The Centaur Window · Index · as of October 27, 2026

Centaur Window Index

How long human direction will still improve results, by occupation, scored on five factors you can reweight.

Companion to the paper The Centaur Window, which explains the five factors, the evidence, and what would change the scores.

A centaur window is the period in which a human steering an AI system beats both the human alone and the system alone on the domain's own scoring function. The Window Score (WS) estimates how long that period has left in each occupation, on a scale from 1 (closed or closing) to 5 (persistent edge). It is the weighted mean of the five factor scores below, rounded to the nearest whole number; the weights start equal, and you can change them. It is not a forecast of job loss. Many roles persist off the frontier.

Window Score

1 · Closed or closingAutonomous systems lead the ranking. Non-use is already uncompetitive.
2 · Short, roughly 1 to 5 yearsFormal outputs, dense feedback. Human steering still helps, briefly.
3 · Medium, roughly 5 to 15 yearsSelection, verification, or framing still change outcomes.
4 · Long or partialContext, integration, or accountability keep a human in the loop.
5 · Persistent edgeEmbodiment, real-time stakes, or open-ended framing dominate.

Five factors, each scored 1 (shortens the window) to 5 (lengthens it)

V · Verification costHow hard it is to score the output objectively. Lean proofs and test suites score 1; taste and trust score 5.
D · Feedback scarcityHow slow or sparse the learning loop is. Self-play and compilers score 1; decade-long outcomes score 5.
C · Residual complementarityWhat a human still adds: framing, context, embodiment, accountability.
G · Capability gapHow far current systems sit below the human frontier on the core tasks, net of economic pressure to adopt.
S · Scorer endogeneityHow much the ranked population controls its own scoring (licensure, peer review, credit norms).
Summary

The ladder by occupational group

Each dot is a group's median composite score; the line spans its lowest and highest occupation. Scores recompute when you change the weights below.

Detail · 35 occupations

Occupations

Rescore: adjust factor weights (default 1.0 each)

V D C G S WS Why, and the next move
Method

How the scores were built

Each occupation is scored 1 to 5 on the five factors. The composite is the weighted mean, and the Window Score (WS) is the composite rounded to the nearest whole number. Equal weights reproduce the paper's index. The last column explains each score in plain words and ends with a next move: one of the five moves in Section 7 of the paper, applied to that occupation. The moves are judgments, like the scores.

The factor scores are judgments informed by task-exposure research. No exposure study measures whether a human-directed system still beats an autonomous one, which is the quantity the window describes. Treat the index as a structured argument you can disagree with one cell at a time.

Sources the scores draw on

Companion to the paper The Centaur Window (October 2026). Scores are directional composites and will be revised as evidence changes.