The Dual-Use Instrument: AI's Golden Age of Breakthroughs and the Erosion of Human Cognition

Sam Kriss wrote that in 1931, Soviet neuropsychologist Alexander Luria trekked to the remote foothills of the Alai Mountains and discovered something profound: basic literacy didn't just teach people to read. Literacy rewired how they thought. Illiterate peasants grouped shapes by lived experience (a circle was the moon, a square a drinking bowl). A few years of Soviet schooling flipped the switch.  These same people grouped by abstract geometry and solved hypothetical syllogisms about white bears in the Far North. Literacy birthed a new mind, one capable of abstractions, counterfactuals, and the kind of imaginative leap that fuels revolutionary politics and scientific progress.

Fast-forward to 2026. We're living through the mirror image of Luria's observation, but in reverse. AI chatbots, transformer models, and large language models (LLMs) are delivering superhuman feats of synthesis, prediction, and creation. Yet they're quietly undoing the cognitive habits that literacy once instilled. The same tools that are saving doctors time and accelerating Nobel-worthy science are fostering what one researcher calls "cognitive surrender" - the outsourcing of reasoning itself.

An interesting case is this Substack post, about a general practitioner (GP) medical doctor and digital health leader. The author dove headfirst into ambient AI scribing for 18 months. The promise was intoxicating: walk into a 10-minute consultation, maintain eye contact, speak naturally, and emerge with a complete, structured note. Burnout dropped. Documentation time plummeted by ~26%. It felt like getting the consultation back.

Then the second-order effects hit. Consultations stretched because he stopped curating in real time—why steer when the machine records everything? Follow-up notes were accurate but ailien: comprehensive transcripts lacked his clinical voice, synthesis, or "illness scripts" (those experiential mental models GPs build through thousands of encounters). The act of writing the note had been doing invisible cognitive work, prioritizing, reflecting, & reasoning. Offloading those thought processes broke the feedback loop. The doctor stopped using the tool not because it failed, but because it succeeded too well, causing subtle, deleterious side-effects.

We're not returning to pre-literate sensory immediacy. We're entering something stranger: a world of infinite generated content where direct experience is mediated by AI, and where abstract, counterfactual thinking is disappearing.

The Golden Age: What AI Has Already Delivered

AI breakthroughs are breathtaking.

Transformer models and LLMs have supercharged science in ways that were science fiction five years ago. AlphaFold and its 2024 Nobel Prize-winning successors, including AlphaFold 3 solved the 50-year protein-folding problem, predicting structures for over 200 million proteins with near-experimental accuracy. It now models how proteins interact with DNA, RNA, small molecules, and ions, unlocking rational drug design at unprecedented speed. AI-powered pipelines are delivering drug candidates 75% faster than traditional methods. Autonomous AI agents like Kosmos are compressing six months of PhD-level research into a single 12-hour run.

In materials science, AI has screened millions of candidates to discover new battery chemistries, carbon-capture materials, and quantum-computing components. Weather models run with hyper-accurate long-range forecasts. Self-improving labs iterate experiments in real time, slashing waste and cost. The 2024 to Q1 2026 period alone saw AI contribute to breakthroughs in fusion plasma control, exoplanet classification, and even new mathematical insights.

Economically, the gains are compounding. Generative AI is projected to add 1.5% to U.S. GDP by 2035, rising to 3.7% by 2075 through productivity alone. Industries with high AI exposure saw 10% productivity jumps, 3.9% job growth, and 4.8% wage increases in 2024-2025. Documentation tools like the one Gooch used are slashing administrative drag across veterinary medicine, human medicine, law, and engineering. Code generation, creative ideation, and data synthesis are freeing humans for higher-order work.

Many of these concrete accomplishments are not hype cycles. They are measurable, peer-reviewed revolutions in capability.

The Regressions: Literacy, Rationality, and the Extended Mind

Just as writing once pulled minds into abstraction, AI is pulling them back toward “fake it til you make it” fluency without friction, answers without effort, and sensory immediacy without synthesis.

Literacy rates and deep reading have been sliding since 2014 and the back slide is accelerating with AI. Elite university students increasingly can't finish a novel or parse a complex sentence without AI. One Kansas study found English majors struggling with Dickens' Bleak House—treating a metaphorical Megalosaurus as a literal dinosaur in Victorian London. These students respond like Luria's illiterate peasants: tethered to immediate, concrete reality.

