There's a founder, somewhere, who walked into a VC meeting this morning with a product that works. A real customer base. Revenue growing modestly. She's pitched this company twelve times. Eleven of those meetings ended with a polite version of "not quite the right fit for our thesis."
Between the eleventh meeting and the twelfth, she rewrote the first slide.
The product is the same product. The code running in production is the same code. Her customers haven't been told anything. But three words that weren't on the slide a month ago are on it now. Platform. Data. AI. "A scheduling tool for specialty clinics" has become, on the new slide, "an AI-native platform for vertical healthcare data."
The twelfth meeting lasts forty minutes instead of twenty. Eight days later she has a term sheet at a valuation thirty percent higher than the round she'd been trying to close for a year. The code hasn't changed. The sentence has.
The sentence is what got priced.
What happened in that meeting happens in every system that hands out any reward. The system pays you to describe yourself in its existing language. The payment comes out of your ability to describe what it doesn't yet have language for. Call it the legibility premium. Call the cost the perception tax.
The NYSE keeps receipts on exactly this trade. There are financial products on that exchange whose only purpose is to make a category tradable. Someone writes down a theme. Someone else gives it a ticker. A handful of stocks get bundled inside. You can buy it today. Robotics, cybersecurity, space, AI, the future of whatever. Call them thematic ETFs. Four economists published a study of them in the Review of Financial Studies in 2023. Specialized thematic ETFs underperform their diversified peers by about six percent a year in their first five years. Roughly thirty percent cumulatively, risk-adjusted.
The finding was sharper than the summary suggests. Sharper than the paper itself spells out, actually. The underperformance came down to launch timing. Fees didn't explain it. Hedging didn't either. Stocks inside a thematic ETF at launch are, on average, already overvalued, because issuers launch a wrapper around a theme precisely when investor extrapolation has peaked. The six percent a year is what people pay to own the sentence describing the thing instead of the thing itself.
Where else does that trade run?
The trade repeats
There's a CEO, somewhere, who got drilled by his IR team the day before his quarterly call. He'd seen the transcripts from Q3. Every other CEO in his peer group had found a way to say AI. His product hadn't changed. The guidance hadn't changed. But three slots in the prepared remarks had been rewritten to put the word in the first two minutes.
He wasn't lying. He wasn't exactly telling the truth either. His company had, in fact, added some machine learning to a few internal workflows, the way every software company has for a decade. What was new was the pressure to foreground it. If he didn't say the word, an analyst would ask. If the analyst asked and he fumbled, that exchange would be the pull quote. The word was already in the room. The only question was whether he put it there.
He said it on minute four.
At the end of calendar 2025, three hundred and thirty-one S&P 500 companies cited AI on their quarterly earnings calls. Sixty-eight percent of every call in the window. Highest in the ten years FactSet has been tracking it, more than double the five-year average. In information technology the share was 94%. In financials, 91. In communication services, 89.
Hundreds of CEOs are not conspiring with each other. Each is responding, in private, to the same local incentive. If the room already has a category, and silence toward the category reads as weakness, every competent operator will eventually say the word, even after it stops carrying information. Category saturation is what happens when the premium on legibility gets paid so uniformly that the signal goes to zero. The word becomes table stakes. The only question is whether you paid.
In April 2026, a wool sneaker company named Allbirds renamed itself NewBird AI, secured fifty million dollars of financing, and declared itself a GPU-as-a-service company. Allbirds had been in the process of selling itself to a brand management firm at a valuation of thirty-nine million dollars, around one percent of its 2021 peak. The announcement produced roughly a six-hundred-percent surge in the stock. Nothing operational had changed. The sneakers in inventory were still sneakers. You could, if you wanted, still order a pair.
Two things happened at the same time. The consumer retail business was about to be sold for parts. A newly-named AI infrastructure business was able to raise more capital, at a higher implied valuation, than the parts-sale would have produced. Two transactions, same underlying assets, radically different prices. Same oats. Same bag. Different letters on the bag. The market wasn't confused. It was pricing the sentence, not the thing.
