AITakeover Tracker
Open Research Project

The Doomsday Clock for Jobs

A single, transparent, reproducible number updated daily — tracking how close AI is to making white-collar work economically obsolete.

What is Takeover Tracker?

Think of it as a composite index — like the S&P 500 for job displacement. Instead of tracking stock prices, we track signals from labor markets, corporate adoption, AI capabilities, economic indicators, public sentiment, and regulation. The result is a single score from 0% to 100%.

0%

Pre-AI baseline

circa 2019

~20%

Current estimate

AI assists, early automation

50%

Role restructuring

Entry-level hiring halved

100%

Economic non-viability

No cost advantage for humans

HOW IT WORKS

Daily Pipeline

Every morning at 6:00 AM UTC, an automated pipeline collects fresh data, analyzes it with AI, scores 24 sector-category pairs, and produces one number. The entire process takes about 15 minutes.

1

Collect

Pull data from 9 sources: RSS feeds, BLS, FRED, GDELT, Reddit, Hacker News, SEC filings, AI benchmarks, and more.

2

Extract

AI classifies each signal by category, affected sectors, sentiment, impact magnitude, and direction.

3

Score

Tier 1 hard data z-scores + Tier 2 math aggregation of displacement levels — no LLM scoring calls.

4

Calculate

Pure math: two-tier 50/50 blend, hype discount, single EMA smoothing, 0.5 rounding.

5

Summarize

AI generates a headline, summary, and key developments for the day's score movement.

Design Principles

Hype Resistant

Reality signals (labor data + corporate adoption + economic indicators) are compared against hype signals. When hype exceeds 2× reality, capability and sentiment scores are discounted.

Fully Transparent

Every data source, weight, formula, and rubric prompt is published. The full methodology is available for anyone to audit, critique, or reproduce.

Source Credibility

Not all data is equal. Peer-reviewed research and BLS data count at full weight (1.0×), while CEO predictions (0.5×) and social media (0.3×) are heavily discounted.

Two-Tier Hybrid

Hard quantitative data (50%) is blended with AI-classified signals (50%). Neither tier can dominate, and the hard data serves as a reality anchor.

Self-Consistency

Each scoring call runs 3 times at low temperature. The median is taken, filtering out outlier responses and producing more stable, reproducible scores.

EMA Smoothing

A single exponential moving average (α=0.08, ~25-day window) filters noise while responding to genuine trends. Scores are rounded to the nearest 0.5.

DATA SOURCES

What Feeds the Index

Every signal is categorized into one of six categories, each with a specific weight reflecting its reliability and relevance.

CategoryWeightSources
Labor Market DataTier 1+2
25%BLS, FRED, WARN Firehose, Indeed
Corporate AdoptionTier 2
25%SEC EDGAR, 25 RSS feeds, earnings calls
AI CapabilitiesTier 2
15%ArtificialAnalysis.ai, Epoch AI, benchmarks
Economic IndicatorsTier 1+2
15%FRED, BLS, GDELT
Sentiment & HypeTier 2
10%Reddit, Hacker News, Google Trends
Regulatory SignalsTier 2
10%GDELT, RSS feeds

Disclaimer

This index is an experimental research project, not financial or career advice. The score reflects a model-based estimate using available public data and AI analysis. It should not be used as the sole basis for any employment, investment, or policy decisions.

Get Started

See how AI affects your career

Search 47,000+ job titles across 1,100+ occupations or take a personalized risk assessment to understand your exposure.