Model changelog & validation

v1.3 — 2026-04-15

Added the evidence layer (real layoff/automation events) and the demand layer (job-posting trends) so scores are anchored to what is actually happening, not just task exposure.

• Introduced an evidence layer: documented layoffs, hiring freezes, and AI-deployment announcements now feed each occupation.
• Introduced a demand layer: tracks job-posting volume so falling demand can corroborate a high score.
• Recalibrated bucket thresholds (low / medium / high) against the expanded signal set.

v1.2 — 2026-03-10

Added the capability clock and the Pressure Index to express not just how exposed a job is, but how fast the pressure is building.

• Added the capability clock: an estimated horizon for when AI capability crosses each task.
• Added the Pressure Index combining exposure, current usage momentum, and timeline.
• Smoothed timeline estimates across related occupations for internal consistency.

v1.1 — 2026-02-12

Split scores into cognitive (AI) and physical (robotics) sub-scores after adding a dedicated physical-automation signal.

• Added the physicalAutomation signal (robotics, self-driving, self-checkout deployment).
• Exposure is now max(cognitive AI exposure, physical automation) instead of cognitive-only.
• Published dual AI / robot sub-scores so the headline number is decomposable.

v1.0 — 2026-01-15

Initial public release of the ReplacedYet AI-Risk Index.

• Cognitive exposure model combining LLM task exposure, current AI usage, AIOE, and routine-cognitive share.
• Physical / social protection discount for hands-on, in-person, high-empathy work.
• Automate-vs-augment split and 0–100 headline score across all tracked occupations.

Cite as: ReplacedYet AI-Risk Index (2026), https://replacedyet.com. Dataset · Methodology