Most of this week's best reads are about AI, and they approach it from different distances. Astral Codex Ten's "AI Safety Is Not A Corporate Plot" is the strongest of the group: it argues from the record that lab leaders raised extinction concerns before their companies existed, and that calls for slowdowns hurt valuations rather than help them. "The 'Last Mile' Problem in AI Safety" (3 Quarks Daily) takes up the communication side, arguing that advocates should describe concrete physical harms like engineered pathogens and infrastructure attacks instead of sci-fi abstractions. Noahpinion's "AI vs. the railroads" covers the financial side, noting that data center capex now rivals the 19th-century railroad buildout and is increasingly funded by debt.
12–18 Oct 2026 35 articles
Reading digest.
week
This week’s reading notesMost of this week's best reads are about AI, and they approach it from different distances. Astral Codex Ten's "AI Safety Is Not A Corporate Plot" is the strongest of the group: it argues from the record that lab leaders raised extinction concerns before their companies existed, and that calls for slowdowns hurt valuations rather than help them. "The 'Last Mile' Problem in AI Safety" (3 Quarks Daily) takes up the communication side, arguing that advocates should describe concrete physical harms like engineered pathogens and infrastructure attacks instead of sci-fi abstractions. Noahpinion's "AI vs. the railroads" covers the financial side, noting that data center capex now rivals the 19th-century railroad buildout and is increasingly funded by debt.
Two LessWrong pieces on China, "Headwinds Everywhere (in the Chinese AI Safety Field)" and "A Chinese Perspective on 'On Social Reality in China'", show how funding rules, censorship and geoblocking limit safety work there and correct some Western assumptions. For technical depth, "You Should Specify Your Code Unambiguously (and Prove That You Did)" uses LLMs with Lean proofs to produce verified Rust, and its Gzip result is concrete. "What's up with GPT's welfare?" and "Three perspectives on corrigibility" look at how models describe themselves and whether they stay correctable. The latter is useful but held back by formatting problems.
Outside AI, Aeon's "The ladder of knowledge" reads Gödel's incompleteness as the open-ended extensibility of reasoning, and Quanta's "How Do Fungi Keep the Planet Healthy?" shows brainless organisms trading like market actors. "AI #190: Scion of the Times" is a link roundup worth skimming, and "Exploring Intrapersonal Game Theory" is interesting but rough in presentation.
This week's reads share a concern with how we know things and what happens when the tools or labels we use stop matching reality. The AI pieces are the largest group, and the strongest is Ramez Naam's "Where's the "intelligence explosion"?" at Noahpinion, which uses newly released lab data to argue that self-improvement gains are roughly 5-10x too weak for a fast takeoff. Scott Alexander's "An Open Letter To Steven Pinker On AI" (Astral Codex Ten) argues the opposite side of the risk debate, and Quanta's interview "He Wants AI Benchmarking To Be a Science, Not a Dark Art" explains why leaderboards stopped being informative. Of the technical LessWrong posts, "Beyond Owls" is the most striking, showing that subliminal learning can carry backdoors and capabilities.
Outside AI, the best pieces are about naming and understanding the body and the world. In Aeon, "Don't use the 'C-word'" makes a humane case for renaming low-grade lesions that will never cause harm, "Life on a hair trigger" shows how chronic danger hardens snap judgments and how READI Chicago's job-plus-CBT approach reduces violence, and "Deskilling the body" argues that intelligence cannot be separated from physical action. In science, Quanta's "Triplet of 'Ghostly' Galaxies Hints at What Dark Matter Is — and Isn't" bears on the dark matter versus MOND debate, and "How Does Life Unfold? A Landscape Metaphor Comes Into Its Own." shows Waddington's old image becoming quantitative biology.
