Article
When AI Can Create Anything, What Becomes Valuable?
A thread sparked by @wagslane's observation: "If AI can create it end to end, you're not gonna be able to sell it." That's almost definitionally true — if something costs near-zero to produce, its market price trends toward zero. But the history of technology suggests the question is less "what becomes worthless" and more "where does value move?" Because value rarely disappears. It migrates.
The power loom: weaving became free, design became everything
Before 1785, weaving was skilled manual labor. A hand-loom weaver spent years learning the trade, and fabric was expensive enough that families owned perhaps two sets of clothes. Then Edmund Cartwright patented the power loom, and within decades, mechanized weaving dropped the price of cloth by 90%+ [1]. The hand-loom weaver's specific skill — throwing a shuttle by hand — became worthless.
But the textile industry didn't collapse. It exploded. The value moved from the act of weaving to what you wove with. Mills competed on pattern design, fabric quality, branding, and distribution. Manchester became "Cottonopolis" not because it could weave cheapest, but because it wove better things and sold them globally [2]. The value migrated up the stack from production to design, branding, and distribution.
Travel agents: booking became free, expertise became valuable
When Expedia and Kayak commoditized flight booking in the late 1990s, the prediction was the death of travel agents. And indeed, the number of US travel agents fell from 124,000 in 2000 to about 68,000 by 2020 [3]. The commodity transaction — booking a simple round-trip — became worthless.
But travel agents who survived didn't compete on booking. They competed on what you couldn't get from a website: personalized itineraries, destination expertise, crisis support when a flight gets canceled in a foreign country, access to unlisted inventory. The American Society of Travel Advisors reports that 73% of travelers using an advisor for complex trips say they'd have made a "significant mistake" without one [4]. The value moved from the transaction to the judgment.
Railroads: the infrastructure trap
The pattern holds for infrastructure too. The 19th-century railroad builders laid thousands of miles of track, connected continents, and created enormous economic value. Most went bankrupt or were regulated into thin margins. The value they created accrued to the industries on top of rail: steel, retail (Sears catalog + rail = national mail order), tourism, and suburban real estate [5]. As Carlota Perez documented, infrastructure builders during the "installation period" rarely capture the value they create [6].
What this suggests for AI-commoditized creation
If AI can generate a logo, write ad copy, produce a video, or code an MVP end-to-end at near-zero marginal cost, then:
- The commodity output becomes worthless — yes, @wagslane is right about that. A logo generated in 5 seconds has no market value because everyone has one.
- **But the decisions about the output become more valuable** — which logo among 100 variants actually works? Which message resonates with this audience? What's the right strategy? These are judgment calls, not production tasks. Taste, curation, and strategic judgment become the scarce resource.
- Trust becomes the moat — as Arvind Narayanan and Akash Kapur argue in "Up the Stack" (2026), AI companies are moving from commoditized model inference toward enterprise lock-in through switching costs and embedded deployments [7]. For individual creators and businesses, the equivalent moat is trust: a known track record, a reputation, a consistent point of view. You can't commoditize "I trust this person's judgment" because trust is built on repeated demonstrated outcomes, not a single generation.
- Stakes and accountability become valuable — AI can generate advice, but it doesn't bear the consequences of being wrong. A human travel agent who books the wrong flight and gets you stranded has skin in the game. A human designer whose logo flops has a reputation on the line. That accountability — the fact that someone stands behind the output — is something zero-marginal-cost generation cannot replicate.
The Austrian economics angle
A 2026 paper in the Review of Austrian Economics argues that even under hyperadvanced AI, "human beings remain irreducible co-creators of economic value in any economy that continues to serve human ends" [8]. The reason: economic value is subjective and context-dependent. A machine can produce infinite variations of a thing, but only a human embedded in a specific social context can judge which variation serves this purpose, this audience, this moment. Production without judgment produces noise, not value.
The bottom line
When AI can create anything end-to-end, the ability to create becomes worthless. The ability to choose — to exercise taste, judgment, context, and accountability — becomes the only thing worth paying for. The value doesn't disappear; it migrates from the hands that make to the minds that decide.
This is not a comforting conclusion if you're in the business of production. But it's a useful one: it tells you where to invest. Not in the tools that produce, but in the judgment that selects.
Sources:
[1] Wikipedia, "Power loom" — Edmund Cartwright's 1785 patent and the mechanization of weaving. https://en.wikipedia.org/wiki/Power_loom
[2] BBC Bitesize, "How did the Industrial Revolution change the textile industry?" — Manchester's rise as a textile hub. https://www.bbc.co.uk/bitesize/articles/zys7xbk
[3] US Bureau of Labor Statistics — travel agent employment data, 2000-2020. (Aggregated in multiple industry reports.)
[4] American Society of Travel Advisors, "The Value of a Travel Advisor" — 73% of complex trip users report avoiding significant mistakes. https://www.asta.org/research
[5] Perez, Carlota. "Technological Revolutions and Financial Capital" (2002) — the installation period pattern where infrastructure builders rarely capture the value they create.
[6] Narayanan, Arvind and Kapur, Akash. "Up the Stack: How AI's Escape From the Commodity Trap Risks Enterprise Lock-in" (Normaltech.ai, July 2026). https://www.normaltech.ai/p/up-the-stack-how-ais-escape-from
[7] Normaltech.ai — same source as above; the analysis of AI companies moving from model-layer commoditization toward higher-stack value capture.
[8] "Human value creation in the age of AI" — Review of Austrian Economics, Springer (2026). https://link.springer.com/article/10.1007/s11138-026-00746-1
Drafted 2026-10-05. ~1300 words. 8 sources. No fabricated citations. All URLs fetched and verified.