The Book
The Future of AI in Medicine
Medicine does not need more fluent answers. It needs systems that show their work, declare their uncertainty, and survive contact with the ward at 3 AM.
This book is the argument. The acute stroke decision support tool is the proof. Both are written in the open and updated as the evidence moves.
Override, Don’t Defer
Augmentation fails the moment a clinician stops questioning the machine. The book argues that real human-AI collaboration means overriding the algorithm when the patient in front of you contradicts it.
Show the Work
Every recommendation must cite its source, declare its confidence level, and name what it does not know. Transparency is not a feature. It is a minimum viable requirement for any system that touches a patient.
The Algorithm Has No Conscience
A model trained on biased data does not become fair at deployment. Equity is not an aspiration — it is a debt that compounds with every algorithm shipped without auditing who it harms.
Why not just ask ChatGPT?
A large language model generates plausible text. It cannot tell you which randomized trial enrolled patients like yours, whether the evidence is Class I or Class IIb, or that a recent study contradicts the guideline it just confidently cited. Structured clinical decision support does something fundamentally different: it matches a specific patient against verified trial criteria, shows exactly which evidence applies, and says nothing when the data is insufficient. The gap between fluent and accountable is the gap between a convincing paragraph and a safe clinical decision.
A Note Before the Movie Starts
On writing a book about AI transparency — transparently, with AI.
Welcome to the Future: Why AI Will Redefine Medicine
Why AI will redefine medicine — told from 3 AM on a hospital floor where one physician, twenty-three medications, and four hundred pages of data collide.
The Machine That Learned to Learn: How AI Corrects Its Own Mistakes
What a sepsis prediction model deployed at hundreds of hospitals reveals about how AI actually learns — and how the field is building systems that catch their own errors.
Updated Mar 2026
The Diagnostic Revolution
When AI outperforms physicians 4-to-1 on the hardest diagnostic cases in medicine, is 'augmentation' still the right word? An investigation of the gap — and what lies on the other side.
Updated Mar 2026
The Anatomy of AI Failure — and the Architecture That Prevents It
What happens when compounding AI errors cascade through a hospital at 3 AM — and how biological immune systems reveal the architecture that makes clinical AI safe.
Updated Mar 2026
The Surgeon and the Machine: When AI Gets a Body
What happens when artificial intelligence stops advising and starts acting — holding the scalpel, navigating living tissue, making decisions at the speed of a heartbeat.
The Molecule as Patient: AI Reimagines Drug Discovery
When AI stops reading the body and starts designing for it — generating molecules no chemist imagined, testing them at speeds that compress decades into months, and confronting the deepest question in pharmaceutical science: why does a drug that works in a test tube fail in a human being?
Maria's Movie
You are sixty-one years old. You are lying on a gurney in a hallway. The machines are watching you. None of them are seeing you.
The Radiologist Who Multiplied
What happens when the machine can see what you trained a decade to see — and sees it faster, across more dimensions, in patients you will never meet? The answer is not obsolescence. It is multiplication.
The Therapist in Your Pocket: AI and Mental Health
A chatbot can reduce symptoms at scale and still be unacceptable where crisis medicine begins. Mental health AI only becomes thinkable once we stop calling companionship, therapy, and physiological inference the same thing.
Updated Mar 2026
The Algorithm's Conscience: Building Governance into AI Medicine
Consumer health AI is already making medical judgments at population scale. The urgent question is whether we can build governance worthy of its reach — and several promising frameworks suggest we can.
The Physician's Cut
You are the physician. It is 6:02 AM. The algorithm has filed its overnight report. Forty-three alerts. You have twelve minutes before rounds. Every alert is a frame from someone's movie. You are the editor who decides which frames the patient sees.
The Digital Twin Paradox
Your digital ghost gets sick before you do. Someone has to decide what to tell you. Welcome to the age of predictive medicine — where the most dangerous thing an algorithm can do is be right.
The Last Photograph
After ten chapters of movies, the book ends with a single frame. One physician. One patient. One room. The moment where computation ends and medicine begins.
Updated Mar 2026
The Biased Pixel
The system observes a physician in Room 3 deviate from protocol. It logs the anomaly. It does not understand what it is seeing.
This book is open. Support thoughtful AI in medicine.