Cognitive science gives insight into the relationship between literacy and cognition. "Extended mind" theory shows that writing isn't just recording thoughts; writing completes these thoughts. Gooch's experience mirrors these findings: AI notes broke the loop that built clinical expertise. Broader studies show AI users exhibit "cognitive offloading," that in turn causes declines in working memory, analytic reasoning, and critical thinking. People accept faulty AI outputs 73% of the time in controlled experiments lending evidence to the deleterious effects of "cognitive surrender.” Attention, skepticism, and the scientific method itself erode when LLMs do the synthesizing.

Steven Pinker's Rationality (2021) warned that we aren't born rational; we build rationality through deliberate practice in logic, probability, Bayesian reasoning, and causal inference. AI short-circuits and prevents skill growth from practice. Why wrestle with a syllogism when the model answers instantly? Why cultivate skepticism when the output is fluent and confident? The Enlightenment ideals, including empiricism, skepticism, & evidence-based progress rely on the very cognitive muscles we're letting atrophy with AI.

Backsliding in Human Potential

This phenomenon is not only about doctors or students. It's about the quiet diminishment of what makes a life meaningful.

In the arts, AI generates poems, paintings, and music at superhuman volume. Why labor through the frustration of original creation when the machine delivers polish? Personal artistic potential, the struggle that forges voice and vision fades in our “Age of AI.”

Economically, wealth creation has always come from human ingenuity: spotting unseen opportunities, iterating through failure, building novel systems. When AI handles the ideation and execution, the incentive to cultivate deep expertise or entrepreneurial grit weakens.

Philosophically, the great explorations of ethics, metaphysics, epistemology demand solitary wrestling with ambiguity. AI offers instant summaries and counterarguments. The joy of building your own worldview erodes. We risk a generation fluent in AI outputs but starved of the internal scaffolding that once produced  Socrates, Newton, Kant, Einstein, da Vinci.

Politics itself may regress from reasoned debate over imagined futures to tribal sensory immediacy. Streamers repeat formulas rather than abstract justice.

A Realistic, Guardedly Optimistic Future

None of these cognitive declines and their effects on society are inevitable.

We stand at an inflection point with enlightenment, wisdom, and agency. The post-literate age does not inevitably mean cognitive collapse. The “Age of AI” can mean augmented humanity if we treat AI as a seductively dangerous instrument, not a prosthesis.

Imagine "AI literacy" curricula that teaches not just prompts, but deliberate cognitive preservation: handwritten notes alongside AI drafts, Socratic challenges to model outputs, mandatory "unplugged" reasoning drills. Doctors like Gooch can (and some do!) use AI for transcription while reclaiming note authorship as a sacred clinical craft. Scientists could let agents run rote experiments while humans focus on the counterfactual leaps AI can't yet replicate.

The same transformers accelerating AlphaFold could help restore rationality.  Personalized tutors that drill children in skepticism, Bayesian thinking, tools that flag our own cognitive biases in real time. Economic gains could fund equal universal infrastructure: clean air, clean water, sanitation, energy, roads, railways, & telecommunication, that could free more people wealth creation, and personal “pursuit of happiness,” including creative and philosophical pursuits.

In the end, the future will not be purely post-literate or pre-AI literacy. It will be hybrid: a new extended mind where AI handles the volume and humans reclaim the voice, the synthesis, the purpose. The kids raised by AI dolls won't inevitably lose abstraction; they will almost certainly evolve a meta-literacy we cannot yet imagine.

We know that society shapes thought. Now you and I get to shape this society. The instrument is in our hands. We must use it carefully to cut away drudgery while sharpening the blade of the human mind. The breakthroughs and progress are themselves accelerating.  

The Death of Ivan Ilyich and other Stories by Leo Tolstoy (1886)

I never read any Russian authors while growing up because of the Cold War, so Russian literature was completely new territory for me when I retired. We had read a translation of Aleksandr Solzhenitsyn’s A Day in the Life of Ivan Denisovich in school, but that was framed more as Cold War history than literature. After finishing Anna Karenina (which disappointed me; see my review here: https://mitchwyle.blogspot.com/2024/11/anna-karenina-by-leo-tolstoy-1875-1878.html) I complained about it over dinner with friends. One of them suggested I try Tolstoy’s short stories instead, so I picked up this collection.

The stories are technically excellent. The plots are tight and strong, the characters are painted in rich three-dimensional detail, and there are moments of sharp humor that I genuinely enjoyed. This volume includes the famous novella The Death of Ivan Ilyich along with several other shorter pieces that showcase Tolstoy’s skill at concise storytelling.

