Medicine can repair the body — but what about the part it can't fix? These four books each stand on a different boundary: the scientific, the mortal, the experiential, the ethical.
2026 · Book Recommendations · No. 23
Medicine is one of the modern world's most deified powers, yet it has clear limits. Mukherjee writes cancer's biography and shows the enemy is no outside invader — it's a Darwinian mutiny staged by our own cells. Gawande reveals that medicine is trained only to "fix," so when it can't, its default reflex of more intervention actually robs us of a good ending. Kalanithi goes from doctor to patient and finds that third-person knowledge can't answer "how should I live the time I have left". Skloot, through the HeLa cells, asks who bears the cost of medical progress and who reaps its rewards. By the end you'll hold a clearer map of what medicine can and cannot do.
| Book | Author | Year | What it nails |
|---|---|---|---|
| The Emperor of All Maladies | Siddhartha Mukherjee | 2010 | Cancer isn't an outside invader — it's our own cells gone Darwinian, using all of life's own tools, just with the brakes cut |
| Being Mortal | Atul Gawande | 2014 | Medicine is trained to repair the body, but the end of life is no engineering problem; when it can't fix you, more intervention steals the good ending |
| When Breath Becomes Air | Paul Kalanithi | 2016 | A neurosurgeon diagnosed with lung cancer at 36 — going from the one holding the knife to the one lying down; he spends his last time answering what knowledge couldn't |
| The Immortal Life of Henrietta Lacks | Rebecca Skloot | 2010 | A poor Black woman's cells, taken without consent, built modern biomedicine — while her family had no idea and couldn't afford a doctor |
We're used to imagining disease as invasion: bacteria, viruses storming in from outside. The most subversive premise of Mukherjee's "biography" is that cancer isn't an invader — it's us. It's a normal cell whose growth switch gets jammed permanently to "on," then begins dividing, copying and spreading without restraint. Cancer uses all of life's own tools — proliferation, evolution, adaptation — just with the brakes cut.
Writing it as a "biography" isn't a rhetorical flourish. Mukherjee traces four thousand years, from Egyptian papyri recording breast tumors as "no treatment," to the radical mastectomies of the 19th century (cut more, cut deeper, on the belief that thoroughness would win), to chemotherapy, targeted drugs and immunotherapy. The through-line of this history is humanity repeatedly misjudging the enemy: believing cancer was local and could be cut out at the root, when it was in fact systemic and molecular.
Why is it so hard to cure? Because cancer cells obey pure Darwinian logic. A dose of chemo kills 99% of cancer cells — and the surviving 1% are precisely the ones carrying drug-resistant mutations. You've just run a round of natural selection by your own hand, and the survivors are stronger. Cancer keeps evolving inside the patient; that is the root reason it is "immortal" and so hard to eradicate. It is, in essence, a complex adaptive system running inside your body.
But this is no book of despair. Mukherjee's other thread is how human understanding accumulates — each generation of doctors inching toward the truth through its errors, until cancer is seen as "a disease of the genome" and treatment shifts from "how much to cut" to "which mutation to target." Understanding the enemy's full biography is itself the precondition for defeating it.
It's enormous (nearly 600 pages), and the narrative occasionally bogs down in clinical detail. Published in 2010, it predates the past decade's immunotherapy breakthroughs (PD-1, CAR-T), so today's reader must supplement it. Mukherjee favors heroic narrative and devotes relatively little space to the suffering of patients behind the failed therapies.
Mukherjee's most useful move for a technical person is "write the enemy's biography before hunting for a fix." Facing a hard problem you're grinding on — an opaque new technology, a distributed system that keeps failing, even AI alignment — most people lunge straight at the solution. To try next week: pick one such problem, don't solve it, and spend two hours writing its "history of understanding" — how it was first discovered, how each generation misread it, why every "we've got this now" later collapsed. You'll likely find that what makes the enemy stubborn hides in the one point that keeps getting misjudged; see that line clearly and the direction of the fix surfaces on its own.
Gawande is a surgeon, and he admits it honestly: medical training taught him to repair the body but never to face the moment it can't be repaired. A doctor's default objective function is "keep the patient alive, longer." So as a person nears the end of life, the system's reflex is more tests, more chemo, more tubes. Each step looks like "for your own good"; together they form a chain of actions that pushes a person into the ICU and strips away their final time.
The root is a misaligned objective. The system optimizes for "safety and survival," but what the patient actually cares about is often something else: being able to use the toilet alone, to go home, to talk lucidly with family. Visiting nursing homes and hospice, Gawande found a counterintuitive fact: patients who choose hospice and forgo aggressive treatment sometimes live longer — and better, because medicine's over-intervention itself wears a person down.
What he offers isn't a slogan but a set of operational questions. When serious illness strikes, the thing to ask is not "what treatments can we still do" but four questions: What is your understanding of the situation? What are your fears and what are your hopes? What trade-offs are you willing to make, and which will you never make, for the sake of more good time? And which course best fits that understanding? These four hand decision-making back to the patient.
He also distinguishes two doctor–patient relationships: paternalistic (the doctor decides for you) and informative (the doctor lists options and you order off the menu). He argues for a third — "interpretive": the doctor helps you clarify what you truly value, then decides with you on that basis. This isn't only about dying; it's wisdom that holds in any situation where something can't be fixed, only traded off.
It rests on the American medical and elder-care system; the institutional detail of nursing homes and hospice doesn't map cleanly onto a Chinese family context. It says too little about the economic base a "good goodbye" requires — behind the four questions sits a set of care resources not everyone can afford.
