# Uncertainty 7 - Complement erosion: the scarce list has a half-life

← [Part VI](README.md) · [Index](../README.md)

---

[Uncertainty 5](learned-verification.md) attacks the first half of the spine - *capability grows fastest where verification is cheap*. This entry attacks the second half: **value accrues to what intelligence cannot manufacture**. That claim silently assumes the set of things intelligence cannot manufacture is *stable* over the corpus horizon. The strongest reasons to doubt that come from inside the corpus itself, which is exactly why this page exists: without it, the spine's second half had no adversarial entry, and invariant 6 was being satisfied by accident rather than by audit.

## The corpus supplies its own ammunition

Reread the [complement list](../09-macro/assets.md) as a target list:

- **Physical presence and skilled trades** - the corpus's own [swing variable](../03-domains/physical/robotics/cost-curves.md). If robot cost curves run, the one *low-concentration* complement (the row that made the distribution story partly hopeful) becomes manufacturable first.
- **Energy and land-with-power** - scarce on permitting timescales, not on 2030s technology timescales. The [inference collapse](../01-substrate/inference-economics.md) attacks demand per unit of capability from one side; AI-accelerated supply (storage learning curves, FOAK geothermal) attacks from the other. Today load growth is winning - [energy](../01-substrate/energy.md) documents the price prints - but "currently winning" is not "inelastic."
- **Distribution and default placement** - agent-mediated commerce disintermediates exactly the placement premium the list prices. ([Game 5](../02-games/5-information.md) argues verification scarcity *rises*, which cuts the other way for trust; the tension is real and unresolved.)
- **Regulatory licenses and liability capacity** - not eroded by technology at all, but repricable by a single legislative act. Political scarcity is the least durable kind, because it has an owner with an incentive to sell.

What erodes slowest: land and zoning, proprietary *real-time* flow data (self-renewing by construction), and liability capacity while [Uncertainty 6](correlated-risk.md) stays unresolved.

## Dated horizons (working table)

If the objection is right, each row needs a horizon rather than a permanent label. Working estimates for the base case - not probabilities stamped in Part V, but orderings the [indicators](../07-indicators/) can falsify:

| Complement | Horizon before material erosion | Gate |
|---|---|---|
| Physical presence / skilled trades | 2032–2040 (structured first) | Robot cost curves + [B12](../07-indicators/diffusion/capability.md) |
| Energy scarcity (not land) | 2030–2035 if supply wins | Permitting + industrial capacity + efficiency; [U2](power-permitting.md) |
| Distribution / default placement | 2028–2032 in agent-heavy verticals | Agent-mediated commerce share |
| Regulatory licenses | Event-driven; can reprice in one session | Scope-of-practice bills; fiscal pressure |
| Liability capacity | Stays scarce while correlation unpriced | [U6](correlated-risk.md); carrier products |
| Land + zoning | Beyond 2040 in most metros | Political, not technical |
| Real-time proprietary flow data | Self-renewing; horizon is access law, not tech | Platform regulation, not model quality |

The table is the operational form of "half-life." A miss on any single row is a local rewrite; a simultaneous miss on the slow rows (land, liability, real-time data) is the abundance scenario [assets](../09-macro/assets.md) does not currently price.

**Do not re-score assets on one robot pilot.** Erosion of physical presence requires B12-scale economics outside structured environments, not a single warehouse win. Warehouse success is the *start* of the sequence, not list-death.

**Land+zoning is the residual bet.** If every other row erodes and this one holds, the distributional story concentrates rather than vanishes - a different map of who wins, not abundance for all. → [Assets](../09-macro/assets.md)

## The historical base rate cuts both ways

Past general-purpose technologies eroded their era's complement lists, and where the surplus went is the empirical question this page turns on. Agricultural mechanization manufactured what labor scarcity had protected, and the surplus mostly reached consumers as cheaper food - farmland rents concentrated, but the *share* of income flowing through the eroded rows collapsed. Electrification commoditized generation while the surplus lodged in the adjacent uneroded rows: distribution rights, appliance manufacture, urban land near the grid. The pattern across cases is that **erosion rarely delivers surplus cleanly to consumers or cleanly to complement-owners - it migrates rent to whichever adjacent row erodes slowest.** Applied here, that favors a middle reading the corpus's distributional pages do not currently price: the assets story can be wrong about *which* complements concentrate wealth while being right that *some* row does, with land-with-power and liability capacity as the historically-shaped candidates for where the rent migrates. That is a weaker claim than the one [assets](../09-macro/assets.md) makes, and if this page fires, it is the claim the corpus retreats to before conceding abundance.