There's a skincare creator, somewhere, with nine hundred thousand followers. She took up skincare two years ago after leaving a job in product marketing. Her posts are half shelf-talk and half personal: what she tried, what she regretted, what she's testing this week. Brands approach her every month now. They aren't paying for her chemistry knowledge, which is mostly pharmacy counter and Reddit. They aren't even paying for her audience reach, which they could buy more cheaply through programmatic. They're paying for her face to be a shape her audience has already agreed to trust, and then for that trust to be rented, by the impression, in the direction of their product.
In 2025, American advertisers spent thirty-seven billion dollars through creators, up twenty-six percent year over year, growing four times faster than the rest of the media industry. The number was $13.9B in 2021. It's $37B now. Roughly 207 million creators worldwide are pricing their own legibility in real time, each one a small exchange between a recognizable identity and the attention of a specific audience. A brand pays a creator the way someone pays for an introduction. The person's name opens a door the brand couldn't open alone. That's what the money buys.
Three rooms. A boardroom, a storefront, a phone. A CEO, a brand consultant, a creator. Different categories, different currencies, one trade. Someone is paying a premium to sit inside a category the system already knows how to see.
Annual Return vs. Benchmark
S&P 500 · Q4 2025
U.S. Advertising 2025
Stacked up, those numbers ask the same question the founder answered between meeting eleven and meeting twelve. If the wrapper costs something, and buyers keep paying it anyway, what are they actually getting?
They're getting borrowed certainty. Not what might be true. What has already been made legible enough to transact. The underlying is incidental. A basket of stocks, a software tool, an audience, a pair of sneakers. What's actually getting priced is the category.
Four layers
The trade runs in four rooms of the same house, and each room pays in its own currency.
Room one is capital. The currency is returns foregone. The founder's deck got priced at six percent of everyone else's returns, compounded over five years. The ETF paper is the aggregate form, executed ten thousand times a year across specialized funds. Capital markets pay cleanly and keep receipts. The loss is eventually knowable because the pricing is public.
Room two is audience. The currency is attention. There's a journalist, somewhere, writing a headline for a story she researched for two weeks. The story is about how a biosecurity program in a midsize city quietly fell apart in 2024. The honest headline is long and slightly boring and locally specific. The headline that will travel above the screenshot cuts the city, cuts the specificity, and uses a word she knows her editor wants in the subject line. She runs the boring version past the desk. The desk asks for a rewrite. The rewrite gets three times the clicks. Audiences pay for recognizable shape first, and the shape is what the system remembers. Truth is a different market, and it moves slower.
Room three is peer. The currency is status assignment. There's a junior analyst at a conference being asked what she does. She has three seconds. She has two versions of the answer, one long and one short. She gives them the short one. The person she's talking to nods, looks up, and flags someone over by name. Peers grant legibility cheaply at first, one introduction or invitation or citation at a time, and compound it. Ten years inside a career wrapper your peers already know how to name is expensive in ways you can't see from inside the wrapper. The cost becomes visible only when you try to step out, and the step out is priced in exactly the currency you've been earning.
Room four is the strange one. The currency is perception itself.
There's a writer, somewhere, who has been trying to finish an essay for a year. She has drafted it nine times. Every version pulls her toward a shape her readers will recognize, which is slightly different from the shape of the thing she was actually trying to say. The essays she published this year were good. This one, the one she keeps circling, is the one whose shape doesn't yet have a public folder. It's also the one that has not shipped. What's been lost isn't visible anywhere. There's no draft in a drawer. There's a perception that hasn't yet been sentenced, and may, the next time she returns to it, already be approximate.
The first three currencies regenerate. A market can correct. An audience can tire of a bit. A peer circle can revise what it rewards. Perception doesn't reissue itself on a schedule. What you stop noticing because it didn't fit a category the system could see is a loss that arrives silent. It's the way a question you were about to ask, right before someone changed the subject, becomes a question you never know you didn't ask. You don't get a statement saying this is what you didn't perceive this quarter. You just gradually stop seeing it.
