The Coming Repricing

There is a quiet conversation happening across the AI industry that hasn't broken into mainstream discourse yet. It goes like this: if artificial intelligence becomes good enough to evaluate any claim against reality — at the moment of decision, in everyone's pocket, for free — what happens to all the businesses, institutions, and credentials whose value rests on people not being able to evaluate them?

The honest answer is: most of them break.

Not all at once. Not symmetrically. But the substrate of the current economy is being repriced, and the repricing is structural — not narrative.

This piece is an attempt to map that repricing concretely. What does the economy look like right now, by structure rather than by category? What's the function-priced share versus the narrative-priced share? What survives when the narrative scaffolding gets evaluated? What inflates? And the question nobody answers honestly — where does the value flow when this finishes?

I'll try to do this without moralizing. The lens I'm using treats reality as structure, geometry, and probability rather than as story. That means describing what is, not what should be. Reality doesn't care which outcome we prefer.

Part 1 — The economy right now, by structure

Global GDP is about $110 trillion. The conventional sectoral breakdown (financial services, healthcare, manufacturing, and so on) tells you what people are doing for money. It doesn't tell you what they're being paid for.

A more useful split is function versus narrative: how much of each sector's value capture is paying for something that materially happens in reality, versus how much is paying for positioning, signaling, brand trust as substitute for evaluation, credentialing, or information asymmetry.

This is the rough ledger:

If you take that ledger seriously — and you should, because it's roughly what every honest operator inside these sectors will tell you in private — roughly 40–45% of global GDP is narrative-priced.

Forty-five trillion dollars per year of activity exists because of information asymmetry, brand trust as a substitute for evaluation, signaling, credentialing, gatekeeping, or coordination friction.

That $45 trillion is the collapse surface.

Part 2 — Why people actually buy

Before the repricing question, an honest read of consumer behavior. If you decompose discretionary spend above subsistence, here's the rough motivation split:

About 95% of discretionary spend is not about function. People know this implicitly. They don't act on it because they don't have the time, expertise, or cognitive bandwidth to evaluate every claim. The brand premium is a tax on cognitive bandwidth.

A personal AI removes that tax.

Part 3 — What the truth-mirror actually does

Imagine a model that runs in your pocket and does the following, instantly and personally:

• →Tells you whether a product claim is functionally accurate or narrative repackaging of a commodity.
• →Tells you whether a service provider is genuinely specialized or templated.
• →Tells you whether a credential signals real skill or is decorative.
• →Tells you whether an investment is statistically likely to beat the index after fees.
• →Tells you whether a news story is reporting events or framing them.
• →Tells you whether a job listing is what it says or something quieter.
• →Tells you whether a contractor's bid is competitive or marked up.

The technology to do all of this exists in 2026. It's not deployed personally and ubiquitously yet — but the trajectory is uncontroversial. By 2030, every meaningful purchase decision routes through a personal evaluation layer first. By 2035, that layer is the default cognitive substrate.

The narrative tax — the friction that lets the $45 trillion exist — gets removed.

Part 4 — What compresses, what inflates

Compresses

Total compression: roughly $30 trillion of current value capture migrates or simply disappears.

Inflates

The new GDP geometry by 2035

Roughly, in current dollars on a ~$170T total:

The $45 trillion narrative surface compresses to a $5–8 trillion residual. Around $35–40 trillion of value either migrates to the AI/compute layer, migrates to atom-movement through relative inflation, migrates to genuine specialists, or simply ceases to exist as economic activity — a productivity dividend that doesn't all get recaptured.

This is the most consequential reorganization of economic activity in industrial history, and it's running on a 10–15 year timeline.

Part 5 — The six tiers

Where does the value flow when this finishes? Honest read:

Tier 1 — The compute oligarchs

Five to ten AI model providers and the compute infrastructure owners behind them. Capture 5–10% of global economic activity through royalties on cognition. Equivalent to the railroad, oil, and electrical-grid barons of their eras combined, because cognition is more fundamental than transport or energy. Maybe 2,000–10,000 people globally hold meaningful equity here. The most concentrated wealth tier in human history by absolute scale.

Tier 2 — The operator class

Individual operators running AI-augmented businesses. A skilled solo operator with good judgment and AI access runs what required 50–200 people in 2020. Maybe 10–50 million people globally. The top tier reaches $10–100M net worth; the broader class reaches $1–10M.

Tier 3 — Genuine specialists

Surgeons, engineers, researchers, top trades, top performers. AI amplifies their capability rather than replacing it. Real wages rise 30–100%. Maybe 50–200 million people globally.

Tier 4 — Atom-movement labor

Construction, manufacturing, logistics, hospitality, food service. Real wages rise 20–40% because supply is constrained by training and physical presence. Maybe 2–3 billion people. Better off in absolute terms. Much worse off in relative terms.

Tier 5 — Displaced knowledge workers

Mid-tier consultants, lawyers, accountants, analysts, copywriters, junior bankers, mid-managers. The compression hits this tier hardest and fastest, roughly 2026–2030. Some retrain into Tier 2. Many don't. Real income falls or stagnates. Perhaps 100–300 million people affected globally.

Tier 6 — The non-participants

Those without specialist skill, operator capability, atom-movement work, or capital. Live on transfer payments, productivity-dividend programs, or informal economies. Largest tier by headcount. Material conditions depend entirely on whether the productivity dividend gets distributed.

Part 6 — The kings question

The structural read is uncomfortable: the gap between Tier 1 and Tier 6 in 2035 is wider than at any point in human history. Possibly wider than at any point in any civilization. Tier 1 individual net worths reach hundreds of billions to trillions. Tier 6 lives on universal income equivalents.

