Reading
Books, papers, and essays that shaped how I think about healthcare, technology, and what comes next. This list grows.
Healthcare, and what AI is doing to it
The books closest to the argument I am making.
The book that framed the question for a lot of us: can AI give clinicians back the time to be human with patients. Written before generative AI, which makes its predictions worth revisiting.
Wachter wrote the definitive account of healthcare's last technology disappointment, so his cautious optimism about this one carries weight. The clearest current picture of AI actually entering hospitals.
Reported from inside institutions trying to make AI work, including the parts that fail. A corrective to both the hype and the doom.
Understanding the technology
What it is, what it can do, and what it costs.
The most practical book on working with AI rather than being replaced by it. Mollick's jagged frontier is the framing I use most.
The containment problem, argued by someone who helped build the thing. Worth reading because it takes the risks seriously.
The economics: AI makes prediction cheap, and everything downstream of prediction gets reorganized. A clean way to think about which parts of healthcare change first.
A memoir and a history of computer vision at once. The human story behind the technical one.
The practitioner's book. What building on top of these models actually involves.
Thinking, deciding, and leading
How ideas arrive, and how decisions get made under uncertainty.
The adjacent possible. Innovation as recombination in connected environments rather than lone genius.
How to make hard decisions well. For anyone choosing between futures rather than optimizing the present.
Practical framing for leaders trying to use AI for thinking rather than just for output.
What it means to be human
The questions underneath all of this.
A novel about an artificial friend who observes human love more carefully than the humans do. Worth reading before designing anything that claims to be a companion.
A journey into consciousness. If we are going to build things that converse with us, it is worth understanding what we mean by a mind.
Papers and essays
Primary sources worth reading in full rather than in summary.
The most specific argument anyone has made about what AI could actually do for human health, including a section on compressing decades of biomedical progress into years. Worth reading alongside the risk literature rather than instead of it.
How agent systems get built and bounded in practice. If you want to understand what an orchestration layer actually has to do, start here rather than with the vision pieces.
Anthropic's CEO on where the technology is in its own development, and what that implies for how we should be handling it.
The published values framework for a frontier model. Useful reading for anyone thinking about what it means to give a system boundaries, which is most of the work in building one.
Technical, and worth the effort. LeCun's line of work on models that learn structure from observation rather than from labels, which is a different bet than the language models get most of the attention.
A plain-language map of model types for people making decisions about them. Useful for anyone tired of every tool being called AI.