One of my old college professors, Ruth Wisse, just gave the Jefferson Lecture (the highest honor given by the National Endowment for the Humanities) at the tender age of 89. One brief quotation, about addressing the cultural drift to grievance over gratitude during Harvard faculty meetings:

The best I ever did, more than once, was to say, “Please remember that democracy is not transmitted biologically.”

That is from an argument referencing the essential daily Jewish prayer, the Shema, and the criticality of reinforcing important ideas more than just prioritizing novelty of content. Things can both be imperfect—even deeply flawed—and still worth celebrating.

If there is to be enduring government of, by, and for the people, the people would have to be instructed and reminded to respect and confidently to perpetuate their precious inheritance.

Nihilism, learned helplessness, and conspiracy don’t have to be a dominant cultural operating system.

// 03.30.26

From Effort to Intuition

03.18.26 // Miscellany, Radiology

True education isn’t just transmitting information. It’s the information filtered through experience that makes it real.

Experience matters because it allows us to convert all that expensive type-two thinking into long-term memory. It can, in a sense, help convert type-two into type-one thinking—or at least less expensive thinking—by taking a complicated world full of many discrete ideas and concepts and chunking them into a smaller number of discrete elements.

When radiologists read a scan, we do not evaluate each individual pixel as we gain experience; we are able to take in larger and larger swaths of structures as units for interpretation and pick up discrepancies that don’t match our mental model of what something should look like.

This is why chess grandmasters can, at a glance, recreate a chessboard after viewing it for only a few seconds, but novices can only remember a handful of pieces. It’s why “1-9-8-5” is a string of four numbers, but 1985 is my birth year.

One of the unappreciated components of a liberal arts education or general skills is that someone who is generally skilled or smart is actually somebody with a large volume of mental models and a large amount of relevant information stored in long-term memory that they are capable of bringing to bear on novel situations.

Learning to think is an organic, holistic process of taking on more and more things and doing the hard, meaningful-but-not-always-fun work to integrate them into the person you are. There is no shortcut for that. That’s why using AI to write something that’s actually important doesn’t really work for many people, and that’s one reason why a score on a standardized test is a helpful but woefully incomplete metric.

As a trainee, I disliked interpreting spine imaging. This is in part because there are a lot of discrete decision points to make at every level regarding canal and neural foraminal stenoses, the location of disc pathology, the degree of joint degeneration, and so on and so forth. That’s a lot of work if each decision requires deliberation, and it’s psychologically unfulfilling when your attending changes all your grades.

But then you read thousands of scans, and much of that type-two thinking is well chunked into a near-automatic type-one process. It doesn’t even feel like thinking—you see a canal, and you see a foramen, and you know what it is. That’s why, as an attending, reading a degenerative spine can have more in common with meditation than nearly any part of my job.

Learning curves are high, and there’s enough internal cognitive load that we can only learn so much at a time. When everything is new and challenging, everything is hard, and everything takes time and energy. Anthropic CEO Dario Amodei is, as yet, very wrong that “the most highly technical part of the job has gone away.” There is as yet no mastery of any topic—no inborn system-one ability or computer skill that you can simply offload parts of the scan to.

The experiences mediate the learning and move the learner down the pathway from effortful to intuitive abilities.

This is why, when a well-trained radiologist reviews a scan, only a small fraction of the findings actually require thought. The rest conform to the library of pattern matching they can bring to bear.

Compromise and Discomfort

03.13.26 // Miscellany

Optimizing is challenging (Herbert Simon argued it was impossible). Some people, in some situations, can optimize for their interests and find their magical right path (their “calling”). I never had a calling; or if I did, I missed it.

I think perhaps the equally important—if less interesting—advice is to satisfy yourself across the spectrum of minimizing excess discomfort.

If there are things that you don’t mind that other people dislike, that is a comfortable niche at the broad job level, the specialty level, and even in your specific role at your company. If some people find a certain aspect of something insufferable, but you find it irrelevant or potentially meditative, then double down. That’s where you add your unique value.

My wife and I bought a house in residency that was two bedrooms/two bathrooms and on the corner lot of a somewhat busy street. You couldn’t really hear traffic noise in the house or anything, and there was plenty of parking, but we were able to get an incredible deal on the house because many people don’t want to live on a corner, and many folks think they need more rooms.

