Every week, I sit with the pieces that caught on me like burrs, the ones I couldn’t skim past, the ones that made me underline, reread, or argue with myself in the margins. Some confirm what I already know, but the better ones reframe it, nudge me off-balance, or open doors I didn’t even realize were closed. I don’t think of this as curating “recommended reading.” It’s more like keeping a record of the arguments, surprises, and mysteries that changed the shape of my week.
This time, the list ranges wide: from the mathematical foundations of deep learning to how contraceptives alter memory, from Havaianas’ engineering of the flip-flop summer to the question of taste itself, from the ethics of surrogacy to the lived resonance of 1948 in Gaza, from creatine’s overlooked role in women’s health to the enduring pedagogy of the transformer. It’s less a neat syllabus than a set of doorways: some opening into clarity, others into discomfort, all worth stepping through.
The Reads
1. Mathematical frameworks for deep learning
This new arXiv preprint represents a push to give deep learning the formal spine it still largely lacks. For over a decade, neural networks have been built on heuristics, rules of thumb, and computational horsepower more than grounded mathematical certainty. They perform astonishing feats- image recognition, language generation, protein folding- yet we rarely understand why they succeed, or more worryingly, why they fail. The paper attempts to systematize a body of mathematical tools that could explain, predict, and constrain these systems, providing not just retrospective clarity but forward-looking safety. Reading it is like looking at blueprints for a house we’ve already moved into, unsettling, but necessary if we’re to live here securely.
2. Emotion regulation strategies differentially impact memory in hormonal contraceptive users
Beatriz Brandao, Madelyn Castro, and colleagues ask a deceptively simple question: how do hormonal contraceptives alter the way we regulate emotions and remember them later? Their findings suggest that contraceptive use changes the interplay between strategies like reappraisal, suppression, and distancing, producing different memory outcomes compared to non-users. The implications ripple out not just for cognitive psychology, but for how we design studies, interpret data, and support women’s mental health. For too long, female biology has been treated as a confounder to be controlled away. This research reframes it as essential context: a reminder that “the mind” can never be abstracted from the body.
3. The tulpa of the internet
A tulpa, in mystic traditions, is a thought-form brought into being through sustained imagination. This essay argues that the internet itself has become a tulpa factory: a collective hallucination engine that takes our memes, parasocial projections, and obsessions, and alchemizes them into something that feels alive. Think of fictional characters treated as real in fandoms, or influencers who persist as archetypes even after they log off. These aren’t just “jokes” or “figments.” They start to act on us, to have agency in culture. It’s a haunting idea: that our feeds don’t just carry information, they birth entities, and we live in a world increasingly populated by ghosts of our own making.
4. Would you wear flip-flops to a job interview? (on Havaianas’ global strategy)
Amy Coded’s Vogue Business profile of Havaianas is one of those rare brand stories that manages to be both fun and strategic. On the surface, it’s about flip-flops becoming high fashion, but underneath it’s about how a Brazilian staple, once regulated alongside rice and beans during hyperinflation, reinvented itself as a global icon. The company’s EU head, Harm van de Camp, lays out an 18-month repositioning strategy that used collaborations (from Gigi Hadid to Zellerfeld 3D printing) to engineer what we now think of as “the summer of the flip-flop.” What struck me most was how carefully authenticity was balanced with aspiration: Havaianas leaned on its deep Brazilian roots while inserting itself into Scandi minimalism, luxury retail, and viral moments. The takeaway is bigger than sandals: trends that look spontaneous are often carefully staged, long-game plays that blend heritage, timing, and cultural listening.
5. Notes on taste
Taste is slippery. We talk about it as if it’s a stable quality when in reality it’s a collage of experiences, exposures, and accidents. This Are.na essay reminds us that taste is not about judgment, but orientation: what you notice, what you keep, what you build your world around. It’s shaped socially, but also deeply personal, often contradictory. In a culture where algorithms flatten preference into recommendation, cultivating taste becomes an act of resistance, a slower, more intentional way of curating meaning. The essay argues that your taste is less about impressing others and more about navigating your own life. What you consume, collect, and return to is identity in practice.
