Notation - recurring shorthand#
Contents
Terms that carry load across the corpus. Definitions are operational, not dictionary.
| Term | Meaning here | Where it matters |
|---|---|---|
| Master asymmetry | Capability grows fastest where verification / ground truth is cheap | Data, all of Part III |
| Ground-truth cost | Cost of knowing whether an output is correct (labels, experiments, P&L, exploits) | Domain ordering |
| Effective compute | Hardware × dollars × algorithmic efficiency; the real capability engine | Compute |
| Inelastic complement | Input intelligence cannot manufacture at will (energy, land, licenses, trust, distribution, physical presence…) | Game 3, Assets |
| Red Queen | Adoption mandatory; surplus competed away to customers in competitive markets | Game 3 |
| Apprenticeship gap | Junior tasks automated → firms stop hiring juniors → seniors missing in 2040 | Game 4, Uncertainty 3 |
| J-curve | Productivity statistics lag adoption until workflows redesign | 2028–2032 |
| Leaky bucket | Frontier leads diffuse (weights, people, distillation) ~30–50%/yr | Game 2 |
| Compute-governance trap | Export controls = competitive lever and only future verification tool; using one spends the other | Bipolar, Game 2 |
| Salient incident | Forcing event that opens an ~18-month rule-writing window | Game 2 |
| Three RSI governors | Verification, physical supply chain, financial (neutral-rate) limits on takeoff | Uncertainty 1 |
| Two economies | Cognition-intensive deflation vs physical/care/energy inflation | Prices |
| Financial governor | Success raises returns → raises discount rates → dearer next capex round | Rates, Capital |
| Shelf-readiness | Having a drafted architecture before the incident window opens | C2 |
| Outcome pricing | Billing for results not seats; revealed belief in reliability | B5, 2026–2028 |
| Two-year moat | Any fixed capability level becomes ~free within ~24 months; frontier access is a wasting asset | Inference economics, Game 3 |
| Revenue bar / capex test | Does AI revenue reach ~$400–700B/yr by 2030, net of financing circularity | Compute, Capital, A2 |
| Verification scarcity | Post-collapse regime where generation is free and checking is the priced good | Game 5, Compressed, C8 |
| Complement half-life | Scarce-complement rows erode on dated horizons, not forever | Uncertainty 7, Assets |
| Gray-zone Taiwan | Soft fail of fab continuity (insurance, licenses, slip) without crisis headline | Uncertainty 4, A6 |
| Process vs outcome verification | U5 middle case: process checks learnable, outcome truth still expensive | Uncertainty 5, Law |
| Behind-the-meter | Captive generation / bilateral power routing around public interconnect queues | Energy, Uncertainty 2 |
How a term earns a row#
A term is admitted here when it does real work in at least two parts of the corpus and would otherwise be re-derived in each. The definitions are operational on purpose: "inelastic complement" is defined by what happens to its price under demand, not by a category list, because the category list changes (Uncertainty 7) while the operational test does not. When a page uses one of these terms in a way that fails the operational test, that is a bug in the page, not a looseness in the term.
The vocabulary also serves scoring. A prediction phrased in a pinned term cannot quietly migrate its meaning after the fact - "the apprenticeship gap widened" resolves against the definition in this table, not against whatever the phrase has come to mean in 2032. Shared shorthand is a commitment device before it is a convenience.
Three families#
The table sorts into three kinds of object, and the kind tells you how a term can fail. Asymmetries and structural facts (master asymmetry, two economies, leaky bucket, Red Queen) describe gradients; they fail by flattening, and each has an indicator watching for that. Governors and traps (three RSI governors, financial governor, compute-governance trap) describe negative feedback; they fail by being bypassed, which is usually a bigger event than the term itself - a bypassed governor means a branch of Part IV is live. Gaps and windows (apprenticeship gap, J-curve, salient incident, shelf-readiness) describe timing mismatches; they fail by closing early, which is generally good news and is tracked in Part VI as upside skew. A term failing in a way its family does not predict - a gradient that reverses rather than flattens, say - is a sign the framework, not just the term, needs the dependency index treatment.
Cross-links#
- Spine: Thesis · Compressed
- Falsifiers: Uncertainties · Steelman
- What rests on what: Dependency index