But the floor question — what happens to Tier 6 in absolute terms — depends on which scenario we end up in.

Scenario A — productivity dividend distributed

Personal AI becomes utility-priced like electricity. Cheap or free at baseline. Open-source models stay viable. Atom-movement wages absorb displaced workers over a painful decade. Tier 6 has a much higher absolute standard of living than Tier 3 had in 2020 — better healthcare, better information, better decision support, better baseline cognition. The gap explodes, but the floor rises significantly.

Scenario B — productivity dividend captured

Compute access stratifies. Top AI is gated by price and policy. Open-source models are deliberately neutered or out-classed. Tier 1 extracts maximum rent. Information asymmetry returns at a higher level. The gap is severe and the floor stagnates or falls. Tier 6 conditions deteriorate.

Which scenario unfolds depends on

• →Regulation of model providers — whether antitrust pressure or regulatory capture wins.
• →Open-source model viability — DeepSeek-style projects continuing to keep frontier accessible.
• →Whether compute commoditizes on a Moore's-Law curve or stabilizes as oligopoly.
• →Energy supply and pricing — compute is energy-bound; cheap energy distributes the dividend.
• →Political response to Tier 5 displacement — historically, when mid-tier work disappears, revolutions follow.

The economic gravity pulls toward concentration. Only deliberate counter-structures distribute the dividend.

Why this concentration is more severe than past tech revolutions

• →Cognition is more fundamental than energy or transport. Every other technology amplified human capability. AI replaces or amplifies the human capability layer itself. The substrate becomes more valuable than anything built on it.
• →Network effects and data moats compound. Each provider gets stronger with use.
• →Capital intensity at the frontier. Training the next-generation model costs $1–10B+. Only oligarchs can compete.
• →Regulatory capture incentives are enormous. Incumbents are already lobbying for the structure that locks them in.
• →There is no inherent competition mechanism for the frontier. Once one provider is six months ahead, that's a permanent practical lead. Catch-up requires national-scale investment.

This is the part the casual AI discourse doesn't reach yet. The conversation is mostly about will AI replace jobs. The structural conversation is about who owns the cognitive utility layer of the new economy. That question is being decided right now — in policy, in market structure, in capital allocation. Most people are not at that table.

Part 7 — What this means if you're alive in this transition

If you're a mid-tier knowledge worker — and “mid-tier” here means anyone whose work is reasoning, analysis, writing, or coordination that another human could review and verify — your compression window is roughly 2026 to 2030. Probably faster. Your two viable paths are:

1. Up into Tier 2 (operator)

Become an operator who uses AI as leverage rather than competition. Run a small business or operating unit with AI doing the work of an extended team. The skill stack: judgment, taste, distribution, customer relationship, willingness to make decisions. Many existing knowledge workers have most of these but lack the operator instinct — the willingness to own outcomes rather than process inputs.

2. Up into Tier 3 (specialist)

Become a genuine specialist whose skill is hard to fake and benefits from AI rather than competes with it. The skill stack: deep technical capability, verifiable output, embodied expertise. A harder pivot for most because it requires years of training.

If you're already in Tier 3 or 4

The trajectory is reasonably good. Real wages rise. Demand for your skill grows. The competition isn't AI — it's other humans noticing this is the right place to be and crowding in over a decade. Move now while supply is constrained.

If you're in Tier 1 or near it

Your structural question is different: how does the era you're building hold together? Concentration this severe with no political safety valve does not end peacefully in any historical precedent. The savvy players in Tier 1 will spend serious capital on distribution mechanisms — not because they're moral, but because they're strategic.

If you're in Tier 6

Your single highest-leverage move is to acquire a personal AI literacy that lets you operate above the Tier 6 floor regardless of what employers do. The most underpriced education in the world right now is “how to operate as a human with AI as your leverage.” It costs almost nothing to learn. It is worth more than every credential underneath it.

Part 8 — The position I'm taking

I run Junction AI. We install AI operating systems for individual operators and small businesses. From the structural map above, that means we are infrastructure for Tier 2 formation.

I'm not pretending that's a neutral position. Equipping the operator class is a bet on Scenario A — the bet that if enough people learn to operate AI as leverage rather than being operated on by it, the productivity dividend distributes more broadly. Whether that bet pays off depends on factors well outside what any individual operator controls.

But it's the bet I can actually run. Building one more compute oligarch is not on the table. Building one more displaced Tier 5 worker is the default trajectory if we do nothing. Equipping Tier 2 emergence is the lever that's available, at the scale I can operate it.

The teaching mission and the business mission are the same mission: equip people to operate above the floor, on the upper side of the curve, without contributing to the lower side falling.

That's the operator-leverage thesis, honestly named.

Closing

There's a version of this conversation that treats the coming repricing as a thing to be afraid of. There's another version that treats it as inevitable progress. Both are forms of narrative comfort.

The structural read is that it's a reorganization. Some people are positioned for it. Most aren't yet. The window to reposition is shorter than most realize and longer than the doomers say — roughly 2026–2032 for individual operators. After that the tiers stratify and motion between them becomes much harder, the way motion between class tiers became hard after 1900, after 1950, after 1980.

This is the decade.

If you want to think clearly about your position in it, the questions I'd ask are:

• →What in my current work survives evaluation by a personal AI?
• →What in my current work is paying me for narrative scaffolding?
• →Where do I currently sit on the six-tier map?
• →What's the closest reachable tier that's structurally above my current one?
• →What's the next concrete move toward it?

Reality measures everyone. The framework that lets you exempt yourself from measurement is not the framework. It's a costume of the framework.

Reality will let us know how this played out.