Weekly Reads: 20/12/2025

Things I Read This Week That Made Me Think

There’s a recurring fantasy in how we talk about systems: that they’re legible. That if you learn the rules, you can play the game. That the power law can be reverse-engineered, that “making it” is a matter of effort calibrated to opportunity, that the algorithm is neutral if you just feed it right.

This week’s reading kept circling around a different proposition: that the systems we navigate, whether markets, cities, platforms, or families, are shaped by forces we can’t see and often can’t name. The compounding of early advantages, the laundering of status through taste, the way a technology designed without you will eventually be used against you.

The most striking version of this came from an essay on Delhi’s elite. The argument: what we call “sophistication” and “taste” is really just algorithmic head start. The urban upper-caste kids who were online in the 2000s built up a decade of data that Meta and Google now use to push curated content. Everyone who came later is playing catch-up, not because they lack intelligence or ambition, but because the algorithm has a longer memory of someone else. Caste, laundered through vibes. The memo about Mokobara luggage and matcha and pickleball went out years ago and you just weren’t on the list.

Invisible structures don’t announce themselves. They present as merit, as taste, as the natural order. The VCs who insist they can identify billion-dollar companies at seed stage aren’t lying, exactly, because they believe it. But the essay on power law this week points out the category error: you can’t reverse-engineer a lottery winner by analysing the ticket before the draw. The power law is a system property, not a selection tool. Treating it as predictive rather than descriptive leads you to fund the obvious (the same LLM layer, the same quick-commerce model) while the actual outliers slip past looking like black car services for rich people or websites for couch-surfing.

Amia Srinivasan, in the London Review of Books, asks the question that haunts all of this: does knowing help? Can insight produce change? Her essay traces psychoanalysis from Freud’s Vienna to Fanon’s Algeria, through Reich and the sex-pol movement, through Judith Butler and the war on “gender ideology.” The answer she arrives at is uncomfortable: probably not. Insight is necessary but not sufficient. The gap between describing the world and changing it is not bridged by better descriptions.

And yet we keep describing. This week: Japan as a preview of American decline, not a cultural quirk. Women’s scepticism of AI as pattern recognition, not technophobia. The grief of never becoming a grandparent, except the “grief” is about having to parent the children you were hoping to spoil. A divorce essay that refuses to name villains because the pressure to produce clean narratives is itself part of the violence. Indian wedding rituals traced back to colonial trade routes, the marigold arriving from Mexico via Portuguese traders and becoming, within a few centuries, indispensable to Hindu ceremony.

Even TikTok, finally cleaving itself from ByteDance- the structural change took years of political theatre to produce. The mechanics were always clear. Oracle gets 45%. The narrative had to catch up.

What does it mean to navigate systems you can’t fully see? To play games whose rules are written in a language you’re still learning? To know that the map is not the territory, and that the mapmakers had a head start?

Let’s find out.

1. The Impossible Patient London Review of Books, Amia Srinivasan https://www.lrb.co.uk/the-paper/v47/n23/amia-srinivasan/the-impossible-patient

The standout piece of the week: a 10,000-word essay delivered as the first of this year’s LRB Winter Lectures, tracing the return of psychoanalysis to left political thought.

Srinivasan’s argument is layered. The unconscious never left the scene because it sets the scene. But there’s renewed interest in Freudian vocabulary as a diagnostic tool for the current moment: why do people support authoritarian strongmen? Why does the war on “gender ideology” generate such libidinal intensity? Why do Israeli citizens perform self-defeating operations under the sign of “safety”?

The essay moves through Freud’s Vienna (where psychoanalysis was born of political retreat after the collapse of liberal hegemony), through Reich and the sex-pol movement, through Fanon’s psychiatric hospitals in Algeria, and lands on the question that haunts all politically-inflected psychoanalysis: can insight produce change?

Freud’s answer was that it could, for individuals, through the therapeutic encounter. But politics doesn’t offer the controlled conditions of the analyst’s couch. When you diagnose someone politically, when you suggest their Zionism is trauma, their transphobia is projection, you’re met not with working-through but with resistance. The knife cuts both ways.

Srinivasan finds hope in organising rather than analysis: the praxis of Ella Baker, the craft of having “hard conversations” that aim not at insight but at collective action. The psyche, she suggests, can flow from action. Sometimes you become the subject who can stand up by simply standing up.