History and culture supply the quieter highlights. Astral Codex Ten's "Your Book Review: We Present Television" finds today's telepresence already imagined in 1940, while its "Your Book Review: Storm Of Steel" pairs Jünger's war memoir with a harrowing NICU story. Aeon's "The ghetto on the lagoon" shows how Venice's first ghetto grew out of a compromise between economic need and prejudice, and The Marginalian's piece on Semmelweis and Wollstonecraft ties the history of handwashing to a personal loss. On the economics side, Noahpinion's "Stop spending money on smaller classes" is the most useful.
AI is the largest category this week, but the best pieces ask what it means rather than what it can do. In "San Francisco State University in Twilight" (LessWrong), an AI proves an open math conjecture in nine minutes, and the author mourns what that does to human understanding. "Can a Million AI Agents Create Superintelligence?" (Noahpinion) uses benchmark data to argue that more agents buy speed, not intelligence. "Against Against Handover" (Astral Codex Ten) makes the case for handing governance to aligned AI, and it is worth reading to argue with. In science, "AI Proved the Quasi-Riemann Hypothesis. What Happens Now?" (Quanta Magazine) covers a similar math result, and "Malaria Parasite Caught in the Act of Breaking and Entering" (Quanta Magazine) is a clear structural-biology explainer.
In economics and history, the standouts are "China's growth machines" (Works in Progress), which credits competing local governments for China's industrial rise, and "Beijing Will Not Pace the Frontier" (SemiAnalysis), which finds little published safety work from Chinese labs. "What Will Happen to Indian IT?" (Asterisk Magazine) looks at AI's effect on a services economy. In history and culture, Devereaux's "Teaching Paradox: Hearts of Iron IV, Part IIIb" (ACOUP) uses a strategy game to argue about how we remember WWII atrocities. "Horrible, unmanly, indispensable" (Aeon) is the best of the social histories.
The personal and literary pieces round out the week. "Dear Al Pacino" (Psyche) and "The Uncanny Catharsis of the AI Emomercial" (The Paris Review) are the strongest essays. "Your Book Review: The Ghost In My Brain" (Astral Codex Ten) asks how medicine should learn from patients who all differ. For poetry, "Friday Poem" (3 Quarks Daily) is a short elegy that stays with you.
This week's reads keep returning to the question of what we do with loss, limits and change, and to the stories we tell about them. In philosophy, Aeon's "Love with nowhere to go" is the standout: a personal essay that widens grief beyond death to cover estrangement, illness and other losses nobody recognises. Psyche's "When giving up teaches you more than persevering ever could" and Philosopheasy's "Count the Grains Before They Become a Heap" make smaller, lighter versions of the same move, treating quitting and moral drift as things worth looking at closely.
In history and culture, the strongest pieces correct popular myths. Asterisk Magazine's "Why Don't We Live in an Art Nouveau World?" offers a theory of why scalable styles win. ACOUP's "Teaching Paradox: Hearts of Iron IV, Part IVb: Iron Ends" uses a video game to show what WWII's powers were fighting for. Works in Progress's "Roman love" and Aeon's "How to be a (Roman) man" take apart modern ideas about Roman sexuality and masculinity. In science, Quanta's "How Voting Is Like a Klein Bottle" is the clearest explainer. Its spider-silk and interstellar-medium stories are good too, and Nautilus's graviton interview is a bold but lighter read. In arts and literature, Astral Codex Ten's Smollett review and The Paris Review's piece on Cynthia Buchanan's *Maiden* are the liveliest. Poetry runs from Stevenson's "Windy Nights" at Dead Poets Daily to Sarton's poem at The Marginalian.
AI and economics share a more anxious thread. Noahpinion pairs a hopeful piece ("A golden age of science is coming") with an alarming one ("The internet breaks"). The Pragmatic Engineer's "What's the ROI of software engineers' spending on AI?" is the most grounded of the group, though it is cut off by a paywall. In economics, Conversable Economist's "A Supply-Shock Era?" and "Financing the AI Physical Capital Buildout" are useful summaries, and Aeon's "How to build real peace" is the most vivid. Several of these are digests of other people's work, so the original sources are worth following up.