What wore me down, however, was the relentless moralizing. Every story hammers the same theme: oppressed Russian peasants suffer under the weight of blind, self-serving oligarchs and bureaucrats, while the powerful remain oblivious to the obvious immorality and unsustainability of the system. The Christian theological preaching and sense of outrage felt heavy-handed and repetitive. Unlike Anna Karenina, these stories do not linger on the neurotic inner lives of the elites, but the sermon-like tone still grated on me.

I am glad I read the collection for the craftsmanship and the characters, but it has put me off Tolstoy entirely. I will not be moving on to War and Peace or his other major works. I may explore Pushkin next, though I worry I lack enough historical context to appreciate him fully.

3/5 stars.

First Command by Michael Simon (2025)

I picked up the first volume because the series has been getting press as a fun, accessible space opera. Unfortunately readers' complaints about the weak science turned out to be much worse than readers said. Spaceships list and roll like sailing vessels, belch smoke and flames in hard vacuum, and crash into asteroids that are packed so closely they might as well be in a traffic jam. The sheer volume of basic physics violations made it impossible to stay immersed.
The story itself is also relatively weak. The antagonists are cartoonishly one-dimensional — evil simply because they are evil — while the protagonists are much better drawn: heroic, self-doubting, and colorful. The core space-opera premise could have been enjoyable if the science had not been so distractingly awful, but the writing never rises above serviceable.
2/5 stars. Disrecommend

Look Homeward, Angel by Thomas Wolfe (1929)

This book was my first Thomas Wolfe novel, which feels embarrassing to admit because he was my mother’s favorite author. I still have her copy of You Can’t Go Home Again sitting on my shelf, untouched. In high school American Literature we read only excerpts, but we did learn the colorful details: Wolfe was a giant of a man (6'6", 2m), used a refrigerator as a writing desk, filled yellow legal pads with a Cross pen, writing just 3–4 words per line, every third line, and shipped entire orange crates of manuscript pages to his editor, who would eventually telegram him to stop so they could carve a book out of the mountain of prose.

Look Homeward, Angel is a classic “roman à clef” — essentially Wolfe’s own life turned into fiction. The protagonist Eugene Gant is a thinly veiled version of Wolfe himself, and the large, chaotic family, the small Southern town (based on his native Asheville, North Carolina), and many of the events are drawn directly from his upbringing, though heavily exaggerated and melo-dramatized.

The novel is driven by Eugene’s intense longing for meaning, his obsession with time, memory, and mortality, and a soaring, almost mystical romanticism. Wolfe’s prose is dense, lyrical, and richly descriptive — it frequently reminded me of Proust in its lush detail and of Thomas Hardy in its emotional weight. The writing is immersive and often beautiful, even if the book itself is very long and occasionally meandering. This book is a powerful, passionate, and deeply personal coming-of-age novel. It is worth reading for the sheer force of the language alone.

4/5 stars.

Children of Strife by Adrian Tchaikovsky (2026)

This fourth volume in Tchaikovsky’s wildly inventive Children of Time space opera series continues the story with a universe overflowing with grotesque, wildly imaginative biology. This time we encounter planetary-scale living systems, distributed intelligences, and growth rates 50 million times faster than physics would allow. The rapidly growing and morphing biology strongly reminds me of Neal Asher and his vivid, vivisection-like obsession rapid transformation.

Getting past the hard-to-swallow science, the story and character arcs remain strong. Tchaikovsky writes compelling characters you end up genuinely caring about, even when the overall tone stays bleak and pessimistic. I do not enjoy dystopian stories as a rule, but the quality of the writing and the depth of the personalities kept me invested in what happened to them.

That said, the heavy pessimism and the particular brand of biological horror on display this time made this my least favourite entry in the series so far. It is still well-crafted and thought-provoking, but it did not click with my taste.Skip this one unless you are a committed fan of the series and do not mind dark, visceral, and deeply strange far-future biology.

3/5 stars.

A brief history of intelligence by Max Bennett (2023)

This was another recommendation from Senthil that had been sitting in my queue since December 2025. I expected the book to focus heavily on AI, using neuro-physiological and neuro-anatomical insights to guide the development of more human-like, self-improving systems. Instead, it delivers a broad evolutionary history of intelligence itself.

Bennett traces how life on Earth gradually evolved the ability to survive and thrive, beginning with social "politics" among early organisms and the manipulation of their environment. What we call cognition, sentience, qualia, and even our (possibly illusory) consciousness emerge as accidental side-effects of traits that simply proved useful for survival and reproduction. Intelligence, in this view, is less a grand design and more a useful byproduct of fitness.