Consider search funds, the cleanest case of layered pricing in American finance. The wrapper is peer-priced. Elite apprenticeship, operator status, the MBA-to-CEO arc. The underlying is a plain cash-flowing small business that usually sells car washes, dental practices, or HVAC routes. Stanford's 2024 study tracked 681 funds, 35.1% aggregate IRR, and a record 94 launches in 2023. The median acquisition was purchased at about 7x EBITDA, roughly the middle of the market for small operating businesses, which means the underlying is reality-priced: no multiple expansion on entry. Most of the return comes from operating the business, not from the wrapper. Which is the point. The wrapper doesn't generate returns. It pre-qualifies entrants for capital the underlying couldn't raise on its own, at the cost of years those entrants spend learning to describe themselves in the vocabulary of a category the market already understands. The wrapper fits before the wrapper earns anything. That is not a small thing.
Every search fund entrant has, at some level, chosen the legibility of a category that already exists over an opportunity the category doesn't yet describe. Some produce extraordinary returns. Some don't. What they share is the decision to pay the wrapper's quoted price. At the system level, that price isn't subtracted from some pool of capital. It's subtracted from attention. Every year spent inside the wrapper is a year spent looking in the direction the wrapper already points.
Legibility is the pre-tax form of validation. The system can pay it because the system can see it. What it can't see, it can't price.
Markets only move goods they can name. Audiences only amplify claims they can compress. Peers only reward paths they can describe. Every system runs on what it can parse. The premium on legibility is not a bug; in most forms, it's just the price of transacting in language. A market with no shared vocabulary can't settle a trade. Some legibility floor is load-bearing for every system in the house.
The failure mode shows up when legibility stops being the surface of a system and starts being the underlying, at the point where the wrapper becomes the asset the system is actually pricing. Selection pressure inside that system then runs inverse to the thing the system claims to select for. Capital starts choosing companies for their names instead of their cash flows. Audiences start choosing sentences for their screenshot-ability instead of their accuracy. Peers start rewarding careers for their fit inside existing categories instead of the work they actually produce. At the self layer, attention starts sorting signal by how nameable it already is, which means new signal, the kind that doesn't yet have a word, gets sorted out before it surfaces.
This has a body count in other domains.
In 2021, two economists at UC San Diego published a study in Science Advances. They examined papers in top-ranked psychology, economics, and general-interest journals whose findings could not be replicated. Non-replicable papers were cited roughly 153 times more than papers that replicated cleanly. In Nature and Science, the citation gap widened to 300 times. The mechanism was ordinary. When findings were more interesting, reviewers applied lower standards. Only about 12% of post-replication citations acknowledged the failure. The most legible findings, the surprising, the counterintuitive, the shareable, propagated the hardest and survived the longest, regardless of whether they were true.
Inside the academic literature that built the modern social sciences, the most read, most taught, most cited papers of the last fifty years correlate inversely with the thing science claims to produce, which is reliable knowledge. The signal the system is optimizing for is the legibility of the finding. Reliability is a different problem the system has chosen, at the margin, not to solve. Science has the same bug gossip has: the interesting version is the one that travels.
The strange part isn't that anyone in those rooms is behaving badly. The strange part is that they're all behaving well. The reviewer is not cynical. The CEO is not lying. The investor in the thematic ETF is not being defrauded. Everyone is responding rationally to a local incentive. The failure is at the population level, across enough iterations, until the thing the system was supposed to select for disappears into the thing it actually selects for.
What it selects for
The effect compounds. The first three currencies can be reissued, and the fourth one cannot. The attention you point at categories the system already knows how to name is the same attention you can't point anywhere else. The perceptions that didn't get categorized didn't get remembered. They didn't get a word.
At the population level, this stops being a psychological problem and becomes an evolutionary one. The system selects for people whose attention is already aligned with the categories it rewards. Those people have more students, more hires, more followers, more of whatever version of reproduction their layer uses. The next cohort of capital allocators, audiences, peers, and selves is marginally better tuned to the existing categories than the last one was. More legible to the system. Less likely to notice what the system doesn't yet know to name.
The way a net with one mesh size catches one size of fish. Over enough seasons, that's the only fish left in the water.
The obvious counter is that legibility is also how quality scales. Apple is the most legible company on earth and among the most valuable. Berkshire's annual letters are a legibility machine that has allocated capital better than most active managers for fifty years. The replication crisis itself was surfaced by making non-replication legible through meta-analysis. Legibility creates value at every layer. The question is not whether. The question is whether there is a threshold.