35 articles · by category
AI11
Why this one
It directly speaks to AI alignment and security concerns, illustrating how subtle misalignments or hidden capabilities can persist through model distillation pipelines even when training text contains no semantic cues.
- Fit
- 8
- Subst.
- 7
- Orig.
- 7
- Impact
- 7
- Writing
- 6
Beyond Owls: Subliminal Learning Can Transfer Learned Capabilities and Backdoors
A striking empirical demonstration that subliminal learning can transfer complex capabilities, conditional backdoors, and agentic hacking behaviors through semantically unrelated data.
Why this one
It offers concrete empirical insight into AI control techniques while thoughtfully questioning whether patching monitors genuinely improves long-term alignment or merely masks underlying risks.
- Fit
- 7
- Subst.
- 7
- Orig.
- 7
- Impact
- 6
- Writing
- 7
Screwtape: Adversarial Transcript Editing for Monitor Red-Teaming
A grounded technical breakdown of automated red-teaming for AI monitors, offering valuable methodological honesty about the limits of LLM-as-judge evaluation.
Why this one
It offers an insightful epistemological perspective on how machine learning actually progressed without rigorous theory, directly addressing the core problem of modern AI evaluation.
- Fit
- 8
- Subst.
- 6
- Orig.
- 7
- Impact
- 7
- Writing
- 7
He Wants AI Benchmarking To Be a Science, Not a Dark Art
A sharp, accessible interview that cuts through standard complaints about AI benchmarks to explain why they historically worked and why modern LLMs finally broke them.
Why this one
It provides rigorous empirical evidence on the difficulties of transferring safety-relevant reasoning across models, directly informing debates on capability diffusion and recursive self-improvement.
- Fit
- 8
- Subst.
- 7
- Orig.
- 6
- Impact
- 6
- Writing
- 6
Projection sampling preserves capabilities but diminishes gains from alignment fine-tuning
Worth reading for researchers interested in empirical safety distillation, as it delivers a clean negative result on applying projection sampling to alignment reasoning.
Why this one
It directly tackles core AI alignment theory, re-evaluating foundational MIRI-era problems like corrigibility through modern concepts such as shard theory and singular learning theory.
- Fit
- 9
- Subst.
- 5
- Orig.
- 5
- Impact
- 4
- Writing
- 4
Three perspectives on corrigibility - shards, simulators, singular learning
An accessible bootcamp essay applying contemporary alignment concepts to the classical corrigibility problem, though weighed down by massive formatting noise and somewhat loose formal analogies.
Why this one
It offers rare, ground-level qualitative insight into the ecosystem of Chinese AI alignment, a pivotal yet opaque dimension of global AI governance.
- Fit
- 8
- Subst.
- 6
- Orig.
- 7
- Impact
- 7
- Writing
- 6
Headwinds Everywhere (in the Chinese AI Safety Field)
A concise, illuminating field note that exposes the practical, financial, and political roadblocks facing AI safety researchers inside China.
Why this one
It touches directly on core alignment debates, timelines to recursive self-improvement, and the security implications of autonomous frontier models.
- Fit
- 7
- Subst.
- 6
- Orig.
- 5
- Impact
- 5
- Writing
- 6
AI #190: Scion of the Times
While it offers sharp, opinionated commentary on recent safety debates and rapid automation, this is fundamentally an ephemeral weekly link roundup that is worth skimming rather than reading closely.
Why this one
It directly targets the core AI alignment debate over digital sentience, model welfare, and how different laboratory training regimes influence model self-understanding.
- Fit
- 8
- Subst.
- 6
- Orig.
- 6
- Impact
- 5
- Writing
- 7
What’s up with GPT’s welfare?
A thoughtful exploration of how post-training and character engineering shape model self-reports on consciousness, moral patienthood, and welfare trade-offs.
Why this one
It addresses a central strategic debate in AI governance and effective altruism: how to explain catastrophic risk to the public and policymakers without losing credibility.
- Fit
- 8
- Subst.
- 5
- Orig.
