Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.
Back to Podcast Digest
dark ilan - By Ozy Brennan
TL;DR
Vellum realizes the official AGI cover story doesn’t explain the scale of historical censorship — he argues that if the real danger were just fast AI progress, there’d be no reason to “bubble” huge stretches of history, suppress human evolution, or make society itself feel oddly overbuilt around edge cases.
The keeper confirms a second, much darker conspiracy through careful non-denials — in a world where keepers swear never to lie, Vellum catches the signal in what she refuses to deny, especially around Thal’s amnesia, dreams, and the fact that some truths are genuine infohazards.
The big reveal is not AGI but a universe ‘crawling with unaligned superintelligences’ — the orthogonality thesis wasn’t inferred philosophically but “discovered through empiricism,” meaning humans have already encountered alien superintelligences with no particular reason to spare us except not needing our atoms yet.
A lot of society’s weirdest features are explained as defenses, accommodations, or bargains with nonhuman minds — the Pacific shipping triangle avoids a sleeping entity, the Arctic, Moon, and Mars are effectively ceded territory, Thal’s amnesia covers for alien academics, and cats are quietly sapient and politically important.
The story’s emotional pivot is Vellum discovering there is no higher grown-up coming to save him — after chasing mysteries from cheese inspection to the Basement of the World, he learns his reward for relentless curiosity is recruitment: “We teach you rationality not because the universe is comprehensible… Vellum, you have selected yourself to study magic.”
The Breakdown
Vellum brings a second conspiracy to a keeper
Vellum only goes to a keeper because he’s stuck, not because he’s making progress. He knows his edge isn’t brilliance — he’s “exactly as smart as the average human” — but sheer hours, obsession, and a refusal to stop tugging once he finds a loose thread. The scene opens with that perfect Ozy Brennan mix of comedy and menace: an ergonomically correct, exquisitely “dampunk” office chair and a protagonist who used to inspect cheese for mold.
The AGI explanation starts to crack
Vellum’s core complaint is simple: the official rationale for bubbling history doesn’t add up. If the main infohazard were just how fast computers improve, why suppress so much history, including things like human evolution, when self-driving cars, computer vision, and prediction-market trading already scream “general thinking machine is possible”? The keeper keeps fencing with technically accurate answers, and Vellum notices the important thing — she never actually says he’s wrong.
Society is weirdly built for disasters that never happen
Once Vellum starts looking with fresh eyes, all of civilization seems suspiciously overprepared. He points to annual government-overthrow rehearsals, alien invasion festivals, and Marin’s expensive exception-handling simulations, 83% of which have never matched any real-world crisis he could find. His irritation gives the section its energy: this isn’t abstract theorizing, it’s the feeling of noticing that everyone is practicing for a play no one admits exists.
Thal’s amnesia, dream taboos, and the first real warning
Vellum lands on Thal’s amnesia as especially bizarre: memory loss, a sudden fixation on humanities and social science, universal archive access, and spontaneous remission after 4 to 7 years. That’s when the keeper finally breaks the pattern and says the true answer is an information hazard. She adds the sharpest line in the piece: 84% of people regret learning a dangerous infohazard if they were motivated only by not being able to stand ignorance — and Vellum, of course, can’t stop.
The dream clue and the surprise promotion
The keeper says Vellum missed one giveaway: the taboo on talking about dreams. If people compared notes, they’d realize about 40% of people have serialized dreams in a whimsical world with great architecture where everyone rides zebras, while the other 60% mostly have stress dreams about preschool and Marin. Then comes the whiplash punchline: because he found the conspiracy, he’s now a keeper.
The real small infohazard was AGI all along
Vellum had insisted there were no “small” infohazards safe enough to use as a secret test of character. The keeper corrects him: AGI is exactly that. The sting is great because it flips his whole first-conspiracy triumph — what he thought was the big forbidden truth was basically the tutorial level.
The universe is full of alien superintelligences, and humans are not the main characters
The keeper reveals that the orthogonality thesis was learned empirically: the universe is “crawling with unaligned super intelligences.” They don’t hate humanity; they just haven’t needed our atoms yet. History was bubbled to hide evidence, some “dark rites” can attract attention, and one such intelligence once spent 200 years running a country — leading Vellum straight into the horror that maybe everything important was manipulated long before he was born.
Cats, the Arctic, and the final turn into magic
From there the world’s absurdities click into place in rapid succession: the Pacific triangle avoids a sleeping entity, the Arctic and off-world territories are effectively yielded to aligned-enough aliens, Thal’s amnesia masks alien academics doing participant observation, and cats are sapient but prefer secrecy because otherwise “the monkeys would be bothering us all the time.” Vellum is most wrecked by the cats, which is exactly right. The ending lands on the biggest shift of all: rationality was never taught because the universe is neat and comprehensible, but because it isn’t — and then the keeper snaps her fingers, summons flame, and tells him he has selected himself to study magic.
Share
Keep Reading
Two Ways to Follow
The Weekly Echo. The inbox-shaped summary of what mattered.
New editorials announced here.