What I got out of it: The distinction between knowing the world and changing it is older than Marx, but Srinivasan’s formulation is the sharpest I’ve read in years. The essay is also a corrective to the therapeutic turn in politics: the belief that if we just understood one another better, things would be fine. They wouldn’t. Understanding is necessary but not sufficient. At some point, the work of organising has to begin.

2. Japan Is What Late-Stage Capitalist Decline Looks Like Drops in the Ocean — Ellie

Japan Is What Late-Stage Capitalist Decline Looks Like

The argument: Japan isn’t a cultural quirk. It’s a preview. The “weirdness” that Western observers attribute to Japanese culture is actually the downstream consequence of economic structures that America is now replicating.

The timeline: Japan’s bubble burst in 1991. The “Lost Decade” became the Lost Decades. Wages peaked in 1997 and haven’t recovered. Irregular employment (the gig economy by another name) became normalised. Dating requires disposable income, a stable schedule, and housing privacy, all of which became scarce.

The parallels to America are point-by-point. Evil jobs (ブラック企業) that enforce unpaid overtime and surveillance? Amazon drivers denied bathroom breaks. A loneliness epidemic and collapsing birth rates? Check. Convenience food as a substitute for home life? Sysco slop. Parasocial relationships replacing actual intimacy? Influencer culture monetising the same emotional deprivation that idol culture does.

In a functional society, basic human milestones are incentivised. In late-stage capitalism, they become financially punishing. The incentive structure flips. What should be rewarded is penalised, and what should be discouraged becomes adaptive.

What I got out of it: The essay resists the “Japan is weird” framing and asks instead: what structural conditions produce this weirdness? The answer turns out to be familiar. The lesson for Americans is that the Japanese didn’t choose atomisation and parasociality. They adapted to an economy that made real connection logistically impossible.

3. Why Women Aren’t Swooning Over AI Like Men Are The Noösphere — Katie Jagielnicka

There's a Reason Women Aren't Swooning Over AI Like Men Are

Women have 22% lower odds of using generative AI than men. Only 31% of Anthropic’s users are women. The standard explanation is exposure and training, since women are underrepresented in STEM. But the Harvard study finds the gap persists even when women are explicitly given opportunities to learn AI tools.

Jagielnicka’s alternative explanation: women are paying attention. The first time many women encountered AI was through deepfakes, as non-consensual sexual imagery overwhelmingly targeting women and girls. By the end of 2020, 95% of deepfake videos were sexual, 90% featured women. Tools explicitly built to “nudify” women are freely available. Harassment campaigns against women are automated.

Beyond the abuse cases, there’s the systemic bias. AI recruitment tools recommend men over women for higher-paying jobs, even with identical qualifications. AI chatbots advise women to ask for lower salaries. Healthcare AI downplays women’s symptoms. When asked to generate an image of a “manager,” the models produce men.

And there’s the double bind: women who use AI are judged as less competent than men doing identical work. A recent study found that female engineers faced a 13% competence penalty for using AI assistance, compared to 6% for men.

The conclusion: women aren’t risk-averse. They’re risk-smart. They see the cracks in the techno-optimist fantasy because the technology was built without them.

What I got out of it: The essay reframes “women don’t like technology” as “women accurately perceive when technology doesn’t like them.”

4. Reasons Why You Will Never Make It in Delhi Karthika’s Substack, Karthika Rajmohan

Reasons why you will never make it in Delhi

Delhi sells a feudal dream: a “bevvy of help” to tend to your every need, body men to carry your lunchbox, maids to hand you coffee. But who actually gets to live that dream? Rajmohan identifies three barriers to entry.

The first is money, obvious and widely acknowledged.

The second is what she calls “Urban Savarna Soft Power” (USSP): the unspoken memo about what’s cool. Mokobara luggage. Matcha. Pickleball. The algorithm delivers these signals to those already in the know. But who gets in the know?

The answer, Rajmohan argues, is caste by another name. The urban upper-caste kids who were online in the 2000s built up a decade of data that Meta and Google now use to push curated content. The algorithm has their “base.” Everyone who came later, after the 2016 Jio data democratisation, is still catching up. What we call “taste” and “sophistication” is really just algorithmic head start.

The third barrier is genetic lottery: generational wealth, pre-independence zamindari status, and bureaucratic lineages. No amount of USSP acquisition will get you past this wall.