- 4
- Impact
- 5
- Writing
- 6
The “Last Mile” Problem in AI Safety
A sensible but largely familiar critique of AI safety messaging, arguing that catastrophic risk advocates must articulate concrete physical mechanisms rather than relying on sci-fi abstractions.
Why this one
It addresses a central, recurring trope in AI governance debates with historical receipts and public choice economics, speaking straight to rationalist and effective altruist concerns.
- Fit
- 9
- Subst.
- 7
- Orig.
- 6
- Impact
- 7
- Writing
- 8
AI Safety Is Not A Corporate Plot
A sharp, evidence-rich takedown of the prevalent cynical narrative that AI safety advocacy is just a 4D-chess plot for regulatory capture.
Why this one
It brings together the reader's interest in moral philosophy and AI alignment, and it does so carefully and with historical depth.
- Fit
- 9
- Subst.
- 7
- Orig.
- 7
- Impact
- 7
- Writing
- 7
The Dream of an Aristotelian Machine
A thoughtful essay connecting Aristotle, MacIntyre and Anscombe to present-day character training in AI, though its conclusion is more a call for further work than a firm argument.
Arts & culture2
Why this one
It's slight, but well made. It's a finely observed literary miniature that brings Milton and Joy Williams into a reflection on attention and devotion, the kind of beautifully written cultural essay this reader values.
- Fit
- 7
- Subst.
- 6
- Orig.
- 7
- Impact
- 4
- Writing
- 8
My Tamagotchi Diary
A short, wry, quietly moving diary piece that turns a toy into a meditation on care, attention and failure.
Why this one
It suits a reader who likes art history and cultural essays: a vivid, compact look at a forgotten figure, with a neat observation about how art becomes a commodity.
- Fit
- 7
- Subst.
- 5
- Orig.
- 6
- Impact
- 4
- Writing
- 7
Rudolf Kalvach’s Wiener Werkstätte Postcards (1907)
A short, well-written piece that brings back a little-known Viennese graphic artist and makes a sharp point about collectible art that was never sent, though it is thin on analysis.
Economy5
Why this one
It directly addresses the macroeconomics of AI and tech infrastructure history, examining financial vulnerability across major AI hyperscalers.
- Fit
- 8
- Subst.
- 6
- Orig.
- 5
- Impact
- 6
- Writing
- 7
AI vs. the railroads
A sharp, accessible macroeconomic check on how the AI infrastructure buildout is shifting toward debt financing and historical bubble risks.
Why this one
It's a readable explanation of a current Nobel that puts it in the history of economic ideas, aimed at curious generalists rather than specialists.
- Fit
- 7
- Subst.
- 6
- Orig.
- 4
- Impact
- 6
- Writing
- 7
The Nobel Prize to Hall and Woodford: A Macroeconomic Journey
A clear, accessible history of macroeconomics that explains why Hall and Woodford won the Nobel, though it leans heavily on the Nobel committee's own background paper.
Why this one
It joins AI economics, psychology and the history of technology in vivid, humorous prose, which is the kind of cross-disciplinary essay this reader enjoys.
- Fit
- 8
- Subst.
- 6
- Orig.
- 6
- Impact
- 6
- Writing
- 8
Claude, don’t touch my lump
A witty, readable case against AI-unemployment panic that ties labour economics to the psychology of desire and imagination, though its argument leans more on charm and analogy than on rigorous evidence.
Why this one
It offers an engaging, intuitive history of economic thought detailing how modern central banking models married rational expectations with real-world price frictions.
- Fit
- 6
- Subst.
- 6
- Orig.
- 4
- Impact
- 5
- Writing
- 7
A Nobel for New Keynesian macro, at last!
A clear, accessible primer on the evolution of New Keynesian macroeconomics and central banking, though it largely covers standard textbook history.
Why this one
It is worth reading even though it plays down the costs and leans heavily on existing scholarship. It fits the reader's interest in economics, policy and progress studies: it gives a concrete, well-evidenced and counterintuitive account of how institutional incentives drive growth.
- Fit
- 8
- Subst.
