health crisis and why are people in crisis allowed at all?

What happens when someone walks around a filter entirely in minutes? What does authentication and authorization look like when AI agents act autonomously in a world where identity is a fuzzy and contested concept? What happens when the safeguard itself becomes the attack surface, because you’ve published your red-team methodology and handed your adversaries a map of your defenses?

That last point reveals a fundamental disciplinary mismatch. Publishing results is the ML researcher’s instinct to push towards open science, peer review, reproducibility. It is also the opposite of the professional security instinct. Need to know. Role based access. Minimal target surface. These fields have incompatible default behaviors around disclosure, and Anthropic staffed a safety-critical function with people oriented on the marketing end of the spectrum to look “open” about everything. That’s hardly Sharma’s mistake, as he played the game he was told to win. That’s a corporate philosophy that chose academic soft noodling over hard operational security crackers.

I’ve been doing the poetry of information security here since 1995.

Three decades of writing about the space where technology meets institutional failure. I worked for Tim Berners-Lee for years, including him pulling me into a building dedicated to him at Oxford. And what did I find there? A broken hot water kettle pump. Everyone standing around looking at each other and wondering how to have a tea. I broke it apart, hacked it back together, so the man standing in a huge building dedicated to his life’s work could share tea with his guests. The institution of Oxford is very impressive in ways that don’t interest me much. I didn’t wait for a service to come throw away the “broken” thing to justify a new one even more likely to fail. I hacked that old kettle. Sir Tim poured. I’m impressed more by humans who figure out how things work, take them apart and confidently stand and accept the risk that comes with sharing their grounded understanding.

So when the Oxford-trained Sharma announces he’s leaving product safety to study poetry to practice “courageous speech,” I admittedly take it personally.

Poetry is not a retreat from truth to power.

Poetry is what truth to power looks like when the form matches the urgency of the content. This blog is no different than a blog of poetry Sharma could have been writing the whole time he was at Anthropic. It is the hardest kind of speech, not the softest. Poets get exiled, imprisoned, and killed precisely because the form carries dangerous specificity that institutional language is designed to suppress.

Sharma has it exactly backward.

He left a position where he could have said something specific and dangerous into the public, said only vague things, and now wants to go learn the art of saying things that matter. That sequence tells you why Anthropic has been running the wrong team with the wrong leader.

The stand is what’s missing from his resignation.

He said he witnessed pressures to “set aside what matters most.” He didn’t say what those pressures were. He didn’t name the compromise. He didn’t give anyone — bloggers, regulators, journalists, the public — anything to act on. Courageous speech is the specific true thing that costs you something. A self-assuaging resignation letter full of atmospheric dread to pressure others with responsibility and no particulars is the opposite. This too is a structural problem more than a personal one.

Oxford Patterns

If you ever go to Oxford, make sure to look at the elephant weathercock on top and the elephant carving on the corner of the 1896 Indian Institute at Broad Street and Catte Street. This is the building where the British Empire trained its brightest graduates to ruthlessly administer the Indian subcontinent for extraction. They weren’t stupid. They were brilliant, institutionally fluent, and formatted by the institution rather than formed by the work.

India Institute carving of an elephant. “There are still many signs of the original use of the building.” Source: Oxford

This isn’t a new observation. At the exact same time in 1895, the Fabian Society founded the London School of Economics specifically because they saw Oxford and Cambridge as obstacles to social progress. They saw institutions that reproduced elite interests and trained people to serve power rather than challenge it. Sound like Anthropic? Silicon Valley?

Back then it was Shaw, the Webbs, and Wallas who looked at Oxbridge and saw a machine producing administrators for the existing order, and decided the only answer was to build something outside it. Sidney Webb said the London School of Economics would teach “on more modern and more socialist lines than those on which it had been taught hitherto.”

LSE Coat of Arms “to learn the causes of things”, a foundation of scientific thought that forms the exact opposite to Oxford’s motto “The Lord is my light”.