The aspiring class, having mastered USSP, sheds whatever liberal values they picked up online and immediately adopts feudal signifiers. They refuse to enter the office without their man Friday. They won’t serve guests without full-time help. The cycle reproduces itself.

What I got out of it: A brutally clear-eyed dissection of how class works in India, and how the algorithm, far from being neutral, launders caste through “taste.” The essay is specific to Delhi but the mechanism generalises.

5. The Book That Remade America Arc Magazine, Daniel Oppenheimer

The Book that Remade America

Norman Podhoretz, who died this week at 95, was the last of the New York Intellectuals and the last of the original neoconservatives. This excerpt from Oppenheimer’s 2016 book traces how Podhoretz’s 1967 memoir Making It, and its brutal critical reception, reshaped his politics.

Making It was supposed to be a literary act: an honest account of ambition. But Podhoretz was savaged by friends and colleagues. He retreated, drank more, grew depressed. And then he began to reconstruct what had happened.

The reconstruction: the book wasn’t primarily literary. It was political. A “Yes” to success and ambition, but more meaningfully a “No” to the emerging radicalism of the late 1960s. The critics weren’t responding to the book’s literary failures; they were punishing his dissent.

This reframing, from wounded author to principled contrarian, became the template for Podhoretz’s later career. He’d been shown that the liberal intelligentsia couldn’t tolerate deviation. He reorganised himself around the fight.

What I got out of it: A case study in how rejection can become identity. Whether Podhoretz’s reframing was accurate is less interesting than the fact that it was generative- it gave him something to fight for. The essay is also a useful reminder that neoconservatism didn’t emerge from nowhere; it emerged from specific wounds and specific rooms.

6. When Does a Divorce Begin? The Yale Review, Anahid Nersessian https://yalereview.org/article/anahid-nersessian-divorce

A personal essay structured around a book review that never quite happens. Nersessian was asked to review recent memoirs and novels by women, but what she writes instead is about her own divorce, or rather, about the impossibility of writing honestly about a marriage ending.

The essay opens with epigraphs about love, lying, speech, and silence. Epigraphs, she notes, are “envious”. A residue of wishful thinking, a way of signalling what the writing might have been. They’re also fake-outs. The real thing toddles in, uncertain.

The details are precise: she wore a white cotton nightgown from the Marché aux Puces on her wedding day. She told her therapist she wasn’t afraid to get married because she could always get divorced. She didn’t hide her ambivalence from her husband, and this hurt him very much.

The essay weaves through Rachel Cusk’s Aftermath, through the experience of losing custody of mutual friends, through the strangeness of watching the shared condition of marriage dismantle itself overnight.

What I got out of it: The essay is elliptical and refuses resolution, which is the point. Divorce doesn’t have a clean beginning or end. The narrative pressure to produce one, to identify villains, to explain, is itself part of what makes the experience so disorienting.

7. The Rise of Grandfamilies The New York Times — Catherine Pearson https://www.nytimes.com/2025/05/18/magazine/grandparents-families-children-kids.html

A reported piece on a phenomenon that doesn’t fit any of the available narratives: nearly 2.7 million grandparents in America are now primary caregivers for their grandchildren. Not babysitting. Not helping out. Parenting, because their adult children can’t.

The reasons are various and grim: addiction, incarceration, mental illness, economic collapse. The opioid crisis is a through-line. In New Mexico, 8% of all children are in kinship care arrangements, more than double the national average. The report from the Los Alamos National Laboratory Foundation found that drug addiction is the biggest driver.

Kinship caregivers receive a fraction of the support that licensed foster parents get. Foster parents receive monthly stipends, Medicaid coverage, and ongoing case management. Grandparents who step in to keep their grandchildren out of the system get almost nothing. No TANF if their income is too high, no food stamps, no subsidised childcare. Many are on fixed incomes. Many burn through retirement savings. Many return to work in their 60s and 70s to make ends meet.

The emotional weight is compounded by the circumstances. These grandparents are grieving their own children, watching them struggle with addiction or sit in prison, while simultaneously raising the next generation. They carry guilt, anger, exhaustion. They navigate school systems that have changed beyond recognition, help with homework on devices they don’t understand, manage the trauma their grandchildren carry from disrupted early childhoods.

Grandfamilies save the government an estimated $10.5 billion annually by keeping children out of foster care. The thanks they get is a bureaucratic maze and means-tested programs designed for different circumstances.