- 7
- Orig.
- 6
- Impact
- 7
- Writing
- 8
China’s growth machines
A vivid, well-sourced essay arguing that China's growth comes from competition between local governments rather than central planning.
History5
Why this one
It combines sharp literary and historical analysis of a controversial war memoir with deeply felt narrative nonfiction about stoicism, duty, and vulnerability.
- Fit
- 8
- Subst.
- 7
- Orig.
- 7
- Impact
- 5
- Writing
- 8
Your Book Review: Storm Of Steel
A moving and unusual book review that pairs a close textual analysis of Ernst Jünger’s World War I memoir with a raw personal account of surviving a premature birth in the NICU.
Why this one
It's worth reading even if you've never played the game. It brings together history, historiography and cultural criticism in a long, reasoned argument backed by scholarly sources, which is the kind of work this reader values.
- Fit
- 8
- Subst.
- 8
- Orig.
- 7
- Impact
- 7
- Writing
- 8
Teaching Paradox: Hearts of Iron IV, Part IIIb: The Opposite of Love
A forceful, well-sourced essay that uses a strategy game to make a serious case about how culture remembers the war's atrocities.
Why this one
It joins classical history, literary sources and present-day cultural argument in a clear long-form essay, a mix that suits this reader's interest in history and ideas.
- Fit
- 8
- Subst.
- 6
- Orig.
- 6
- Impact
- 5
- Writing
- 7
How to be a (Roman) man
A well-sourced, readable piece of classical history that pushes back on the manosphere's picture of Rome, though its main argument lands without much surprise.
Why this one
It fits the reader's interest in progress studies and the history of technology, delivered as vivid, accessible long-form writing with real science in it.
- Fit
- 8
- Subst.
- 6
- Orig.
- 6
- Impact
- 5
- Writing
- 8
Everything rhymes with porcelain
A charming, well-told essay that links the history of porcelain to modern materials science, though the material on modern ceramics reads more like a sketch than an argument.
Why this one
It combines deep human biography with the history of technology and scientific discovery, themes central to your appreciation of progress and intellectual history.
- Fit
- 7
- Subst.
- 6
- Orig.
- 5
- Impact
- 5
- Writing
- 8
The Madman Who Saved Mothers: The Invention of Handwashing and the Bittersweet True Love Story Behind ‘Frankenstein’
A lyrical, poignant weaving of biography and the history of science that connects Mary Wollstonecraft's fatal childbirth fever to Ignaz Semmelweis's tragic crusade for antisepsis.
Philosophy3
Why this one
It addresses core rationalist and alignment decision theory themes—such as updateless decision theory, credible commitments, and counterfactual bargaining—using formal non-cooperative game theory.
- Fit
- 7
- Subst.
- 6
- Orig.
- 5
- Impact
- 4
- Writing
- 4
Exploring Intrapersonal Game Theory
A conceptually interesting formal note on modeling intrapersonal coordination and credible threats via game theory, marred by unparsed MathJax CSS artifacts and exploratory conclusions.
Why this one
It offers readers passionate about logic, philosophy of mathematics, and epistemological foundations a fresh, practice-based perspective on Gödel's incompleteness theorems.
- Fit
- 8
- Subst.
- 7
- Orig.
- 6
- Impact
- 6
- Writing
- 8
The ladder of knowledge
A lucid and beautifully narrated intellectual history that reframes Gödelian incompleteness not as an epistemic dead end, but as an open-ended constitutional practice of linguistic extension.
Why this one
It offers an articulate, historically grounded survey of philosophy of mind and metaphysics that directly addresses the hard problem of consciousness.
- Fit
- 8
- Subst.
- 7
- Orig.
- 6
- Impact
- 6
- Writing
- 8
Is everything conscious?
An exceptionally clear, thoughtful essay tracing how an uncompromising commitment to rational explanation leads directly to panpsychism.
Poetry1
Why this one
It fits the reader's interest in contemporary poetry and how poems get made, and offers a short reflection on how elegy works.