Everyone compromises, repeatedly. Continually. Incessantly and unavoidably. But you get to choose some of those compromises, and some folks’ big line in the sand might be no big deal to you. A lot of personal value is tied up in noticing and seizing those opportunities.

From Developmental Editing by Scott Norton:

Authors who resort to derivative thinking usually don’t realize they’re doing so. They tend to fall into one of two categories: converts and preachers. The convert is one for whom an existing framework is a recently discovered window into reality.

Being a boring convert seems like an easy trap, which is perhaps no surprise because it’s so ubiquitous in popular (as well as presumably unpopular?) nonfiction.

// 03.04.26

Job Satisficing

03.03.26 // Miscellany

From Herbert Simon’s Models of My Life:

The real economic actor is, in fact, a satisficer, a person who accepts ‘good enough’ alternatives, not because less is preferred to more but because there is no choice.

Since my world picture approximates reality only crudely, I cannot aspire to optimize anything; at most, I can aim at satisficing. Searching for the best can only dissipate scarce cognitive resources; the best is the enemy of the good.

Combine the important economic concept of satisficing with the classic Goodhart’s Law: when a measure becomes a target, it ceases to be a good measure.

True optimization is essentially impossible, but even those futile attempts to select for specific features (compensation, time off, etc) can be fraught. These are all important factors of a job, to be sure. But they are also important because they are easily measurable and comparable.

In a competitive market, the grass is always greener somewhere, and there is a lot of green grass out there at the moment. One component of job satisfaction is, of course, attitude:

Hard days aren’t necessarily bad days, though they can be.
Sometimes a day so busy that even the bathroom seems too far away can nonetheless be filled with meaning and purpose.
Loafing is not an inherent good, free time doesn’t intrinsically bring happiness, and work—even extra work—doesn’t always have to mean frustration, tiredness, burnout, despair, etc.

You get to choose your attitude and whether you believe that what you do matters.

Whether you quiet quit or actually quit, just make sure you understand your role in the intolerable situation. It’ll help you make the most of your next opportunity.

From a short essay written by programmer Greg Knauss:

People will argue that speaking English to LLMs is just another level of abstraction away from the physics of how the machine actually works. And while that’s technically true — the worst kind of true — it also misses the point. Industrialization fundamentally changes things, by quantum degrees. A Ding Dong from a factory is not the same thing as a gâteau au chocolat et crème chantilly from a baker which is not the same thing as cramming chunks of chocolate and scoops of whipped cream directly into your mouth while standing in front of the fridge at 2:00am. The level of care, of personalization, of intimacy — both given and taken — changes its nature. Digging a trench is a very different thing than telling someone to dig a trench. Assembling a clock is a very different thing than asking Siri for the time.

// 02.27.26

A Problem With Doctors’ Software

02.26.26 // Medicine, Miscellany

Daniel Cook on Mixing Games and Applications:

Why do games have such a radically different learning curve than advanced applications? It turns out that games are carefully tuned machines that hack into human beings’ most fundamental learning processes. Games are exercises in applied psychology at a level far more nuanced than your typical application. Implicit in this description of interactivity is the fact that users change. More importantly, the feedback loops we, as designers, build into our games, directly change the user’s mind… The person that starts using a game is not the same person that finishes the game. Games and the scaffold of skills atoms describes in minute detail how and what change occurs.

This is a pretty big philosophical shift from how application design is usually approached. We tend to imagine that users are static creatures who live an independent and unchanging existence outside of our applications. We merely need to give them a static set of pragmatic tools and all will be good. Games state that our job is to teach, educate and change our users. We lead them on an explicitly designed journey that leaves them with functioning skills that they could not have imagined before they started using our application. Our games start off simple and slowly add complexity. Our apps must adapt along the user’s journey to reflect their changing mental models and advanced skills. Failure to do so results in a mismatch that results in frustration, boredom and burnout.

Interesting applications to enterprise software including EHRs, PACS, etc, where there is a distinct lack of gamification. This dovetails nicely with Jason Fried’s point about software design between making features obvious, easy, and possible.

Jony Ive, in a 2025 fireside chat at Stripe:

I think the spiritual thing is that I believe that when somebody unwrapped that box, and took out that cable and they thought somebody gave a shit about me. I think that’s a spiritual thing.

