Weekly Reading List: 21/09/2025
Things I Read This Week That Made Me Think
I read widely, but what I save here each week is about stickiness. The pieces that followed me into conversations, the ones I argued with in my head while commuting, the ones I found myself screen-shotting for friends.
Some weeks, the thread tying them together is obvious. Other weeks, like this one, the links are looser but no less meaningful: the origins of the firm, the social fabric of unemployment, the allure of Ozempic, the myth of AI’s inevitable ascent, the decline of Maoists, and the fragile construction of masculinity. What unites them is the reminder that our institutions, whether companies, movements, or identities, are more brittle and more invented than they seem.
The Reads
1. The Nature of the Firm (Ronald Coase, 1937)
Why do firms exist at all? Coase’s answer, transaction costs, feels almost obvious now, but at the time it reoriented economics. What stuck with me on reread is how timeless the tension is: when is coordination more efficient inside an institution, and when is it better left to the open market? Startups, DAOs, and even AI agents keep circling this question.
2. What does “luxury” mean today? (Washington Post)
WaPo explores how luxury fashion has shifted from heritage and craftsmanship toward vibes, scarcity, and marketing theater. A $2,000 bag isn’t just about leather, it’s about signaling belonging to a moment. The piece asks: is luxury still about quality, or has it become a game of cultural fluency? It reminded me that “luxury” has always been less about the thing itself, and more about what it lets you say without words.
3. AI for Creative Testing (Notion playbook)
A hands-on workflow guide to using AI for ad creative: generating variants, testing messaging, and iterating at scale. What’s compelling is its matter-of-factness. AI not as a grand revolution, but as an accelerator of experiments. It’s a reminder that “creativity” in business often isn’t about one brilliant idea, but about testing enough shots on goal that one hits.
4. How I Got a Job at DeepMind Without a ML Degree
A candid career story that doubles as critique: prestigious AI labs demand elite credentials, yet here’s someone who hacked their way in by building sideways: projects, persistence, and proof of ability. It highlights the gap between gatekeeping and actual skill. The lesson isn’t “ignore degrees,” but that orthogonal paths exist if you’re stubborn enough to carve them.
5. . Some Fragmented Thoughts on Our AI Future
This is, for me, a theological meditation on AI as both false salvation and potential tool. The author argues that Silicon Valley’s pursuit of AGI has taken on quasi-religious tones, promising either utopia or apocalypse while erasing the human vocation to meaningful labor. What’s striking is how the piece reframes AI not as an engineering challenge but as a question of ontology: are we creators, or are we surrendering to what we’ve created? It ends with a call to treat AI as a properly ordered tool of dominion in service of human flourishing, vocation, and culture, and not as a deity. Whether you agree or not, it’s a reminder that the AI debate isn’t just technical or economic. At its core, it’s spiritual.
6. Beyond the Sum: Unlocking AI Agents’ Potential Through Market Forces (arXiv)
Jordi Montes Sanabria and Pol Alvarez Vecino argue that LLM-powered agents could one day act as autonomous market participants, trading, negotiating, and even coordinating at speeds and scales that humans can’t. The catch: Today’s infrastructure isn’t built for non-human actors. Identity, payments, interfaces: all assume people in the loop. The paper is half technical, half speculative, but the thought experiment is provocative: what happens when markets aren’t just human playgrounds anymore?
7. Long-Term Unemployment and College Grads (NYT)
NYT looks at the growing number of educated young people stuck in long-term unemployment. Beyond stats, the piece captures the emotional toll: when the “college = career” story collapses, identity collapses with it. It reminded me that jobs are scaffolding for self-worth. Strip that away, and the void is existential, not just financial.
8. The Worst Argument Against Ozempic (Cremieux)
Cremieux dismantles the hand-wavy critiques of weight-loss drugs as “unnatural.” The essay reframes Ozempic as a tool in a landscape where obesity has been moralized, stigmatized, and commodified. What’s refreshing is how it refuses to romanticize either side. Instead, it asks: why do we accept heart surgery as “medicine,” but judge GLP-1s as vanity? The answer reveals more about us than about the drug.
9. Ask Before You Intro (Elad Gil)
Elad Gil on why intros should never be casual. A sloppy introduction isn’t neutral; it wastes time and burns trust. I’ve been on both ends, and he’s right: good intros create lift, bad ones create drag. His simple advice, “always ask first”, is a professional kindness that compounds.