- Fit
- 7
- Subst.
- 5
- Orig.
- 5
- Impact
- 4
- Writing
- 7
Making of a Poem: Geoffrey G. O’Brien on “Ray of Light”
A short, affecting interview about writing an elegy, with some nice close-reading moments, though it's slight and the poem itself isn't included.
Psychology3
Why this one
It's a well-written personal essay on identity, choice and how social conditions limit decisions, and it complicates both sides of a polarised debate. It's a narrative rather than a sustained argument.
- Fit
- 6
- Subst.
- 6
- Orig.
- 7
- Impact
- 5
- Writing
- 7
Deciding to detransition
A candid first-person account that offers an unusual angle on detransition: the reasons it describes are social and economic rather than ideological, which makes it worth reading.
Why this one
The reader cares about psychology and decision-making, and this piece shows how a research paradigm changes the way we think about willpower and inactivity.
- Fit
- 7
- Subst.
- 6
- Orig.
- 5
- Impact
- 6
- Writing
- 7
How mental disorders play havoc with our motivation
A clear, humane explainer that draws on the author's own meta-analysis, though it stays mostly at a popular level and covers familiar ground.
Why this one
It offers an engaging synthesis of systems biology and embodied mind theory to counter computationalist hype around AI and mental outsourcing.
- Fit
- 7
- Subst.
- 6
- Orig.
- 5
- Impact
- 6
- Writing
- 7
Deskilling the body
A well-written defense of embodied cognition that challenges the Cartesian assumption that human intelligence can be separated from physical action.
Science2
Why this one
It will appeal to readers interested in biological game theory, decentralized information processing, and empirical research bridging economics and evolutionary biology.
- Fit
- 6
- Subst.
- 6
- Orig.
- 6
- Impact
- 6
- Writing
- 5
How Do Fungi Keep the Planet Healthy?
While the conversational podcast transcript format dilutes the depth, the discussion offers fascinating scientific insights into how brainless organisms execute strategic economic behaviors.
Why this one
It offers an accessible, well-grounded look at cutting-edge astrophysics and the ongoing empirical debate between dark matter models and modified gravity theories.
- Fit
- 7
- Subst.
- 6
- Orig.
- 5
- Impact
- 6
- Writing
- 7
Triplet of ‘Ghostly’ Galaxies Hints at What Dark Matter Is — and Isn’t
A lucid science dispatch explaining how three rare, dark-matter-deficient galaxies provide compelling evidence that dark matter is a physical substance rather than an artifact of modified gravity.
Social science1
Why this one
It directly addresses a major strategic question within AI safety and rationalist communities—how to engage with Chinese researchers and thinkers—while avoiding superficial cultural essentialism.
- Fit
- 8
- Subst.
- 6
- Orig.
- 6
- Impact
- 6
- Writing
- 7
A Chinese Perspective on “On Social Reality in China”
A thoughtful, grounded critique of Western rationalist assumptions about China that offers useful context on censorship, intellectual subcultures, and AI risk discourse.
Technology2
Why this one
It offers a practical, working prototype at the intersection of AI capabilities, formal verification, and AI control, demonstrating how proof checkers can safely bound autonomous optimization loops.
- Fit
- 8
- Subst.
- 7
- Orig.
- 7
- Impact
- 7
- Writing
- 7
You Should Specify Your Code Unambiguously (and Prove That You Did)
A compelling, concrete demonstration of using LLMs alongside formal proof assistants to generate verified, ultra-fast code from high-level specifications.
Why this one
It touches the reader's interest in the history of technology and the hardware behind AI, but it reads as dense equipment-industry analysis rather than a broader essay.
- Fit
- 4
- Subst.
- 7
- Orig.
- 7
- Impact
- 5
- Writing
- 6
ASML Is Building a Hybrid Bonding Tool!
A sharp, patent-based industry analysis that's worth it for semiconductor watchers, but it's specialist, investor-focused and paywalled before its challenges section.
All caught up.
No stories match these filters.