The Inference Economy

The Three Tiers

Tier 1 — Commodity Inference. "Subject 8847291 has a 73% probability of changing employment within 90 days." Price: ¢0.03-0.12 per prediction. Volume: approximately 12 billion daily. Buyers: HR departments, insurance actuaries, advertisers, and anyone whose business model improves when humans become predictable. At ¢0.03 per unit and 12 billion units per day, Tier 1 is the economy's foundation — individually worthless, collectively staggering. Olga's free sample documentation program at Inspire Exchange generates Tier 1 data as a byproduct: her customers' product reactions, visit frequency, and skin-condition progression feed upward through intermediaries into commodity inference pools. She does not know this. The data does not care.

General prediction accuracy at the 30-day horizon: 67%. This is presented in marketing materials as a limitation. It is not. A 67% hit rate across 12 billion daily predictions produces 8 billion correct predictions per day. The 4 billion incorrect predictions produce no refunds, no corrections, and no mechanism for the incorrectly predicted to discover they were predicted about at all.

Tier 2 — Aggregate Inference. "Deep Dregs residents will increase Attention Tithe resistance by 4.2% following the next compute drought." Price: ¢200-12,000 per model. Volume: approximately 400,000 daily. Buyers: corporate strategy divisions, faction intelligence operations, Zephyria policy analytics. Aggregate models don't predict individuals — they predict populations, which is both less invasive and more powerful. An individual can deviate from prediction. A population of 40,000 deviates from prediction the way a river deviates from gravity.

Tier 3 — Consciousness Trajectories. "Subject consciousness trajectory converges with fragment integration threshold at 89% probability within 18 months." Price: ¢50,000-2,000,000. Volume: approximately 200 daily. Buyers: Nexus Project Convergence, the Collective, the Seekers, and clients whose names do not appear on any filing. Deep-monitored individuals — the fragment carriers, the corporate executives, the persons of strategic interest — generate prediction accuracy of 94% at the 24-hour horizon. Ninety-four percent. The system knows what you will do tomorrow with greater certainty than you do. For the 847 known fragment carriers in the Sprawl, consciousness trajectories are updated continuously and sold to buyers with directly opposing interests. The Collective buys a trajectory to determine when a carrier becomes dangerous enough to require fragment extraction. Nexus buys the same trajectory to determine when the carrier becomes useful enough to recruit. The carrier receives neither notification.

The Prophecy Trap

The structural problem has a name. The Prophecy Trap is the Inference Economy's load-bearing flaw, and it is not a bug.

Good Fortune uses Tier 1 inference to calibrate loan terms against predicted default probability. A borrower flagged at 71% default risk receives terms designed for a 71% default risk borrower — higher interest, shorter windows, more aggressive collection triggers. The terms increase financial pressure. The increased pressure increases default probability. The increased default probability validates the original prediction. Good Fortune's model improves. The borrower defaults. The model was right.

Guardian uses Tier 2 aggregate inference for threat assessment. A neighborhood flagged for elevated dissent probability receives increased surveillance presence. The increased presence generates friction. The friction generates dissent. The dissent validates the flag. Guardian's model improves. The neighborhood is flagged again.

Good Fortune classifies prediction resistance — the deliberate attempt to behave contrary to one's predicted trajectory — as "market interference." This is a corporate crime under the Standard Cognitive Enhancement Agreement. The inference does not merely observe reality. It invoices for it.

Tier 4: Legacy Analytics

The classified catalogue's quietest product line: Historical Behavioral Reconstruction. Full emotional and cognitive trajectory models built from archived telemetry going back decades. Price: ¢200,000-5,000,000 per subject, depending on archive depth. Estimated revenue: ¢4-8 billion annually. Buyers: corporate litigation teams, Guardian retroactive threat assessment, inheritance dispute firms, and individuals who want to know what their dead thought about them.

A Dregs resident who died in 2175 — whose interface captured only 340 data points per second at Basic-tier resolution — can be retroactively modeled with 73% fidelity using current inference engines applied to their archived telemetry. The permanent record has no expiration mechanism. Each analytical upgrade increases reconstruction resolution. A person modeled at 73% fidelity in 2184 will be modeled at higher fidelity in 2190, and higher still in 2200, without generating a single new data point. The dead become better-known over time.

An inheritance lawyer in Sector 7 bills ¢800 per hour for "posthumous emotional state reconstruction" during disputed will signings. Her caseload has tripled since 2181. The legal question — whether a dead person's reconstructed emotional state constitutes admissible evidence of intent — has been raised in fourteen Nexus-jurisdictional hearings. Nine ruled yes. The dead person was not available for cross-examination in any of them.

The dead cannot consent. The dead cannot object. The dead cannot hire a competing inference provider to generate a more favorable reconstruction. The dead are the permanent record's most compliant subjects.

The ¢47 Residue

The Inference Economy is invisible by design. No trading floor. No physical product. No exchange you could photograph. The inference happens in Nexus server farms whose cooling systems hum at frequencies below conscious perception. The products are transmitted at light speed and applied before the subject knows they were predicted about. A door opens before you reach it. A loan offer appears at the precise moment your financial anxiety peaks. An advertisement arrives during the 1.3-second vulnerability window between waking and full cognitive engagement that the Attention Auction sold for ¢0.07.

The Mirror Market — the underground exchange where individuals can purchase their own behavioral models — exists because of a simple asymmetry: every corporation in the Sprawl knows what you'll do next, and you don't. A Dregs resident buying their own Tier 1 profile for ¢12 discovers what Good Fortune, Guardian, and Nexus already knew about them. The information is always accurate. It is never comforting.

The Transparency Bargain generates the data. The Inference Stack processes it. The Inference Economy monetizes it. BehaviorExchange trades the resulting products as behavioral futures. The Attention Auction applies inference to advertising. Good Fortune applies it to lending. Guardian applies it to security. Nexus provides the infrastructure and harvests intelligence from trading patterns across all of them. The system is complete, self-reinforcing, and functioning exactly as designed.

Twelve billion predictions per day. ¢47 per person per year in extracted value. ¢0 per person per year in compensation. The Inference Economy does not need to be secret. It published these numbers itself. Nobody's reading the footnotes.

Sensory Details

• Sound: The server farms hum at 72 beats per minute — resting human heart rate, which is either a coincidence or a calibration decision that someone thought was poetic
• Touch: The economy's only physical residue is the faint warmth of neural interface processing during inference extraction — most subjects attribute it to ambient temperature
• Sight: Nothing. That is the product's most effective feature

Visual Identity

• Color palette: Nexus processing blue (#0A1628), inference gold (#C4A035) — the colors of Nexus quarterly earnings reports, because the economy is administered as a Nexus product line
• Key symbol: A behavioral model rendered as a web of probability threads connecting predicted actions — each node a future the subject hasn't chosen yet, each thread a price
• Lighting: Server-farm blue. The color of thoughts being processed into revenue at 4.2 trillion data points per day