What used to depress me was this sense that solving a functional imperative then we’re done. But of course, that’s not enough. That’s not the characteristic of an evolved society.

Sir Jony argues even the packaging should reflect consideration. I’ve always thought the tight physical tolerance of Apple’s packaging was fantastic (and subsequently industry-changing, now ubiquitous). If only software would follow suit. I would say it is, on the whole, getting worse (Apple included).

// 02.24.26

Signal v Noise

02.23.26 // Miscellany, Reading

From The Signal and the Noise: Why So Many Predictions Fail-but Some Don’t by Nate Silver:

Distinguishing the signal from the noise requires both scientific knowledge and self-knowledge: the serenity to accept the things we cannot predict, the courage to predict the things we can, and the wisdom to know the difference.”

A clever formulation of the Serenity Prayer. There is a classic Mark Twain aphorism, “It ain’t what you don’t know that gets you into trouble. It’s what you know for sure that just ain’t so.” Ain’t that the truth:

One of the pervasive risks that we face in the information age, as I wrote in the introduction, is that even if the amount of knowledge in the world is increasing, the gap between what we know and what we think we know may be widening.”

[…]

“As the statistician George E. P. Box wrote, “All models are wrong, but some models are useful.” What he meant by that is that all models are simplifications of the universe, as they must necessarily be. As another mathematician said, “The best model of a cat is a cat.”…The key is in remembering that a model is a tool to help us understand the complexities of the universe, and never a substitute for the universe itself.”

The model is a tool, not a reality. If only we were better at remembering that most metrics and dashboards are proxies for things we care about and not the actual things we care about.

We are always filtering. The more inbound, the more we rely on crude and deforming filters.

Meanwhile, exposure to so many new ideas was producing mass confusion. The amount of information was increasing much more rapidly than our understanding of what to do with it, or our ability to differentiate the useful information from the mistruths. Paradoxically, the result of having so much more shared knowledge was increasing isolation along national and religious lines. The instinctual shortcut that we take when we have “too much information” is to engage with it selectively, picking out the parts we like and ignoring the remainder, making allies with those who have made the same choices and enemies ofthe rest.”

We are also, as Daniel Kahneman argued, always trying to construct causal explanations of the world. We understand life as “the steady flow of hindsight in the valley of the normal.”

“The most calamitous failures of prediction usually have a lot in common. We focus on those signals that tell a story about the world as we would like it to be, not how it really is. We ignore the risks that are hardest to measure, even when they pose the greatest threats to our well-being. We make approximations and assumptions about the world that are much cruder than we realize. We abhor uncertainty, even when it is an irreducible part of the problem we are trying to solve.”

We see this failure in Goodhart’s law & Taleb’s Black Swan. Conditional probability, when used appropriately, remains one of our best tools:

“Bayes’s theorem deals with epistemological uncertainty—the limits of our knowledge…What isn’t acceptable under Bayes’s theorem is to pretend that you don’t have any prior beliefs. You should work to reduce your biases, but to say you have none is a sign that you have many. To state your beliefs up front—to say “Here’s where I’m coming from”—is a way to operate in good faith and to recognize that you perceive reality through a subjective filter.”

The Bayesian is a huge part of making good decisions as a doctor.

Good Questions

02.19.26 // Miscellany

On the importance of the right questions:

The trouble with trying to solve questions posed by communities is that all the good ones are gone, especially if they can be answered. Why then do not we see the scientists around us spend their nights hunting for questions? Instead, it seems that having a clear question is a scientist’s natural state of mind; after all, the storyline of almost every scientific paper starts with a clearly defined question and then proceeds directly to the answer. In reality, the way scientists retell their discoveries may reflect much more how humans communicate knowledge than how those discoveries were actually made.

[…]

[If] a scientist proposes an important question and provides an answer to it that is later deemed wrong, the scientist will still be credited with posing the question. This is because the framing of a fundamentally new question lies, by definition, beyond what we can expect within our frame of knowledge: while answering a question relies upon logic, coming up with a new question often rests on an illogical leap into the unknown—the hallmark of night science.

[…]

Why, then, does it not seem this way? Why do questions appear secondary to answers? It may be because a new question is so powerful that it transforms our reality. A new question tends to erase its own origin; it is hard to imagine that there ever was a time when the question was not there.

We are living in a world where answers seem ever easier to come by. But answers are only half the question.