Imagine a library that holds every possible arrangement of letters ever — every true sentence, every false one, every novel that could be written and every one that never will, all of it, shelved.1 You walk in looking for something good. You find nothing, because everything is here. The masterpiece is on a shelf somewhere, but so is its garbled twin, and the ten million near-misses around it, and you have no way to tell which is which. Total abundance produced total worthlessness.
That is Koe's setup for the skill he calls taste. The counterintuitive core: in a world that can generate anything, the valuable act is not making more — it is choosing what's worth keeping.2 Curation beats creation.3
Taste is discernment — the ability to select what matters and hand it to someone else.4 Strip away the word "taste," which sounds like a vibe or a luxury, and the mechanism is plain: it is filtering. Out of an enormous field of options, you point at the few that are worth another person's attention, and you are right often enough that people start trusting your pointing.
Koe builds the case with two thought experiments. The infinite library is one. The other is the infinite-monkey problem: put infinite monkeys at typewriters for a billion years and eventually one types Shakespeare — but the result is worthless, because the monkey has no idea what it wrote.5 Both stories make the same point from opposite ends. When anything can be produced, the production stops being the valuable part. What's left is selection.
Think of creation as a faucet and taste as the drain you choose to leave open. When the faucet was expensive — when writing a book or making a film took years and money — owning the faucet was the whole game. Now the faucet is free and runs at infinite pressure. Everyone has one. The scarce thing is no longer the water; it is the person who can look at a flood and say this cup, not that one.
Koe's monkey makes the logic crisp. The monkey produced Shakespeare, but it had no idea what it was writing, so the output means nothing.6 The meaning was never in the text alone — it was in a mind that understood why those words and not the billion others. That understanding is taste. It is the difference between a thing existing and a thing being chosen by someone who knew why.
This is why Koe insists curation matters more than creation in the AI era specifically.7 When generation costs nothing, the bottleneck moves to the only step a generator can't perform: caring which output is good.
Taste hands the vault its answer to the abundance problem the AI pages keep circling. AI Commoditization Raises the Human Premium argues the human is what stays valuable when content floods; this page names which human capacity does the work — the filter, not the faucet. It gives Taste/Judgment/Labor its first term in spoken-Koe form. And it sits one rung above Agency as the Meta-Skill in Koe's own stack — agency unlocks the skills, taste is the first one you reach for.8 The link to Iterators Beat Perfectionists is the part worth dwelling on, and the Handshakes section does.
Koe's sharpest line is a question he poses about AI itself: it can spit out five billion words of beautiful prose — which ones do you select to stick around?9 Slow down on that, because it's the whole concept in one move.
The five billion words already exist, hypothetically perfect. Creation is solved. What's not solved is the pointing. Somebody has to read the flood and say these, these are the ones, and that somebody is doing the only job left. The case study isn't a person who wrote something; it's a person standing in front of an infinite output, making cuts. The value migrated from the writing to the cutting. And notice — the cutter has to know why, or they're just the monkey again, pointing at random. Taste is selection plus the reason behind it.
It's Thursday, late, and you've got eleven drafts open — eleven versions of the same piece, some yours, some the model's, all decent. The temptation is to make a twelfth. Resist it. Making more is the easy move and it's the wrong one.
Instead you read all eleven slowly and you feel for the one that lands. You can't fully explain it yet, but one of them has something the others don't. You close the other ten. That small, almost-painful act — deleting good work because it isn't the work — is taste in motion. Then you ship the one, you watch what happens, and you notice whether the thing you felt was right actually was.10 Tomorrow your feel is slightly calibrated. That loop — build, ship, see what works, iterate — is how Koe says you grow taste in the first place: not by studying it, but by making real cuts and getting feedback on them.11
A creator without taste has tells, and the AI era amplifies all of them:
This is one creator's framework in a YouTube transcript — [PARAPHRASED], [POPULAR SOURCE]. The infinite-monkey theorem and Borges-flavored infinite-library are real ideas Koe deploys as illustrations, not as cited scholarship; he gestures at them as things "some of you may have heard."14
A tension worth naming: Koe says you develop taste by building and shipping and seeing what works.15 But "what works" usually means what got attention — and attention rewards the familiar, the trend, the already-validated. So the feedback loop he prescribes may train you toward what's popular rather than what's good, and those diverge. He doesn't address how a taste calibrated on engagement metrics avoids collapsing back into trend-chasing — the very thing his agency page warns against. The open question: is there a way to develop taste that isn't downstream of the crowd's verdict?