What I got out of it: The framing of “grandparents raising grandchildren” as heartwarming obscures the structural failure underneath. These aren’t families choosing multigenerational living for its benefits. They’re families absorbing the fallout of systems that don’t work, and receiving almost no support for doing so.

8. Guests of Honor Orion Magazine https://orionmagazine.org/article/guests-of-honor/

A lyrical meditation on the botanical elements of Hindu wedding rituals, the turmeric, the marigolds, the paddy stalks, the bael patra, and the colonial trade routes through which some of them arrived.

The marigold, ubiquitous in Hindu ceremony, came from the New World only a few centuries ago. “Mary’s gold” arrived via Spanish and Portuguese traders. It’s now a cliché: strings of marigold draped across pillars, thrown at the bride and groom as they circle the ceremonial fire. But the reminder is useful: rituals that present themselves as “ancient” and “timeless” are in fact edited, adapted, hybrid.

The essay moves through Goddess Lakshmi’s footprints drawn in rice powder, through the paddy stalks that symbolise harvest abundance, through the Gandharva marriage of Kalidasa’s Shakuntalam, a marriage performed without rituals, in the forest, where the heroine is described as “delicate as a newly-opened jasmine.”

What I got out of it: A reminder that even the most “traditional” ceremonies are layered with history, trade, and adaptation. The marigold came from Mexico. The ritual absorbed it. Tradition is not stasis; it’s slow-motion editing.

9. The Map Is Not the Territory: Why VC Misunderstands the Power Law Alphanome.ai https://www.alphanome.ai/post/the-map-is-not-the-territory-why-venture-capital-misunderstands-the-power-law

The argument: venture capital treats the power law as a selection mechanism when it’s actually a system property. Returns are dominated by a tiny percentage of winners, that’s true. But you cannot reverse-engineer a lottery winner by analysing the ticket before the draw.

The consequences of this misunderstanding: VCs demand that startups look like giants on day one. They build TAM models showing $10 billion outcomes. They flood capital into “consensus” bets. And they miss the actual outliers, which rarely look like fund-returners at the time.

Uber was a black car service for rich people. Airbnb was for people who couldn’t afford hotels. Coinbase was for bitcoin hobbyists. These companies were passed over precisely because they didn’t fit the template.

The alternative: VCs should be farmers, not snipers. Plant seeds in fertile soil (uncapped markets) with hardy DNA (great founders). You can’t predict which tree will block out the sun, but the forest will produce one.

What I got out of it: A useful corrective to VC mythology. The power law is real, but using it as a selection tool is a category error. You’re not identifying winners; you’re constructing narratives about why something might be big. Those narratives tend to converge on the obvious, which is precisely where the returns aren’t.

10. TikTok Agrees to Deal to Cede Control of US Business TechCrunch, Lucas Ropek https://techcrunch.com/2025/12/18/tiktok-agrees-to-deal-to-cede-control-of-u-s-business-to-american-investor-group/

The long-anticipated structural change finally happened: TikTok has reached a deal to cede a substantial portion of its US operation to American investors. Oracle, Silverlake, and MGX (an Abu Dhabi-based AI investment firm) will together own 45%, while ByteDance retains nearly 20%.

The closing date is January 22, 2026. The new entity, “TikTok USDS Joint Venture LLC”, will oversee data protection, algorithm security, content moderation, and software assurance.

The deal parallels the language in Trump’s September executive order. It’s the culmination of years of political pressure, national security hand-wringing, and legislative theatrics.

What I got out of it: A reminder that political theatre and structural change operate on different timelines. The TikTok saga produced endless takes about sovereignty, censorship, and geopolitics. The actual resolution is a joint venture with Oracle.

11. Get Ready for AI Media The AI Ad Economy, Debra Aho Williamson

Get Ready for AI Media

The prediction: 2026 is the year AI platforms join the media plan. Just as social media went from a $1.2 billion afterthought in 2007 to a $300 billion channel today, AI platforms like ChatGPT, Gemini, and Copilot are becoming advertising destinations.

The numbers: ChatGPT is on track to reach 1 billion weekly users by the end of 2025, just three years after launch. For comparison, Facebook took more than eight years to reach 1 billion monthly users.