10. What If AI Doesn’t Get Much Better Than This? (New Yorker)
The New Yorker asks us to consider the plateau scenario: what if GPT-5-level is as far as it goes? It’s still transformative, but it breaks the narrative of inevitable exponential curves. I like this because it punctures hubris, forcing us to plan not just for endless growth, but for the possibility that today’s tools are the revolution.
11. Parents Giving HGH to Kids to Grow Taller (RealClearScience)
A disturbing archive piece: wealthy parents dosing their kids with human growth hormone to boost height. It’s not about health, it’s about advantage. Reading it now, in an Ozempic-saturated moment, it feels like a precursor: biology as a battleground for status. The question isn’t “does it work?” but “what kind of arms race are we entering?”
12. The Decline of the Maoist Movement (Deccan Herald)
Deccan Herald traces the retreat of India’s Maoist insurgency. Once a powerful force, it’s now fragmented and fading. What stands out is how inevitabilities erode: movements that once looked unshakeable weaken in plain sight. History is full of these declines, but living through them often feels like watching a shadow disappear without noticing the moment it went.
13. Meta + Ray-Ban Smart Glasses (The Verge)
The Verge gets hands-on with Meta’s latest glasses. Better battery, sharper design, but the adoption question lingers. For years we’ve been promised AR as “the next platform.” Yet every iteration feels just shy of tipping over. It makes me wonder if the bottleneck is technical, or if it’s cultural. Maybe we just don’t want to live under constant capture.
14. American Masculinity (GQ)
GQ’s special issue explores masculinity in flux, anchored by Glen Powell. Glossy as it is, the subtext is confusion: nostalgia and reinvention, bravado and vulnerability, all jostling at once. What I liked most was the implicit admission: masculinity is not a fixed archetype but an evolving argument.
15. VCBench: Benchmarking LLMs in Venture Capital (arXiv)
A new benchmark designed to test whether LLMs can predict founder success better than humans. The dataset, 9,000 anonymized founder profiles, shows models like DeepSeek and GPT-4o outperforming both market indices and, in some cases, tier-1 firms. The implications are wild: if models can out-pick VCs, what does that say about judgment, pattern recognition, and the future of investing? It’s part research paper, part existential threat to my industry (and job security lol).
16. Planetary Intelligence (The Atlantic)
An older piece, but still expansive: can Earth itself be thought of as intelligent? The essay frames ecosystems as information-processing systems, coordinating survival like a neural net. It’s speculative, yes, but valuable as a lens. If nothing else, it loosens our anthropocentric grip on “intelligence.”
17. Life is Poker, Not Chess (Robin Guo)
Guo argues that life resembles poker more than chess: incomplete information, probability, and luck matter more than perfect calculation. It’s a metaphor I return to often. Too many of us punish ourselves for not “solving” life like a chess game. Poker wisdom, play the odds, manage risk, adapt, feels truer, and more forgiving.
Closing
If there’s a thread across this week’s list, it’s valuation. What do we call “luxury” in a world where scarcity is manufactured? What do we call “success” when models might out-pick investors? What do we call “progress” if AI never leaps past today?
The answers aren’t fixed. They’re negotiated, fought over, reshaped. And that’s what makes them fragile. Firms, markets, movements, even identities aren’t permanent structures; they’re provisional agreements we keep remaking. That can be unsettling. But it’s also liberating, since if these categories were built, they can be rebuilt.
If something here made you rethink what’s valuable, share it. And if you’ve read something that belongs in next week’s stack, send it my way. The best maps are the ones we draw together.
Multiple good finds, Harnidh. Look forward to your weekly list.
On Coase’s classic: insights tend to feel obvious in hindsight, because they explain the reality of what we’re experiencing in the world well. But it is that clarity that’s valuable because it helps us figure out what’s noise. A few years from now, my bet is we’ll see that firms that found leverage in AI by using it to increase consumer welfare will outperform the rest on revenue. I wrote about examples and why here: https://meryam.substack.com/p/the-welfare-arbitrage-ai-is-most
On AI for creatives: I work in digital advertising and this is one of the applications of AI that I am the most bullish about. What doesn’t get talked about often is the marketplace structure of the top websites/apps where ads are served. While advertisers can start testing and submitting multiple creatives, much of the value they aim to generate for their brands comes from the marketplaces more effectively matching the ads to their viewers. So, marketplaces that prepare for the influx of creatives should outperform the rest ( a la the welfare argument in the linked post ).