The ETF data marks it cleanly. A diversified fund compounds at roughly the market return with minor drag. A thematic fund underperforms non-linearly: the steeper the narrative premium at launch, the longer the five-year drag. The inflection is measurable, and it sits at the moment a wrapper gets launched specifically because a theme is the most legible it has ever been. That is the moment the wrapper's price is furthest ahead of the underlying. Call it the saturation point. Below it, legibility is selling the thing. Above it, legibility is selling itself.
Every room has a saturation point. A career wrapper tips the year it stops describing the work you still want to do. A piece of writing tips the moment the sentence that would travel contradicts the sentence that would be true. An institution tips when the question it can ask out loud no longer includes the question it actually needs to answer. These aren't hypothetical. They are the ordinary experience of working inside any category long enough.
What is at stake, across those tipping points, is not abstract.
At the capital layer, the stakes are compounding. A portfolio paying six percent a year in legibility premium retires with roughly half of the portfolio that isn't. Over thirty years, the six percent becomes fifty.
At the audience layer, the stakes are publication. The essay that would have been read by fewer people but would have said something only you could say is the essay that never got written, because every paragraph pulled toward a shape the audience already had a folder for. You built an audience. The audience you needed would have been built by the essay you didn't write.
At the peer layer, the stakes are the career the system never let you build. The market didn't reject it. Your peers didn't have a sentence for it yet, and a sentence is the minimum viable unit of peer allocation. The promotion you got pointed at the wrapper. The promotion you didn't get is the one the wrapper couldn't see.
At the self layer, the stakes are every question you never knew you didn't ask. Every color you mixed a little differently in your head and never named. Every observation you almost made at dinner that would have become a pattern in five years if you'd held onto it long enough to develop vocabulary for. These don't show up anywhere. They don't become drafts in drawers. They become not-yours.
Stacked up across a life, that's what's actually being priced. The legibility premium doesn't just cost returns or engagement or status or awareness. It costs the version of your work, your judgment, and your attention that the system doesn't yet have a folder for. That version was the one with the asymmetric upside.
Medallion is the far pole of the illegibility spectrum. Renaissance Technologies' fund returned about 39% annualized, net of fees, between 1988 and 2018. A hundred dollars at inception would have compounded to roughly four hundred million. A hundred dollars in the S&P reached about nineteen hundred. The fund is closed to outside capital. Its strategies are never explained. The edge is a recipe that works because nobody wrote it down. The moment you could hand it to someone else in a deck, a ticker, or a sentence an LP could repeat at dinner, it would stop working.
That's what an edge looks like when it survives. Most edges eventually become legible, which is why most edges eventually dissolve.
What gets seen gets selected. What hasn't been categorized yet gets traded for what has.
Six months after the round closed, the founder is updating her LinkedIn bio.
The three words that got her round are on the page in front of her. Platform, data, AI. They had worked in the deck. They worked in the press release. They worked in the onboarding materials her Head of Sales sent every new enterprise prospect. She had heard herself use them at three conferences, always in the same order. Her customers had started describing the product back to her in the same three words.
She reread the draft. It was accurate enough. It was also a sentence written by the round that closed eight months ago, not by the product she's building now. The product has moved. It stopped being a scheduling tool sometime around month two, and drifted, without anyone quite naming it, into something closer to an inference layer for the billing workflows of specialty networks. The sentence she wrote nine months ago hasn't moved at all. She sits for a minute and tries to remember what she would have said if asked a year ago, before the rewrite.
She can't quite get the words back. Anyone who has watched a product quietly outgrow its pitch knows the particular uncomfortable silence of that moment.
A painter sees a color no catalog lists and mixes it before naming it. Name the color first and it becomes the nearest color in the catalog. Mix it first and the catalog has to expand to describe it. The founder named her next move before she could mix it. Eight months of work inside the wrapper has, gradually, replaced her vocabulary for the thing she was actually building with her vocabulary for the thing she sold. She is not confused. She is exactly as legible as the round required her to be. What she is realizing, six months in, is that the legibility has a denomination, and the denomination is measured in a currency she didn't notice she was spending.
Somewhere else this afternoon, a doctor's aging report sits closed on his desk. Half a million dollars of work is waiting to be named before it can be paid. Same distance. Different wrapper.
Most of what matters now lives in that distance. The distance is the only thing the premium can't reach.