Predictions

A research programme should make bets it can lose. These are the programme’s falsifiable predictions: three deposited on the Open Science Framework with explicit refutation conditions before the underlying paper was submitted for review, and the structural theses the papers stake out. Original deposit dates remain visible on OSF regardless of any later revision. This page will record refutations as plainly as vindications.

Pre-registered

Deposited publicly with vindication and refutation conditions, from The Shrinking Synthesis.

openPre-registered · May 2026Central estimate ~2042

AI's foundational institutional response arrives 2037–2049

A criterion-meeting "Phase 4" institutional event for AI — a named formal statutory, constitutional, regulatory or treaty instrument establishing a new category of governance authority, responsive to documented chaos rather than anticipatory — will emerge in the window 2037 to 2049. The central estimate of ~2042 is anchored in cohort-replacement arithmetic; the paper’s headline 2037–2047 window names the empirical clustering of the radio, television and internet cycles.

Refuted if

  • A criterion-meeting event emerges before 2037 — the generational-floor mechanism is disconfirmed.
  • No criterion-meeting event has emerged by 2049 — the historical compression pattern does not extend to AI.

The EU AI Act (2024) is excluded as anticipatory regardless of its regulatory content. All three outcomes are scientifically informative.

OSF pre-registration From: The Shrinking Synthesis Evaluable from 2037; fully resolvable by 2049
openPre-registered · May 2026EFPE cohort: born ~1997–2012

The settlement will be architected by the cohort that grew up with AI

The decisive intellectual roles in AI’s Phase 4 institutional design — principal drafters, rapporteurs, frame-setters — will feature members of the Early-Formative-Partial-Exposure cohort (those whose neuroplasticity windows encompassed ChatGPT’s November 2022 diffusion onset), within mixed-cohort coalitions where older cohorts provide political leadership and sponsorship.

Refuted if

  • Decisive intellectual roles show no measurable EFPE-cohort presence, being filled instead by cohorts whose plasticity windows closed before 2022.
  • AI Natives (born ~2025 onward) appear as principal Phase 4 architects — the framework predicts they consolidate the settlement, not design it.

Refutation would disconfirm cohort replacement as the operative driver of the generational floor, leaving the timing prediction potentially vindicated but its mechanism undefined.

OSF pre-registration From: The Shrinking Synthesis Assessed against the eventual instrument’s drafting record
openPre-registered · May 2026GDPR-era institutional grammar

The settlement’s content will trace to internet-era individual-rights frames

AI’s Phase 4 event will build on the individual-rights frames developed in the internet cycle — privacy, fairness, due process for automated decisions, dignity in algorithmic systems — with institutional grammar traceable to the GDPR-era frameworks. A novel mechanism-form (compute governance, model auditing, capability registration) still confirms the prediction, provided the underlying conceptual claim is the protection of individual rights.

Refuted if

  • The Phase 4 instrument abandons the individual-rights conceptual lineage altogether, establishing governance categories with no traceable connection to those frames.

Refutation would disconfirm organisational learning as the mechanism of cycle-to-cycle institutional uptake, removing one of the framework’s three theoretical legs.

OSF pre-registration From: The Shrinking Synthesis Assessed against the eventual instrument’s conceptual lineage

Structural theses

Falsifiable in direction, staked out in the papers — not yet pre-registered with dated windows.

Firms that adopt AI as pure augmentation — keeping headcount and layering AI on top — will be structurally undercut on cost by AI-native entrants. The middle position is permanently more expensive, not a stable destination.

FalsifierAugmenter firms sustaining cost-parity with AI-native competitors at scale, across sectors, without converging on the AI-native operating model.

From: Adopt or Be Undercut

Displaced cognitive workers will not be re-absorbed at scale onto a new rung of human work, because AI removes the condition that made every previous absorption work: the new work being work only humans could do.

FalsifierThe emergence of a durable, mass-employment category of new work that AI systems cannot perform more cheaply — a genuine new rung rather than a transitional niche.

From: The Decoupling of Productivity from Human Hands