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.

Becoming Kinder

02.17.26 // Miscellany

Following up from our brief discussion of yoyu, an excerpt from George Saunders’ 2013 convocation speech at Syracuse, presented without unnecessary commentary:

When young, we’re anxious — understandably — to find out if we’ve got what it takes. Can we succeed? Can we build a viable life for ourselves? But you — in particular you, of this generation — may have noticed a certain cyclical quality to ambition. You do well in high-school, in hopes of getting into a good college, so you can do well in the good college, in the hopes of getting a good job, so you can do well in the good job so you can . . .

And this is actually O.K. If we’re going to become kinder, that process has to include taking ourselves seriously — as doers, as accomplishers, as dreamers. We have to do that, to be our best selves.

Still, accomplishment is unreliable. “Succeeding,” whatever that might mean to you, is hard, and the need to do so constantly renews itself (success is like a mountain that keeps growing ahead of you as you hike it), and there’s the very real danger that “succeeding” will take up your whole life, while the big questions go untended.

So, quick, end-of-speech advice: Since, according to me, your life is going to be a gradual process of becoming kinder and more loving: Hurry up. Speed it along. Start right now. There’s a confusion in each of us, a sickness, really: selfishness. But there’s also a cure. So be a good and proactive and even somewhat desperate patient on your own behalf — seek out the most efficacious anti-selfishness medicines, energetically, for the rest of your life.

Radiology Attention Residue

02.16.26 // Radiology

The radiology job of “reading scans” is actually two different jobs in rapid alternation: pattern recognition (pixels, asymmetry, gestalt) and translation (turning that visual certainty—or uncertainty—into clean prose with a defensible impression).

The vast majority of my work is diagnostic these days, but there have been times in my life when I wore lead, did a lot of procedures, and planned for a career as such. Interventions were much more physically demanding and certainly had moments of emotional intensity that I haven’t felt in years at this point in my career. But I always found diagnostic radiology more mentally exhausting, especially earlier in my career, when a greater fraction of cases required a greater amount of Type 1 thinking.

A lot of that end-of-day exhaustion isn’t from volume alone; it’s from the accumulated switching cost (attention residue) of bouncing between visual cognition and written language, case after case. If you want real efficiency (and fewer mistakes), you don’t just read faster—you reduce forced context switches so your brain can stay in one mode longer.

We all need to work on our own personal version of the iterative loop and efficiency/ergonomics.

This is one of the missing joys of working without a better process—whether that’s macros, pick lists, templates, or even better physical integration with your computer. A streamlined, efficient process minimizes interruptions, and that seamlessness allows you to move through the cases instead of fighting to get each one done.

From writer George Saunders’ recent interview in the NYT Magazine:

[Chekhov] says a work of art doesn’t have to solve a problem — it just has to formulate it correctly… My job, rather than answering your question, is to allow [the characters] to make the best possible case for their view…I wrote myself into a place where the question got more and more profound, and I found myself less and less capable of giving a definitive answer. That’s not for an artist to do. You ratchet the question up, and you go, “Yeah, that’s a tough one.”

Yoyu

02.12.26 // Miscellany

There is perhaps no career/life advice given more often in recent years than to say no to things. To be sure, the Power of No is important. We are finite creatures with limited time. We simply cannot do it all, and no one will respect our time for us. (And to be fair, people also often advise new careerlings to say yes to everything in order to build a reputation and discover what you like).

But I also believe there is a pendulum component to people‘s attitude about work that is a perhaps overly profound sense of scarcity. That we win by doing the bare minimum. That we cannot give of ourselves without being taken advantage of. While I agree that the institution doesn’t love you back, I nonetheless think this is a depressing way to approach a massive fraction of your life.

There is a Japanese concept that Craig Mod refers to in his book Things Become Other Things called yoyu, which he describes as the extra space in your heart for other people. I think there is a missing sense of abundance in many people, a deep conviction that there is simply not enough of “you” to give of yourself. That our love and attention and focus and work are too limited a precious resource to be shared.

Time is a zero-sum game, and we all only have 24 hours a day to deploy, but at the same time, those hours are not equally effective or meaningful if we are willing to share ourselves generously. Adam Grant describes in Give and Take that “givers” are overrepresented at both the top and bottom of the corporate ladder. Giving in self-destructive ways to selfish, bad people is an easy path to resentment and burnout, but giving generously is also a way to increase the number of possible paths to fulfillment.

Appreciation is a fantastic currency.

Trust does not appear in GDP

02.10.26 // Miscellany

From Adam Butler’s “The Strip-Mining of Trust“:

The economists call trust “social capital,” but the name obscures the crucial asymmetry. Trust is not like money. It is more like topsoil—built up over generations through countless small depositions, easily washed away in a single season of careless extraction, and once gone, recoverable only on timescales that mock human planning. The farmer who strip-mines his soil for a decade of bumper harvests will leave his children a desert. The society that strip-mines its trust for a generation of efficiency gains will leave its children something worse: institutions that still stand in their accustomed places but have been hollowed out from within, and a friction industry selling services to navigate the rubble.

There is a story about a nurse in part three that is so compelling that I will not deprive you or Mr. Butler of some of the essay’s potency by trying to blockquote a reasonable portion of it. Just go read it.

This is not a failure of implementation but a structural limitation: to manage at scale what it cannot see, the system must convert discretionary judgment into measurable proxies. And here Goodhart’s Law begins its corrosive work. Once the proxy becomes the target, it ceases to measure what it was designed to capture. What gets measured gets managed; what gets managed gets gamed; what gets gamed gets hollowed out.