June 2026 - What AI is amputating, negative capability, and the lost art of the bag-dragger
I’ve named wonder as my word for the rest of 2026. It arrived in early May and it’s been sitting with me since, partly because wonder is the one posture I’ve been least in lately, and partly because I’m starting to think it is an actual discipline. May and the start of June have been about internal work more than operational output: the role at the aerospace and defense company is clarifying, and I wrote and released a Sabbatical guide that found its first readers, and started traveling via word-of-mouth on its own. There was also a phrase a friend used during a cigar session, one that pulled a seven-year practice of mine into focus. More on all of it below.
What I’ve Been Reading
Brent Beshore - “The Question We Forget to Ask” (Substack, free) · Part three of my dear friend Brent’s AI series, and the best thing I’ve read this year on the formation question. His argument is precise: every technology extends some human capacity while amputating another, and AI, uniquely, extends cognition. “What does extending cognition amputate?” His honest answer: formation. The friction of not knowing, the patience built by working through problems, the contemplative attention Simone Weil called the substance of prayer. AI makes all of it optional. “Every time I outsource my attention, I’m outsourcing something sacred.” It’s not some naive screed against AI; it’s an honest reckoning from someone who is thinking deeply from inside the problem. I re-read the section on Sabbath a couple of times. If you’ve been following this email for a while, you’ll know this connects directly to what Sasse said last month, and Brent wrote the piece on Sasse, too. He’s been working this vein steadily.
Andrej Karpathy - Sequoia Ascent 2026 talk (blog post / YouTube, ~30 min) · Karpathy’s clearest articulation yet of Software 3.0: the context window as the new program, LLMs as a new interpretive layer between human intent and digital output. His most important observation isn’t about capability. It’s about what changes when the unit of programming shifts from writing code to delegating macro-actions. “Some apps should stop existing as apps.” The Beshore piece and the Karpathy piece are asking the same question from opposite ends: Karpathy maps what becomes possible; Beshore asks what it costs. I’m knee deep in AI right now, and I’m pondering it as much as I’m practicing it.
“Letter to the Western Man at the End of Their Patience” (anonymous, Substack) · The most quietly unsettling thing I read this month. The thesis: the modern influence operation does not try to convert you to a foreign ideology, because an argument is something you might win. It just rearranges the room you already live in until you can’t stand next to your own neighbor. “You only need to make the people inside that society stop hearing each other.” It reframes Soviet active measures for the algorithmic age: the platforms that were built to connect you soon discovered that “rage holds your attention longer than reason and that humiliation travels further than thought.” Reading slowly, with long books and articles, rather than snackable content, might help. But a real fix is proximity to actual people who, let’s face it, aren’t 100% aligned with your ideology. Another fix: question and analyze the emotion that arrives faster than the thought. Fair warning: it’s deliberately one-sided. But the core mechanism is the truest description I’ve seen of why the social media feed makes us contemptuous. And why you haven’t seen me on any social platform in a number of months.
Annie Jacobsen - Surprise, Kill, Vanish · Seventy years of America's paramilitary history, from the post-WWII birth of CIA covert action through the Phoenix Program to the drone-and-signature-strike present. The access is remarkable, because Jacobsen gets operators and classified sources who don't usually talk. The throughline that stuck with me is the escalation paradox, that capabilities built as alternatives to full-scale war quietly become the default tool, precisely because they're deniable and they work in the short run. The Title 50 versus Title 10 tilt (CIA covert action versus military operation) sounds like legal trivia until you realize the entire accountability gap lives inside it, or unless you’ve actually lived inside of it. Pairs directly with Billy Waugh’s bio from last month: he was the operator on the ground, this was/is the system above him. The uncomfortable question it leaves open is whether making assassination efficient makes the long wars more likely rather than less.
What I’m Thinking About
Josef Pieper made a distinction the medieval scholastics took for granted and modernity almost completely forgot: ratio and intellectus. Ratio is discursive reasoning: analysis, logic, the mode that builds toward conclusions through effort. Intellectus is contemplative understanding, receptive, not effortful, arriving as a gift rather than a product of labor. Modern knowledge work is almost entirely ratio. AI accelerates ratio further, faster, with less friction; it is the perfection of the ratio mode. What it cannot extend is the other mode, the one that requires waiting, receiving, staying in the question without resolving it.
Keats called it Negative Capability: the capacity to remain in uncertainty without an irritable reaching after fact and reason. I’ve been sitting with the gap between what I can produce and what I can actually receive. The OODA loop work from May still applies: orientation is the complex node, and genuine reorientation requires staying in observation longer than is comfortable. Wonder, I think, is the word for the posture that makes intellectus possible. Not curiosity as fuel for the next thing, but wonder as the willingness to be in a question long enough for something to arrive that I couldn’t have generated. I don’t have a tidy conclusion here either. I’m still in it. “Be quiet and wait” is what I’ve been hearing in my heart.
The Sub-Plot
I was smoking a cigar with friends when one of them (Phil, a former Navy submariner - get it? SUB-plot?) used a phrase I had not heard in years: bag-dragger. In Washington, the bag-dragger is the young aide a step behind a congressperson or a cabinet official, holding the briefing book, saying nothing. It looks like the lowest job in the building. It is one of the best a 24-year-old will ever have, because that aide is in the room. Every room. The classified briefing, the tense negotiation, the moment the principal drops the talking points and says what they actually think. The aide is not there to talk. The aide is there to absorb.
I’ve been running a private-sector version of this for about six years, and I have a bag-dragger right now. He joined the week I was closing a hard transaction, with no context and nothing to contribute. The instinct in most companies is to park a person like that on busywork. I do the opposite: you are coming into every meeting with me, and if people look at you, just look at me. He sat through a session with me, our founder, and our COO working a real problem, and at one point the founder turned and coached him directly. Not about the deal. About how to operate. “You could be right all day long, but if you don’t deliver it well, it gets lost, and then people just don’t listen to you anymore.” You cannot teach that in an onboarding deck. You catch it by being in the room when someone who earned the right to say it, says it.
Here is why I put this next to the Beshore and Pieper thread above. We develop young people almost entirely around task output: give them a deliverable, grade it, repeat. But what makes a person valuable early is not the quality of their task work. It is proximity to judgment, and that is tacit knowledge, the kind that does not transfer through assignments. It transfers through nearness. Which is exactly the formation Beshore says AI is quietly amputating, and exactly what Pieper’s intellectus requires. The counter-move is almost embarrassingly low-tech and vulnerable: stand the young person close to where judgment actually happens, then debrief. And the debrief is the part everyone skips, because it is unglamorous. The magic is not walking them into the room. It is the twenty minutes afterward where you make them metabolize what they saw, every time, not just when you feel inspired. I DEFINITELY do not get that part right most of the time.
Two questions this month, since I’ve been circling around “wonder” and “proximity”. What practices are you protecting, the ones that require staying in a question rather than generating an answer? And on the other thread: who let you in the room early, and who are you letting in now? If you read the Beshore piece, I’d also like to know what it cracked open for you.
— eddy —