The Signature Economy: Why Perfect Images Became Worthless and Named Humans Became the Asset

A summary of Sam Altman's recent comments has been ricocheting around the internet, and its claim is stark: AI can now generate images more beautiful than most humans could ever make, and the value people assign to them rounds to approximately zero. Not discounted. Zero. The phrasing in the viral version may be paraphrase, but the verified version of Altman's thinking says the same thing in his own words; in a recent wide-ranging interview he put it plainly: "Human values are valuable precisely because they are human," noting that people simply prefer the human behind the thing. Sit with who's saying it. The man selling infinite image supply is telling you the images aren't where the value is, which, as admissions against interest go, is about as costly a signal as this industry produces.

Here's the thing, though: the art market figured this out centuries before the models did.

The market that always priced the signature

Take two physically identical paintings: same canvas, same pigments, same brushwork quality. Attribute one to the master and one to his workshop, and the price gap isn't 20 percent; it's often a hundredfold. Prove a work is a forgery (an object whose craft, by definition, fooled experts) and its value doesn't drop, it evaporates, even though not a single molecule changed. The most famous painting sale in history, the Salvator Mundi at $450 million, was substantially a bet on attribution: on whose hand, not what surface. The art market has been running the controlled experiment for five hundred years, and the result never varies. Strip the signature and the object collapses. The pixels, the pigment, the perfection, were never the point. Provenance was the product all along.

AI just ran that same experiment at planetary scale in about three years. It made technically flawless images infinite and free, and the market responded exactly as the forgery precedent predicted: perfection without provenance priced itself at nothing. This shouldn't have surprised anyone, least of all the builders who optimized for resolution, fidelity, and realism on the assumption that output equals value. Abundance didn't destroy the value of beauty. It revealed where the value had been hiding the whole time.

Brands were the original signature technology

Now the part that matters for anyone running a company: this is not new territory for branding. It's the territory branding was invented on. Potters pressed marks into Roman amphorae, medieval guilds hallmarked silver, and trademark law exists because a maker's mark (a signature at commercial scale) was how buyers found an accountable human in a market of anonymous goods. A brand, at its origin, is exactly the thing Altman says people are paying for: proof that a specific someone stands behind the object and answers for it.

The industrial century let brands drift into being aesthetic systems, and the drift was affordable while good surfaces were scarce. That subsidy just ended. When every company can generate flawless visuals, flawless copy, and flawless everything, surface excellence carries precisely zero information, and the brand's original job (identifying the accountable human) becomes its entire job again. Abundance didn't make branding obsolete. It fired branding from its decorative side hustle and rehired it for the position it was created for.

Apple just paid extra for the receipt

If the theory needs a contemporary proof at the highest level of brand craft, Apple supplied one. When Apple TV refreshed its identity in late 2025, the new logo animation wasn't rendered, though a render would have been indistinguishable on screen. It was sculpted from solid glass and shot practically, a weeks-long production at a London creative studio working with Apple's design team, with color returning to the mark as an homage to the six-stripe 1977 logo. The sonic logo, the few seconds of sound that now open every show, was composed by Finneas O'Connell, a named, Grammy-winning human attached to six notes. And then came the move that makes this a signature-economy case study rather than a production anecdote: the behind-the-scenes film was published. The glass, the rigs, the hands.

Run the economics. On your television, photons are photons; the glass buys nothing a render couldn't fake. Which means the entire premium purchased exactly one asset: the true, verifiable sentence "humans made this," with names attached. Apple, the most sophisticated brand operator alive, looked at a world drowning in generated perfection and concluded that provenance was worth a weeks-long shoot and a famous composer, and that the making-of wasn't a bonus feature. It was the deliverable. The same news cycle offered the control group: Coca-Cola's AI-generated Christmas ad, produced at a fraction of the cost and mocked for its glitches, surface without signature. One company bought pixels. The other bought proof, and published it.

The signature stack

So what does signing your name look like at company scale? Concretely, in descending order of weight: named humans on the work, meaning real bylines with real records, founders and operators visible and attached to positions someone could disagree with; receipts, the specific claims and numbers a person is accountable for, because evidence is a signature machines can verify; costly signals, the visible bets against your own short-term interest that can't be faked precisely because they hurt; and provenance discipline in the work itself, knowing and showing what's human-authored, what's machine-produced, and who reviewed what. Every post on this site carries a name for this reason, and it's why our published AI rules open with the sentence they do: a named human stands behind every sentence that ships. We wrote that rule before this discourse arrived; the discourse is now explaining why it was never optional. There's an AEO dividend too, and it isn't sentimental: answer engines weight authorship, entity consistency, and verifiable claims, which means the signature stack is simultaneously how humans learn to trust you and how machines decide to cite you.

Signed systems, not signed pixels

None of this argues against using the machines; Altman's own distinction survives scrutiny, and it happens to be our operating model: when a human with real intent uses AI as an instrument, the work keeps its value, because the signature moves up a level. Our client image engines produce hundreds of on-brand images a year, and no one signs each pixel; the signature is on the system: an identity a human authored, constraints a human set, a review lane a named person owns. Same with everything else in the stack. The question was never "did a machine touch this." It's "does a specific human answer for this," and any company that can't answer the second question is publishing forgeries of itself.

The honest limit

The predictable next move is already visible: humanity as an aesthetic. Hand-drawn wobble, confession-toned copy, performative behind-the-scenes, the "undeniably human" style guide. The wave of Apple-imitators is already forming, and it will be slopified within a year, because anything identifiable gets optimized and anything optimized gets cheap; we've written that loop already. What separates Apple's glass from the coming imitation is that the cost was real and checkable: a named studio, a named composer, a physical object that exists. The signal worked because it hurt and could be verified, not because it looked handmade. The signature that survives isn't a texture, it's a liability: a person who was accountable before the work shipped and remains accountable after it fails. That can't be generated, not because models lack the pixels, but because accountability isn't made of pixels. It's made of someone having something to lose, which is, conveniently, the one input that was never abundant.