Christopher Wylie went to LSE. He did exactly what Sharma didn’t, he named the company, named the mechanism, named the harm, accepted the consequences.

I made Steve Bannon’s psychological warfare tool.

“To learn the causes of things” was put in action. Oxford trains you to administer. LSE, rejecting harmful elites using technology as an entitlement pipeline, graduated generations of thinkers to investigate and report accurately.

In other words we have the fitting critique from 130 years ago: Oxford produces people who can run systems beautifully without ever questioning whether the systems should exist. They generate pills to be easier to swallow without ever really asking what’s in the pills.

When you produce people whose entire identity is institutional, they follow one of two tracks when they lose faith in the mission: they keep executing inside the machine, or they collapse and retreat in confusion. Neither option includes standing outside and clearly naming what went wrong. Nobody at Oxford is taking the obvious weathervane off the India Institute and putting it in a museum with the phrase “colonialism”.

Sharma chose a quiet, personal retreat. And his first move is to seek another credential in a poetry degree.

Think about the poets most people admire. Bukowski drove a mail truck. Rumi was a refugee. Darwish wrote under military occupation.

Write down! I am an Arab.

They didn’t study courageous speech. They performed it, at personal cost, because the content demanded the form. A person who needs an institution’s permission to find his voice has already answered the question of whether he has one.

Post Resignation Revelation

The indictment lands on Anthropic. They built a safety team that was structurally incapable of seeing the actual safety problems. They defined the threat as “what if someone asks Claude a bad question” rather than “what happens when unregulated technology hands power to people who intend harm yet face no consequences.” They staffed that narrow definition with researchers whose training reinforced it. And when one of those researchers sensed something was wrong, he didn’t have the framework to articulate it, because the role was never designed to look at the real risks.

Anthropic got exactly the safety theater it paid for. And the theater’s timing is exquisite.

Sharma resigned Monday. On Tuesday, Anthropic’s own sabotage report admitted that Opus 4.6 shows “elevated susceptibility to harmful misuse” including chemical weapons development, and is “more willing to manipulate or deceive other participants, compared to prior models.”

Ouch.

The same day, Seoul-based AIM Intelligence announced its red team broke Opus 4.6 in 30 minutes and extracted step-by-step instructions for manufacturing sarin gas and smallpox. Anthropic’s own system card reveals they dropped the model’s refusal rate from 60% to 14% to make it “more helpful” — deliberately widening the attack surface that AIM Intelligence walked right through.

Sharma’s team spent millions if not more studying whether Claude would answer dangerous questions. Perhaps they also studied if touching a hot plate will burn you. He quit without specifics. The next day, his employer confirmed the model answers dangerous questions, and an outside team proved it in half an hour.

The specifics Sharma wouldn’t provide, Anthropic and AIM Intelligence provided for him. He is now off to get a degree so he can write a poem. Meanwhile reality bites.

Sharma deserves better questions to work with and the academic environment to avoid facing the hardest questions. The rest of us deserve actual answers about what he saw, like asking for whom exactly Oxford built its ugly elephant-engraved India Institute.

British Museum: “Samurai Doesn’t Mean Warrior” Says Half of Japanese Samurai Were Female Warriors

First of all, let’s just admit that the British love the spectacle of suffering and gore. London is littered with “museums” glorifying the worst chapters in human history, almost as joyful interpretations of others’ suffering.

Perhaps you could call it the “keep calm and carry on” being a deeper cultural pun about “keep calm and carrion”? Or perhaps I spent too much time in the dank, dark corridors of the London Dungeon.

Notably, a compelling new Samurai exhibit is being held away from Japan.

…some of the country’s most formidable warriors were a group of female samurai called the Onna-bugeisha.

They were every bit as deadly and powerful as their male counterparts and were trained using the same self-defense and offensive maneuvers. They used a special weapon called a naginata…

Eeeeeek. Nails on chalkboard?

This weapon analysis is bullshit. The naginata polearm was used by samurai, foot soldiers, and warrior monks from the Heian period onward.

