Not the Interface Between the Machine & the Model But the One Between the Model & the Human

Source: J. Bradford DeLong, Grasping Reality · 2026-10-06

Can me + my machines read the books that I will never have time to? I read half as manyserious books as I used to. Can my systems fake it? & are the systems in which I am a component really faking it?

In my prime, I would read perhaps 150 serious books a year.

I would skim over parts that did not resonate, but would still absorb a good deal as I skimmed. I would focus on the themes and passages that resonated. I would think about them. I would then go off and argue with the author, or at least with a spun-up sub-turing instantiation of the author in my imagination. Figure 5 hours per book. 750 hours per year building my own cognitive capacity out of a total budget of 24 × 365 = 8760 hours per year, it was time worth spent.

I am down to perhaps, half of that: 75 books a year, 1.5 a week, of which 40% are re-readings to remind me and refresh my parasocial acquaintance with my intellectual magisters and adversaries. Half of this gap comes from lower energy. Half of the gap comes from time being absorbed by reading and watching other things. Of those, perhaps half of which are things that are worthwhile.

There is a definite loss. There is a great sense in which deep reading and substantial continued engagement builds one own cognitive knowledge structures in an important way. Thus this feels like a gap in my personal information processing life that I should try to repair. The natural way to repair it, of course, is to see whether MAMLM technologies might be useful. Could they create a document, a document that I could then spend an hour with, that will present to me summary points and then focus my attention on passage in the book that I would have focused my attention on had I read it? And might I then be able, by having that document thereafter, quickly refresh my recollection? Thus I will not have the memories I would have had had I spent five hours with the book. But I would have the ability to rapidly simulate and, when asked questions about or thinking about issues relevant to the book, have me and my IT systems rapidly come up with the words I would have said or written if I did have the memories?

In short, could I turn myself and my machines into a Chinese Room? (With the benefits not only of communicating to others, but then actually building memories as my self watches the Chinese Room output.)

Were this to work, my 75 serious books a year could then effectively grow to 375. I, or rather the Chinese Room of which I was a component, would then know more useful stuff, or at least communicate to others (and back to myself) as if it knew more useful stuff.

The answer to the question is, almost surely: No.

But is it worth trying, because (a) it might work—the horse might learn to sing—(b) it might half-work and still thus be useful as a note-system, and (c) it will teach me a lot about the limits of MAMLM technology when one asks it to do more than stochastic parrotage outside the realm of the extraordinary use case that it MAMLM-as-coding-assistant, or rather MAMLM-as-coding-master. Hence I now have:

A conspectus is a short, organized overview of a larger text, giving you the main points and how they fit together, from conspicere, “to look at attentively.” You rarely see the word even in rarified academic English, except in translations from Slavic where konspect is an everyday word for structured study notes.

A summary condenses. A synopsis is a summary that follows the sequence of topics. An overview is casual and broad. A survey covers the range, rather than compresses. The word conspectus combines summary and survey: a compact map of the territory that is comprehensive and structured yet as brief as possible. the content. For me, my conspectus is book-by-book folders for books that I wish I had read (or re-read): notes, briefing, and synthesis documents that I review, flipping back to the text to read the book passages it direct me to as I do, and then afterwards either refer to myself or have it brought to my attention by my systems when it seems relevant to what I am thinking about.

Thus I will not have my memories of having read the book. I will not have read the book. But will the Chinese Room I and my systems construct be a good enough substitute for my memories of having read the book? Will I or my intelligence, for some definition of “I” or of “my intelligence”, have the knowledge that I would have had I read the book of Felix Somary, Menger’s student and classmate of Schumpeter and Mises, who spent fifty years watching Europe come apart and saying so? My machines built me a conspectus: a briefing, a critique, and pointers to the key scenes, including Weber storming out of a Vienna café and Ludendorff interrogating Somary at Kreuznach.

My bottom line: I think so. Give me five minutes’ notice, and I can now construct the memories, the brain states, that I would have had I in fact read the book. Or at least I can construct a very close approximation of them. I have run enough spot-checks against the text to be confident that the hallucination level her is very low. So I can now talk about the book as convincingly as if I had read it. Is that knowledge, or only a good performance of it?

Now let me switch gears and talk about Felix Somary 1889-1956 called by some “the Raven of Zürich” and his autobiography The Raven of Zürich. In 1899 (aged seventeen), he enrolled in the Faculty of Law and Political Science of the University of Vienna, where he studied under Carl Menger (to whom he also served as an assistant) and Eugen von Philippovich. I am told by Wikipedia that his classmates included Joseph Schumpeter, Rudolf Hilferding, Otto Bauer, Emil Lederer, and Ludwig von Mises.

Somary foresaw the 1929 crash, preferred state bankruptcy to inflation, and warned against unconditional surrender. I’ve wanted to read his memoir for a decade and never will.

Here we have yet another of the Vienna figures, like Hayek, Schumpeter, Polanyi, and so forth, who kept trying to grasp and understand the problems of 1900 to 1950—the problems rooted in the replacement of the steampower by the applied-science, and then the mass-production economy, along with the parallel and associated transformations of mass communication and of the social network with the rise of nationalism and ethnicism. I know the book because of its account of Max Weber and Joseph Schumpeter discussing the Russian Revolution:

Great story, even if it is too good to be true in its details. So that put Felix Somary’s The Raven of Zürich on my to-read list. But today it is further from the top of the list than it was a decade ago. It is clear that I am never going to read it. But a properly-harnessed ollama/gemma4:31b-mlx, overseen and guided by a harness constructed by anthropic/claude-opus (itself harnessed by claude-code), can:

Plus I fed it Max Weber’s Economy & Society (which I have read, but long ago), and Slouching Towards Utopia (which, of course, I know very well.

To summarize: I am quite pleased. After spending an hour on the conspectus folder (and then half an hour, checking random pieces of the conspectus folder against the actual book itself), I definitely feel like the Chinese Room of which I am a component could definitely fake it. It can respond as if I have actually read the book, in the way I would. Thus I have hopes that my leafing to the passages it points me to, consulting it when it comes to mind, and having my systems bring it in when they think it is relevant will together be as if I actually have the memories of having read, at least for people looking at me in my Chinese Room from the outside. And from the inside? That is a hard question that I do not yet have an answer to.

But the proof of the pudding is in the eating: What do my systems think is worth noting from Somary? Here is the introductory briefing that it wants to give me:

Felix Somary was the ultimate high-society canary in the coal mine. Trained in the Vienna school—which meant he spent his youth learning that the world is a fragile place and that economic depressions are inevitable (p. 17)—he spent the next fifty years watching the 20th century dismantle itself in real time. He calls the government the “fence that separates the zoo of wild animals from the outside world” (p. 15), and he spent his career observing exactly when and how those fences broke.

The arc of his life is a map of European collapse. In the pre-1914 years, he was playing the great game of international finance, trying to use the Baghdad Railway as a colonial compromise to stop the Anglo-German naval race from turning into a world war (p. 35-38). He failed, of course. During the war, he became a critic of the military’s delusional confidence, specifically General Ludendorff, who thought he could just seize Caucasian oilfields without considering how to actually get the oil out (pp. 66-67).