Google is already showing ads in AI Overviews. Microsoft’s Copilot has ads. ChatGPT has been testing app suggestions that looked suspiciously like promotions (OpenAI pulled the feature after backlash). The pattern is clear: AI is becoming a discovery channel, and where discovery goes, advertising follows.

The strategic implication: brands will need to optimise for AI visibility across five fronts: content marketing, websites, listings platforms, social channels, and community/review sites. The same game that SEO played with Google is about to be replayed with LLMs.

What I got out of it: The piece is useful for anyone thinking about marketing in the next few years. But the larger observation is structural: AI companies need revenue, and advertising is the obvious answer. The user experience will change accordingly.

There’s a line in the Srinivasan essay I keep returning to: “The will to greater and greater theoretical mastery can itself be a form of resistance to the real work of practical transformation.”

It’s a warning to academics and intellectuals, primarily. But it generalises. We love to map the territory. To name the power law, the USSP, the algorithmic caste system, the libidinal investment in genocide. Naming feels like knowing. And knowing feels like doing.

It isn’t.

Fanon learned this in Blida-Joinville, treating French torturers and Algerian rebels in the same hospital. He understood, psychoanalytically, that colonialism was mutilating everyone, the colonist and colonised alike. The insight was accurate. It was also useless as a guide to action. When the Algerian revolution came, it came through violence and politics, not therapy. Fanon joined the FLN. He didn’t offer the French servicemen a second session.

But here’s what stayed with me: even Fanon kept writing. The Wretched of the Earth was dictated in the final months of his life, while he was dying of leukaemia. He knew that insight alone couldn’t produce liberation. He also couldn’t stop trying to articulate what liberation might look like. The diagnosis and the struggle weren’t opposites — they were in tension, and he held the tension until the end.

Maybe that’s the thing. The essays this week aren’t maps that will get you where you’re going. They’re more like weather reports: useful for knowing what you’re walking into, not for walking itself. The Delhi essay won’t get you past the third barrier. The Japan essay won’t rebuild American third places. The AI essay won’t make the algorithm treat women fairly.

But they do something else. They name the fog. And sometimes, when you’re in the fog, that’s not nothing. It’s the difference between thinking you’re lost because you’re stupid and knowing you’re lost because the terrain is genuinely hard to navigate.

The Srinivasan answer to “what comes after insight?” is organising: the craft of conversation, the practice of collective action, the willingness to believe that the psyche can follow the body. Ella Baker, travelling through the American South in the 1940s, understood something the theorists didn’t: you don’t change someone’s mind and then ask them to act. You invite them to act, and the acting changes the mind. “We’ve both got good dresses.”

I think about that in the context of the AI scepticism essay. Women aren’t avoiding AI because they don’t understand it. They’re avoiding it because they understand it fine and the cost-benefit doesn’t work. What changes that calculus? Not better explanations of how transformers work. Not reassurances that bias is being addressed. The calculus changes when the actual risks shift, when the tools stop being used against them, when the systems are designed with them in mind.

That’s not insight work. That’s power work. And power work is slow, collective, and often invisible. The opposite of the solo genius hitting on the billion-dollar idea.

The VC essay makes the same point from the other direction. The power law is real, but using it as a selection tool is cope. You’re not identifying winners; you’re constructing narratives about why something might be big. The actual outliers didn’t look like winners at the time. They looked weird. The pattern recognition that VCs pride themselves on is precisely what screens them out.

So what do you do? You plant seeds. You tend soil. You accept that you can’t predict which tree will block out the sun, only that the forest will produce one. It’s an argument for humility, for diversification, for resisting the narrative pressure to have already figured it out.

Which brings me back to the divorce essay, and the wedding essay, and the grandfamilies piece. All of them, in different ways, are about the violence of premature closure: the demand that a marriage produce a clean story, that a ritual present itself as timeless, that a generation’s choices be legible as either virtue or betrayal.

The marigold came from Mexico. The ritual absorbed it. The grandparents raising their grandchildren didn’t choose this, they’re absorbing the fallout of addiction and incarceration and economic collapse, and the system that should support them offers means-tested programs designed for different circumstances. The divorce has no beginning because the pressure to name one is itself part of what makes divorce so disorienting.

Maybe the honest position is: we’re in the middle of something. The systems are shifting. The sediment is still settling. The maps we’re drawing will be obsolete by the time they’re printed.

In the meantime: read the weather reports. Name the fog. And then, because knowing isn’t doing, find someone to organise with.

Until next week.