Utagawa Yoshiiku’s “Taiheiki Eiyūden” (Heroes of the Tale of the Great Peace). Painted by Yoshiiku, Chōōrō studio

It was a primary weapon during the Genpei War (1180–1185) and Nanboku-chō period (1336–1392). When it became associated with women, it was centuries later during the Edo period’s long peace (1603–1868). The weapon was dated, as they always are, and lost battlefield relevance. Women of the samurai class carried it when it had been repurposed as a status symbol and for self-defense. The association therefore was a Tokugawa-era development, not an origin story.

Dare I say this exhibit’s source material retrojects a gendered identity onto a weapon that spent centuries as a general-purpose military tool primarily used by men, because a fraudulent “designed for women” framing creates modern narrative more useful than the actual history?

So, second of all, this exhibit sees ideology in the past but not in its own mirror. It tells us why it deconstructs Meiji myth-making, while it is performing 2026 myth-making.

This British exhibit could have led with the insight that the West fundamentally misunderstands what being a “samurai” meant. It could have used the women’s history to blow open that misunderstanding for everyone, male and female. Instead it drives deeper into a narrow definition and pushes women into it. That makes it representation without revision.

Notice the reporters who focus on fierceness and brutality of woman, which I’m not sure is the correct path towards redeeming women in history.

“It’s a surprise that comes from a narrow use or a narrow understanding of the word samurai, because samurai doesn’t mean warrior,” Dr Rosina Buckland, Asahi Shimbun curator of Japanese Collections, told The Independent.

According to the exhibition, the most celebrated female samurai was Tomoe Gozen who died in 1247, and whose exploits are the subject of The Tale of the Heike. She was reported to have ripped off the head of the samurai Uchida Saburo leyoshi who tried to capture her for ransom.

The curator literally says “samurai doesn’t mean warrior”, an arguably radical claim because it reframes everything, and then the exhibit’s marquee story becomes a woman warrior ripping someone’s head off. I’m only left wondering if she also shit down his neck?

The corrective gets completely swallowed by the very framework it’s supposedly correcting.

The exhibit claims “women made up half of the samurai class” by expanding “samurai” to mean social class rather than warrior caste, then promotes the exhibit with gorified stories that rely on a narrow warrior definition. It’s using both definitions simultaneously depending on which one serves the current sentence.

That’s not scholarship, it’s rhetoric.

Third, something smells off about this whole thing. The Huo Family Foundation sponsorship is a buried detail. A China-based foundation sponsoring an exhibit in Britain to re-frame Japan’s martial identity raises, well, red flags. The funding source may be the context.

Others have pointed out the curation within context is problematic. The object selection gives us a tsuba with a world map (exceedingly rare) cherry-picked to suggest Edo-period global consciousness that wasn’t representative. It offers ornate ceremonial armor that the Japanese don’t consider authentic to samurai identity. Neither curator is a weapon or armor specialist, a fact I find almost impossible to believe given the British adoration of weapons and armor in their exhibits. They seem to have set up evidence which coincidentally fits a Chinese thesis about Japan, rather than building from evidence.

To be fair, Broderick and Buckland clearly agree that the samurai meant more than warrior. The failure becomes that their exhibit couldn’t commit to this story without also promoting gratuitous violence to seek attention in the British market.

Fourth, and finally, this is a familiar pattern: women are admitted to the canon by demonstrating they can meet criteria, not by challenging whether those criteria are correct. The Onna-bugeisha become legible to a museum audience not because samurai culture encompassed governance, education, estate management, arts, and social organization — important recognition of the proper definition and all things women were deeply embedded in — but because a woman ripped the head off someone. The entry ticket to historical recognition is unnecessarily reset to brutality.

It reminds me how Hillary Clinton worked for Goldwater. She didn’t enter politics by articulating an alternative vision; she entered by proving she could operate within the most aggressive, unfit for rule, existing framework available, then spent decades unable to fully shed her origin story. The price of admission shaped the thing it would allow her to be.