Koe and George Butcher converge tightly here. Butcher's Iterators Beat Perfectionists says you don't think your way to good work, you iterate your way there by shipping and adjusting. Koe's prescription for developing taste is identical in shape: build, ship, see what works, iterate.16 Where they pull apart is the object. Butcher's iteration improves the work. Koe's iteration improves the chooser — each ship calibrates your filter, not just your craft. Hold both and you see iteration does double duty: it makes better things and it makes a better picker, and the second compounds longer, because a picker outlives any single project.
There's a quieter clash with the Taste/Judgment/Labor framework. That framework treats taste as a stable asset you bring to AI work — the human input the machine lacks. Koe's account is more restless: taste isn't a possession, it's a muscle that atrophies if you stop making cuts and rebuilds when you start. The framework files taste under what you have; Koe files it under what you keep doing. Read together they suggest taste is an asset that decays — owned only as long as it's exercised.
Plain version: deciding what's worth keeping is the same move whether you're an editor, an artist, or a person standing in front of an AI flood — and it's worth more now precisely because making things got easy.
Handshake one — to creative-practice, Iterators Beat Perfectionists. Butcher's iterator wins by accepting that growth isn't a smooth climb — you ship rough work, most of it underwhelms, and occasionally a step jumps. Koe's taste is built by exactly this process, but it reframes what the iteration is for. The iterator isn't only improving output; they're training a selector. Each shipped piece returns a verdict from reality, and the creator who pays attention slowly learns to predict those verdicts before shipping — which is precisely what taste is, the ability to call the cut before the crowd does. Put the pages together and a non-obvious thing appears: the perfectionist fails twice. They fail to ship, sure — but more importantly they never train their taste, because taste only calibrates against real feedback, and the perfectionist starves it. So perfectionism isn't just slow; it's a taste-starvation disorder. The iterator's apparent recklessness is actually the only known way to grow the discernment that lets you stop being reckless. Neither the craft framing nor the business framing says this alone: Butcher explains why you should iterate, Koe explains that what you're really iterating is your own judgment, and together they expose perfectionism as the thing that guarantees bad taste forever.
Handshake two — to business, Beta-Editor Review Skill. The beta-editor page is about reviewing and refining AI output — sitting in the editor's chair over the machine's draft. That chair is the taste seat. Stack the pages and the editor skill gets its economic justification: in an era where AI writes the draft, the editor's discernment is no longer a finishing touch, it's the entire value-add. The beta-editor page describes the activity; the taste page explains why that activity became the scarce, paid one. The insight that falls out of the pair: editing and taste used to be separable — taste chose the project, editing polished it — but when generation is free, they collapse into a single function. Choosing which of the AI's five billion words to keep is editing, and editing well is taste. The two pages, read together, show a job role being born from the merger of two skills that abundance fused. Carry the merger forward and the economics get pointed. As long as editing and taste were separable, you could hire cheap editing and reserve your own scarce taste for the big calls — the labor was delegable, the judgment retained. Once generation is free, that split closes: every cut is now a taste decision, because there's no rote production left to delegate beneath it. The editor who merely polishes is automated; the editor who chooses is the only one paid. So the beta-editor page, read against the taste page, quietly predicts which half of the editorial profession survives the flood. The mechanical, grammar-fixing, tidy-it-up editor was always doing labor, and labor is the thing the machine took. The editor who could say this one, kill the rest, here's why was always doing taste, and taste is the thing the flood made precious. The pairing tells a working editor exactly where to stand: not at the polishing end, which evaporated, but at the choosing end, which just became the whole job — and it explains why "I'll clean up your draft" is now worth nothing while "I'll tell you which of your fifty drafts deserves to exist" is worth everything.
The Sharpest Implication. If curation outranks creation, the creative résumé inverts. "I made a lot of things" stops being impressive — the machine makes more, faster, cheaper. "I have reliable judgment about what's worth keeping" becomes the whole pitch. And judgment can't be downloaded; it can only be built by years of making real cuts and being right often enough. The sharpest version: the future's scarce professional isn't a maker, it's a chooser who can be trusted — and trust in someone's choices takes a track record no AI can fast-forward.
Generative Questions.