Gawande's four questions are a tool you can use next week, and you needn't wait for a crisis. BigCat has aging elders nearby; rather than being rushed by a doctor during some hospitalization to decide "save them or not, and how," find a calm conversation now and earnestly ask the first two of the four: "What are you most afraid of?" "If you could keep only one thing, would you rather have more days, or a certain quality of life?" Write the answers down. When the day comes to make a medical decision on their behalf, what you hold isn't guesswork but the priorities they spoke aloud themselves — the most practical lesson in the book.
Kalanithi was 36, about to finish his neurosurgery training, the kind of person his cohort bet on most. Then he was diagnosed with advanced lung cancer. The book's singularity is that he stands on both sides of the boundary at once — both the doctor who understands the disease best and the patient lying in the bed. And he finds that all the knowledge the former accumulated cannot answer the latter's most urgent question.
That question is: when you know your time is short, how should you spend what remains? Medicine can tell him the statistical survival curve, but it cannot tell him "can I still be a surgeon," "should my wife and I have a child," "what, right now, is actually worth doing." This is the gap — uncrossable by data — between third-person knowledge (facts about the group "lung-cancer patients") and first-person experience (how "I" should live).
Kalanithi's response isn't to find answers but to keep acting. He returns to the operating table, turns to writing when his strength fails, and he and his wife have a daughter. You can never reach perfection, he writes, but you can believe in an asymptote toward which you ceaselessly strive — this became his stance toward death: meaning lies not at the endpoint but in the striving itself.
He died before finishing the book; the last chapter was written by his wife, Lucy. Its heaviest passage is his message to his infant daughter: you filled a dying man's days with a sated joy… in this time, right now, that is an enormous thing. This isn't about how to beat death — it's about how, with death already certain, meaning remains possible.
It's short, and the author died before he could revise it, so the second half feels somewhat unfinished. The vantage is highly personal — the way a man who holds medical training, literary sensibility and medical resources all at once meets death won't transfer cleanly to an ordinary patient in very different circumstances.
The boundary Kalanithi draws — third-person knowledge vs. first-person experience — is exactly the one most worth recognizing in the AI age. AI can give you infinite third-person knowledge: facts, methods, statistics, other people's experience in any field. But there's a class of questions that can only be answered in the first person: what matters to me, how I should spend my finite time, what's worth devoting the rest of my life to. To try this week: make a list of "questions AI can't help with" — pick out the ones you keep wanting to outsource to search and models but which only you can answer. Seeing that boundary is what keeps you, amid an ocean of answers, from losing the one question only you can answer.
In 1951, a poor Black woman named Henrietta Lacks died of cervical cancer. Doctors took her tumor cells without telling her and without her consent. Those cells became the first "immortal" cell line in human history able to divide indefinitely outside the body — HeLa. They underpinned the polio vaccine, in-vitro fertilization, the gene map, and countless studies of cancer and viruses.
The cruel contrast: HeLa cells were sold worldwide and spawned an industry, while Henrietta's family had no idea for decades and were too poor to afford a doctor. When they finally learned their mother's cells were "still alive, in labs across the globe," what met them wasn't compensation but more blood draws, more research, more confusion. Skloot's book restores a woman who'd been reduced to the code "HeLa" back into a person with a name, a family and a face.
At the level of mechanism, the book presses a question medical progress can't dodge: who bears the cost of progress, and to whom do the rewards go? Once your tissue leaves your body, does it still belong to you? Informed consent — today's ethical floor — was built up step by step precisely under the pressure of stories like HeLa's. It reminds us that behind every gleaming medical breakthrough there may stand a contributor who was never consulted and never remembered.
Skloot doesn't write it as a simple indictment. Through Henrietta's daughter Deborah, the book presents a more complex truth: anger, poverty, faith, and a longing to understand science mixed with fear of it. Deborah says you can't go into history with hate; you have to remember times were different — and that posture, of not being swallowed by hatred in the face of injustice, is the book's spiritual ground note.
The narrative shuttles between the Lacks family's story and the history of science, and some readers find the author's own presence (Skloot's relationship with Deborah) given too much weight. Focused on ethics and character, it explains relatively little about how HeLa cells actually work in the science.
The HeLa story distills a portable judgment tool — the "HeLa question": whose unconsented contribution is this progress built on? It's especially sharp in the AI age: today's large models are trained precisely on vast amounts of text, images and code used without explicit authorization. Whether you're an AI super-individual using the tools or someone assessing an AI company's long-term value, it's worth asking: is its capability windfall standing on a crowd of "HeLa-style contributors"? This is both an ethical judgment and a concrete risk assessment — copyright litigation, regulatory backlash and the legality of data provenance are becoming ever-heavier variables in AI valuations. Next week, when you look at an AI company, put "where did the data come from, and who consented" on your due-diligence checklist.
Gawande's distinction — between "to be alive" and "to live well." A simple test: the last major health/life decision you made for yourself or family, did it optimize "safety and duration," or what the person themselves said they valued? If you can't even name what the person values, the conversation that should have happened hasn't yet.
Mukherjee's lens: a stubborn enemy tends to win, again and again, at the same point that keeps getting misjudged. Test: can you clearly state how this problem "has been misunderstood across generations"? If not, you're most likely repeating a pit someone before you already fell into, without knowing it.
Skloot's HeLa question. It doesn't ask you to feel guilty, only to see. Qualified awareness is concrete: can you name at least one real group of contributors (data labelers, users whose data was collected, some unattributed labor), and whether they were treated fairly? Only after seeing can you speak of judgment and choice.