6. The baby died. Whose fault is it?
This Wired investigation examines one of the most painful questions in modern reproductive ethics: when a baby dies during a surrogate pregnancy, where does responsibility fall? The parents? The surrogate? The clinic? The contracts? The reporting exposes how legal frameworks struggle to catch up with the lived complexity of surrogacy. Markets can broker eggs, wombs, and legal obligations, but they can’t easily allocate blame or grief. What makes the story haunting isn’t just the tragedy itself, but the way accountability evaporates when reproduction is commodified. It’s a piece that forces readers to confront what happens when human creation is mediated by contracts, and who gets left holding the grief when things go wrong.
7. The death of the public intellectual
The Digital Meadow’s essay argues that what has died isn’t intellect, but stagecraft. Once, thinkers like Arendt or Said could influence society because the platforms of mass culture (magazines, lectures, broadcast TV) elevated them. Today, thinkers exist, but the scaffolding that amplified them has dissolved, replaced by fragmented attention economies and algorithmic platforms. What rises to the top now isn’t complexity but virality. The result: our culture is rich with ideas but poor in forums that let them shape public life. It’s not an obituary for intellect itself, but for the cultural infrastructures that once let ideas move the world.
8. How to study people who are very drunk
On its face, this Economist (via Hindustan Times) piece is lighthearted: how do scientists study people who are, frankly, hammered? But read closely and it’s about something deeper: how states of intoxication reveal what inhibition normally hides. Drunk behavior is messy, unpredictable, sometimes aggressive, sometimes tender. By studying these altered states, researchers glimpse truths about cognition, vulnerability, and social bonding. It’s a reminder that even chaos can be data, and that the everyday rituals of nightlife carry insights into what it means to be human together.
9. Contextualized networks for tumor gene regulation
GenBio AI’s research into contextualized interactomes offers a window into how cancer really behaves. Instead of thinking of tumors as clusters of rogue cells, this approach maps the networks of genes and signals that regulate them, and how those networks shift depending on context. By modeling those dynamics with AI, researchers hope to uncover treatment pathways that target not just individual mutations but the orchestration of the whole system. It’s a highly technical piece, but one that underscores a profound shift in medicine: from targeting static parts to understanding dynamic systems.
10. Homeless and hungry, Gazans fear a repeat of 1948
The New York Times’ reporting from Gaza makes the parallels with 1948 visceral. People describe their current displacement and hunger not as unprecedented, but as a continuation of the Nakba. What’s striking is how memory isn’t historical here; it’s present-tense, woven into the way families interpret their suffering. The piece reminds readers outside the region that conflicts are rarely “new”. They are repetitions, reverberations, and unfinished traumas. History doesn’t just inform identity; it actively shapes how people experience crisis in the moment.
11. Creatine in women’s health: bridging the gap from menstruation through pregnancy to menopause
This comprehensive review reframes creatine from a “gym bro supplement” into a potential women’s health nutrient across the lifespan. Abbie Smith-Ryan, Stacey Ellery, and colleagues trace how hormonal changes across menstruation, pregnancy, perimenopause, and menopause affect creatine metabolism, and how supplementation may support muscle, bone, brain, and reproductive health. The paper highlights links to mood regulation, fetal development, and postmenopausal resilience, calling for more sex-specific research. It’s a radical shift in framing: creatine isn’t just about athletic performance, it’s about equipping women’s bodies for the physiological demands of every life stage.
12. The Annotated Transformer
Still one of the most celebrated teaching tools in modern AI, this project patiently walks through the architecture behind today’s most powerful models. Rather than treating transformers as magic, it annotates each line of code, linking theory to implementation. For students and practitioners, it’s invaluable. For the rest of us, it’s proof that deep innovation often rests on clear explanation. In a field that moves at breakneck speed, the Annotated Transformer remains a rare artifact of pedagogy: careful, open, and enduring.
That’s this week’s list. If something here tugged at you, share it with a friend who’d linger over it too. And if you’ve read something recently that belongs on next week’s stack, hit reply, I love trading links.
Prefer a shorter, more visual version? I post this weekly on Instagram as well: @harnidhk.
See you next week with another set of ideas worth your time.
Love,
Harnidh.
Thanks for sharing these. Out of curiosity, how do you skim through and find these pieces?