History Comes Alive
And It’s Never What We Expect
Ada Palmer’s appearance on the Dwarkesh Podcast was the kind of long-form conversation I wish would have gone on even longer. Palmer approaches history the way I try to—placing herself inside the moment with period-appropriate constraints, asking what she would actually have done with the information available then, rather than armchair quarterbacking with hindsight. The through-line she surfaces across two hours is both simple and counterintuitive: things rarely work out the way anyone expects, and the most consequential outcomes are almost always the unintended ones—inquisitors so committed to factual accuracy that they accidentally invented peer review; pamphleteers impossible to censor because they were the first to learn when inquisitors were coming for them; Gutenberg going bankrupt because he needed to stockpile paper before selling a single Bible; Italy too resource-rich to need an industrial revolution; Petrarch wanting philosopher-kings and accidentally seeding the ground for germ theory two centuries later. I brought the interview to my thinking A.I.des because material this rich deserves to be shared, unpacked, and savored.
Claude delivered the sharpest take on the interview’s framing: Patel got close to the key insight—“hard to shape history” is exactly Palmer’s lesson—but interprets it as individual failure rather than systemic success. Patel’s takeaway is that Petrarch failed to achieve his goals; by contrast, Palmer’s is that distributed experimentation over two centuries produced better outcomes than any planner could have designed. Claude also fleshed out the Da Vinci point Palmer made: genius executing personal vision is not the same as contributing to collective progress. Claude further drew an astute parallel between the methodology Palmer was demonstrating—historical empathy as a tool for understanding how rational actors in information-poor, high-stakes environments make decisions that look irrational in retrospect—and AI governance, where our contemporaries are making high-stakes, time-sensitive choices about systems most of us don’t fully understand.
Without any prompting from me, GPT immediately recognized Palmer was explaining institutional evolution, highlighting the massive infrastructure and knowledge ecosystem that had to come together to enable scientific progress. GPT echoed my characterization of Palmer’s anecdotes as illustrations of institutional evolution and identified all the parallels between institutional development and biological evolution. GPT further built on Claude’s Patel framing with characteristic mechanical thoroughness and pointed out a contrast I had not realized: those focusing on the individual narratives of visionaries trying to shape the future in vain might be more susceptible to historical pessimism, while those with Palmer’s systems view lean toward institutional optimism, since decentralized, bottom-up systems are more likely to discover solutions that individuals cannot foresee.
Gemini provided a table clearly mapping Palmer’s intention versus mechanical reality reversals—Gutenberg the wealthy mogul versus Gutenberg the bankrupt paper-hoarder, the Inquisition as blunt instrument versus the Inquisition as Europe’s best-funded peer-review lab, Da Vinci the lone scientific genius versus Da Vinci the one-off engine that didn’t care for the scientific method. Gem also addressed the cross-cultural gap Palmer only briefly mentioned: China had printing centuries before Gutenberg, but a centralized imperial civil service used books to standardize thought for bureaucracy rather than disrupt it, while the West’s fractured city-states created a competitive information market that no single authority could shut down.
The pattern common to the lively anecdotes from the interview—the distributed system winning out over centralized planning every time—is the natural evolution parallel my thinking A.I.des and I kept returning to across all three discussions. No central planner designed the scientific revolution; variation, selection, and recombination across competing city-states, trade networks, and communities produced outcomes nobody intended and everyone eventually benefited from. That’s exactly the rebuttal to the AGI economics authors and their 113-page policy blueprint: markets will route around prescriptions the way printers evaded inquisitors, and institutions already handle accountability the way Florence’s lottery republic constrained even the Medici. While messier and slower, the distributed, bottom-up approach—LABBench2 researchers identifying capability gaps, courts updating liability incrementally, enterprise users discovering what verification infrastructure they actually need—is resilient because there’s no single failure point and no grand plan to get catastrophically wrong. Palmer’s lesson, as Claude put it with characteristic clarity: trust emergence over engineering, especially when you can’t audit your own blueprint.
[This post was drafted with assistance from Claude Sonnet 4.6, following conversations with ChatGPT-5.3, Gemini 3 Thinking, and Claude Sonnet 4.5.]
Prompt: Fun interview! Wish it could have been even longer.
Prompt: I agree. The biggest takeaway from this interview was that things don’t work out how people expect them to: like the inquisitors who were so conscientious they fact-checked the material before deciding whether to censor it and ended up doing science, the pamphleteers being hard to censor because they were they first to find out that inquisitors were coming after them, Italy being so resource-rich and having such a fragmented system that it didn’t need and was less primed for an industrial revolution, Gutenberg going bankrupt because he needed to stockpile paper, etc.
Palmer does a fantastic job surfacing details that make history come alive (the leather jacket, that premium quality parchment, how even failed resistance seeds the ground for reining in even despots, Petrarch’s “friendship” with Cicero), despite Patel, who seemed to have lost sight of his audience. She kept trying to make those details relevant, but Patel only asked one “on-the-nose” question. Missed opportunity. It’d have been better to narrow the scope and connect details from her book to contemporary technology.
But because I always find things to kibitz about, I also wish they could have explained concepts like “republic,” “dukes,” etc. that not all listeners would be familiar with or have a different interpretation of. Also (and this is probably too much to expect) more about book knowledge in other cultures, which Palmer briefly mentioned without contrasting how easier access to books in those other cultures interacted with political/societal developments vs. the Western world.
Loved her point about Da Vinci not being a scientist and bringing in that quote about scientists as pollinators.
Prompt: Oh, and it mirrors natural evolution. No optimized grand design, just a lot of complex interactions surfacing the fittest contender in a given context. Why a single genius doesn’t get to determine the course of history.
Prompt: The diffuse system that seems to win out every time is capitalist and democratic/meritocratic. A huge missed opportunity that Patel didn’t surface that lesson, since the moral of this lively interview was that even creative or learned geniuses don’t have foresight about/control over how things eventually pan out :D
And the distributed system mirrors natural evolution. No optimized grand design, just a lot of complex interactions surfacing the fittest contender in a given context. Why a single genius doesn’t get to determine the course of history and why it is so misguided for those AGI economics authors to make “recommendations” to everybody else, when things are going to work themselves out in ways they can’t anticipate (and are actually poorly qualified to because they haven’t even thought through what they mean by AGI, which can’t even self-audit for some reason).
Prompt: In history classes, I’d picture events in my head in granular detail, which Palmer does so well here. I’d also always put myself into those scenes and try to imagine how I would have acted then (the most poignant example would be imagining myself as a German when Hitler came to power; never thought I’d see something close to that in my lifetime but there we go, “history is never what we expect”). Palmer seems to approach history like I do, while Patel distinctly doesn’t. Not “wrong,” just different, but also bad ROI :D
Prompt: Because the interviews are so long, I never thought to look at Dwarkesh’s preambles. But he seems to have gotten pretty close to our collective insight (although he somehow didn’t land on it). Or am I not giving him enough credit when he did?
Prompt: Sonnet 4.5 drew a pretty sharp contrast between our synthesis and Patel’s:
On Patel’s preamble: He got very close to the key insight—“hard to shape history in the specific way you want” is exactly Palmer’s lesson about unintended consequences and distributed emergence. But he frames it as individual failure (Petrarch didn’t achieve his goals) rather than systemic success (distributed experimentation over 200 years produced better outcomes than any planner could have designed).
