The Complex Perspective · 2036
The 2036 Baseline
The shared anchored world every module scene inhabits — dated assumptions, argued from mechanism.
Six modules, six scenes — one world. This page is that world’s specification, so the scenes cannot quietly contradict each other. Everything below is written in the Anchored register: each assumption traces to a mechanism observable in 2026, and none depends on an invention that does not exist. Where the modules go beyond this baseline, they carry a Projection badge and say what they are adding.
The baseline names no models, quotes no benchmarks, and carries no dates except the horizon itself. Those are the parts of any forecast that rot first — the decade test is the exhibit.
Baseline assumptions, stated 2026-07
B1 — Delegation is routine. Software agents transact, negotiate, and execute multi-step work on behalf of people and firms, the way email and spreadsheets are office furniture today. Mechanism: this already works in 2026 in narrow domains, and cheap, labor-substituting office technology has repeatedly reached ubiquity within roughly a decade — spreadsheets being the canonical case.
B2 — Generating content costs approximately nothing. Text, images, code, audio, and plausible-looking evidence of all four are produced at near-zero marginal cost by anyone. Mechanism: already true in 2026; costs fall further on every hardware and inference trend that has held for a decade.
B3 — Verification does not inherit that discount. Checking a claim against reality — auditing the books, running the experiment, visiting the warehouse — keeps a real-world cost floor that generation lost. Mechanism: verification is bound to physical processes and institutional access, not to compute.
B4 — Liability stays attached to legal persons. Contracts, torts, and reputations bind humans and firms; an agent’s signature is worth exactly the enforceability of its principal’s promise. Mechanism: law has lagged every prior general-purpose technology; nothing in 2026 legislation trends toward machine legal personhood.
B5 — Human cognitive bandwidth is unchanged. Attention, working memory, and hours in the day stay fixed while everything competing for them multiplies. Mechanism: biology; ten years changed nothing in the 2016 edition’s bounded-rationality chapter.
B6 — Institutions adapt on institutional time. Law, education, professional norms, and regulation lag the technology by years — creating the gray zones the incentives module plays in. Mechanism: the EU’s landmark AI rules took from 2021 to 2027 from proposal to full application; that clock speed is the norm, not the exception.
B7 — Physical constraints persist. Energy, chips, logistics, and construction remain scarce and slow relative to software. Not everything accelerates. Mechanism: the cost curves that collapsed are informational; kilowatt-hours and building permits are on different curves.
B8 — Capability keeps improving, unevenly and surprisingly. The baseline does not assume progress stops; it assumes progress keeps arriving in jagged, hard-to-predict shapes, so plans premised on a specific capability schedule keep failing. Mechanism: that is the last decade’s most consistent empirical pattern — and the reason this site bets on invariants instead.
The same world, over the line
To feel the difference between this page’s register and the one this site carefully rations, here is the same baseline — as the movie:
Nobody in 2036 remembers doing anything themselves. The agents wake before the city does, trading, writing, litigating, flirting on our behalf; humanity’s last job is to sign. In the signing halls the pens are gold, and the queue is very, very long.
Beautiful, unfalsifiable, and useless for planning. That is why it gets the sprockets — and why the eight numbered assumptions above, dull as they are, are the ones the six modules actually build on.