The Compressed Version#
Intelligence becomes cheap and abundant; everything that intelligence cannot manufacture becomes correspondingly precious. Energy, land, trust, verification, licensure, physical presence, and genuine human relationships appreciate. Competent cognitive labor depreciates - especially the junior rung that used to make experts. Firms must adopt and mostly won't profit from it; consumers capture most of the surplus in forms the statistics won't measure well. Aggregate job loss is easier to absorb than it looks (demography); the apprenticeship gap is not. And the deepest structural risk is that our decision-making institutions - legal, political, diplomatic - run at human deliberation speed while the systems they govern increasingly do not.
The single most useful thing to build in this environment is anything that makes verification cheap, since that is the scarce good the entire equilibrium now rotates around.
Concretely, that means:
- Cryptographic provenance at the point of capture - attestation in the sensor, not inference after the fact → C8
- Institutional vouching with skin in the game - reputation that costs something to stake
- Synchronous high-stakes evaluation to replace the dead asynchronous signals
- Systems that preserve legible human accountability at exactly the points where liability and judgment still bind
- Balance sheets and underwriting that can price residual risk - or deployment stops where insurance refuses → Uncertainty 6
These are the complements cheap cognition cannot supply itself, and they are undersupplied for the same reason safety is - each is a public good with private cost, so the market will build them late, and the first one to build them well will own a great deal.
If the paragraph above is wrong, it is most likely wrong in one of two ways: verification itself gets automated, so it never becomes the scarce good (Uncertainty 5); or the list of things intelligence cannot manufacture shrinks faster than the decade this advice assumes (Uncertainty 7). Watch those two pages; the rest of the corpus is elaboration.
A third way to be wrong is quieter: Taiwan soft-fails without a war (Uncertainty 4 gray zone) and every timeline slips in quarters while the spine still looks intact. Watch freight insurance and license denials, not only fleets.
See also: Game 5 · Meaning · Startups - ideas (what to build from this list) · Notation · Dependencies