I enjoyed the book a lot. Bennett does a good job addressing counterfactual evidence and the main controversies surrounding his claims. The references to AI and software are present but clearly secondary. When he does touch on artificial intelligence, he frames it as simulations of biological processes on our planet, which may not always be the optimal path for building the systems we actually want.The book is a thoughtful and well-reasoned exploration of how intelligence arose in nature.

5/5 stars.

Ensh*ttfication by Cory Doctorow (2025)

I've been skimming Cory Doctorow's daily links and rants on Pluralistic for years, so I walked into this book already fluent in most of his ideas. What surprised me was how much richer and more coherent everything became once he had an entire book to breathe. Ensh*ttfication is the definitive deep-dive into the "Ensh*ttocen" - the Great Ensh*ttening era we're all living through, laying out his core theses with far more historical context, evidence, and narrative force than any blog post could manage.

The book is long, dense, and occasionally meanders into broader territory: labour rights, liberty, social justice, and the politics of technology. I actually loved those side-quests. They served as powerful reminders of how extraordinarily lucky I am to have the freedoms and tools I do. Doctorow's sales pitch for Mastodon and the Fediverse was particularly interesting, even if reality has already moved on: walled gardens have largely won, gateways have been shut down, and most of the people I actually talk to will never install Mastodon. (The one bright exception I'm excited to try is matterbridge — https://github.com/42wim/matterbridge. The universal chat client (theoretically) lets one client bridge WhatsApp, Slack, Discord, Telegram, Teams, and more. If I ever find the time, I'll clone, build, and try it.

At its heart, this book delivers a razor-sharp diagnosis of exactly how enshittification works: a three-stage pathology that turns once-great platforms into "giant piles of shit." First, services are genuinely good for users to draw them in. Then they abuse those users to please business customers (think ad-choked feeds and algorithmic manipulation). Finally, they squeeze the business customers themselves to extract every last drop of value for executives and shareholders. Doctorow backs this up with devastating case studies—Facebook, Amazon, Google, Uber, TikTok, and more, showing how weakened antitrust rules, regulatory capture, lock-in tactics, and "twiddling" algorithms created today's techno-feudal system of rent extraction. He doesn't stop at diagnosis; he also maps out concrete cures: aggressive interoperability mandates, breaking up monopolies, stronger data rights, and rebuilding the open internet we once took for granted.

I don't always agree with Cory's values or speculations, and I occasionally have counter-evidence that undercuts some of his stronger claims. Yet I never resent the ride. His extreme points of view and the mountains of data he cites are consistently thought-provoking and valuable. The book is daunting in scope, but it's also comprehensive, well-crafted, and deeply informative. I'm genuinely glad I ploughed through every page.

The book is highly recommended if you want to understand the forces quietly degrading the internet (and society) around us.

5/5 stars.

X Minus One Old Time Radio Episodes (1955)

Similar to the Perry Rhodan space opera archives over which I stumbled and am now enjoying in serialized book format, the old Astounding stories from the 1950s were dramatized into Radio plays.  These Radio Plays were digitized and made available by fans on the public Internet. Some of the stories are timeless and others I had read are very entertaining as audio theater.  4.5 Stars.

The Winter Queen (Азазель1 Azazel) by Boris Akunin (Борис Акунин)

This book was recommended by Swiss relatives who loved the mystery and spy thriller, as well as the historical accuracy of the events.  The AI's all agree that the English translation is slightly better than the German translation so I read the English one.  I agree that the somewhat contrived and "James Bond"-style super-spy vs super-villain thriller events are entertaining.  The Mystery and historiography are also very good.  However, I found the spy vs spy stuff and some of the cartoonish one-dimensional characters a little off-putting.  I may read more books in the Akunin series, but I did not really get deeply immersed in this one. The book would likely work better as a graphic novel or serialized comic book series.  3/5 Stars.

The Internet Con by Cory Doctorow (2022)

This book is a quick read and a precursor to Cory's book on Ensh*ttification he wrote later. This book has some details of the interesting political, legal, and economic underpinnings that incentivize the current state and trends in data sovereignty and computing.  Many of the topics in the book will be familiar to, and redundant for readers of Corey's daily blog entries.  For me personally it was good to get a comprehensive summary in one short volume.  I think there is a little too much hand-wringing, ranting, and Sturm und Drang than specific, prescriptive remedies for actions we should take. But the book is quite informative and short. 3/5 Stars.