## What the objection does and does not break

The claim degrades gracefully rather than snapping. The honest restatement is not "the list is wrong" but **"the list has a half-life, and each row needs a dated horizon"** - value accrues to the *lagging edge of the erosion frontier*, which keeps moving.

**The strongest counter the corpus can offer is that erosion is gated by the rows it would erode.** Manufacturing physical presence requires robots, which require leading-edge fabs, rare-earth motors, and assembly plants sited on permitted land with power - the erosion of row one is a construction project running through rows two and four. Manufacturing energy abundance requires exactly the land, interconnection, and licensed installers whose scarcity it would end. Each act of erosion pays a toll to the surviving complements, which means the list cannot collapse all at once: it erodes in sequence, and the sequence itself re-prices the later rows upward while they wait. This is a genuine negative feedback, not a debating point - but it bounds the *rate* of erosion, not the *fact* of it, which is why this page claims a half-life rather than immortality.

The erosion sequence is also spine-coupled. Which complement gets manufactured first is set by verification cost - robot manipulation improves where success is checkable, energy tech where output is metered - so the first half of the spine governs the decay schedule of the second half. If [Uncertainty 5](learned-verification.md) fires and verification stops ordering capability, erosion stops being sequential and the half-life shortens across every row at once. The two red-team pages are not independent failure modes; 5 firing is an accelerant for 7.

The exposed predictions are distributional. If complements erode faster than they concentrate, the surplus [Game 3](../02-games/3-firms.md) routes to complement-owners flows instead to consumers as deflationary abundance: [assets](../09-macro/assets.md) overstates wealth concentration, and the [prices divergence](../09-macro/prices.md) - the claimed defining political-economy fact of the 2030s - narrows from the late 2030s as the inflating categories get manufactured too.

## Leading indicators

| Signal | Reading |
|---|---|
| Wholesale price at datacenter-heavy nodes falling despite continued load growth | Efficiency/supply winning the energy race; the top row eroding |
| Robot deployment economics outside structured environments ([B12](../07-indicators/diffusion/capability.md)) | Physical presence becoming manufacturable |
| Share of commerce completed agent-to-agent without a placement premium | Distribution row eroding |
| Scope-of-practice expansions or license liberalization passing state legislatures | Political rows repricing |

## What resolution forces

**If erosion runs fast:** rewrite the spine's second half in frontier form; re-score the [assets](../09-macro/assets.md) and [prices](../09-macro/prices.md) claims; Part IX's concentration story softens toward consumer-surplus abundance. Log under [scoring](../08-method/scoring.md) rule 5 as a framework revision.

**If the list holds through ~2032:** the base case strengthens, but date-stamp the rows anyway - this page's claim is about half-life, and a half-life is confirmed slowly.

**Failure mode of this page:** treating any single substitution as list-death. The distribution claim survives in weakened form as long as *one* row stays durable - and land with zoning is a very durable row. The symmetric failure also exists: reading every price decline as erosion. A datacenter-node price falling because demand collapsed ([inference economics](../01-substrate/inference-economics.md) run too far) confirms a different uncertainty entirely; the indicator only counts when prices fall *despite* load growth, which is why the first row of the table is worded the way it is.

---

**Related:** [Game 3](../02-games/3-firms.md) · [Assets](../09-macro/assets.md) · [Robot cost curves](../03-domains/physical/robotics/cost-curves.md) · [Inference economics](../01-substrate/inference-economics.md) · [Uncertainty 5](learned-verification.md)

**Previous:** [Uncertainty 6 - correlated risk](correlated-risk.md) · **Back to:** [Part VI hub](README.md)