the methodology

most AI safety rankings are static snapshots. someone asks a chatbot "which jobs are safe?" and publishes the answer as a list. that's not a system — that's a conversation.

hardhat runs a live index. every profession is rescored daily across 4 quantitative dimensions. the scores drift, respond to market signals, and mean-revert — just like any real index.

the index is deterministic: same day = same scores for everyone. there's no personalization, no A/B testing on scores, no black box. if you and a stranger check the same profession on the same day, you see the exact same number.

we built this because someone had to treat career safety like a real market — not a blog post.

the index covers 350+ occupations — from construction trades to healthcare, mechanics, tech, and services. if you can enter the career through training, apprenticeship, or certification rather than a four-year degree, it belongs in the index.

career safety doesn't exist in a vacuum. when the construction sector booms, electricians thrive. when tech stocks crater, software engineers sweat. the index captures this through four real-world data layers.

every profession category is mapped to a real-world sector ETF proxy — XHB for construction, XLK for tech, XLV for healthcare, VNQ for real estate, and so on. the engine pulls quarterly sentiment data from these proxies and uses it to modulate scores by ~10%.

live sentiment analysis runs daily across 40+ career-focused communities. the system monitors real conversations from workers in every sector — construction crews, nurses, software engineers, truckers — and scores the collective mood using natural language processing. when electricians are posting about overtime and bidding wars, that's bullish. when tech workers are sharing layoff threads, that's bearish. this live signal patches directly into the scoring engine alongside the ETF proxies.

IPEDS training pipelines track how many people are entering each profession through accredited programs. if nursing schools are graduating fewer students while demand grows, that's a supply shortage — and the index boosts demand scores accordingly. if a field is flooded with graduates, scores reflect the oversaturation.

QCEW employment data from the bureau of labor statistics tracks quarterly employment levels and wage trends across every industry sector. year-over-year hiring growth and wage movement feed directly into market stability scores — if your sector is adding jobs and raising pay, that shows up.

this isn't a stock ticker. it's a reality anchor. federal employment data, training pipeline supply, worker sentiment, and market signals all converge to keep scores grounded in what's actually happening — not speculation.

all data is refreshed automatically — federal data sources are pulled via API and regenerated on schedule, while live sentiment updates daily. no manual spreadsheets, no stale data sitting around for months.

every profession on the index carries a GPT exposure coefficient (γ) — a number between 0 and 1 that measures how much of the job's task portfolio is exposed to large language model capabilities.

this isn't opinion. it's derived from peer-reviewed research analyzing thousands of occupational tasks against GPT-4's demonstrated capabilities. a γ of 0.95 means nearly all tasks in that job overlap with what AI can already do. a γ of 0.15 means almost none do.

the exposure score directly calibrates the AI resistance dimension — the single largest weight (~40%) in the survival score. professions with high physical requirements and low GPT exposure (like firefighters, γ = 0.18) score well. professions with high exposure and no physical barrier (like data entry, γ = 0.97) score poorly.

you'll see the γ badge on every profession's detail page. it's the single most important number for understanding a career's AI risk.

• we don't sell rankings. no profession can pay to score higher.
• we don't make predictions. the index reflects current automation exposure, not future guarantees.
• we don't replace human judgment. scores are a starting point — not financial advice.
• we don't personalize scores. everyone sees the same number on the same day. that's the point.

automation doesn't respect borders. if AI replaces translators in germany, it replaces them everywhere. the index reflects that.

every data point feeding into the survival index is collected from a global scope — not just US markets. we aggregate labor statistics, automation research, wage data, and industry reports from OECD nations, emerging markets, and international labor organizations to build a complete picture of how each profession is evolving worldwide.

this matters because local snapshots lie. a profession might look safe in one country while getting automated across the ocean. global collection catches early signals — if warehouse workers are being replaced in south korea today, the ripple hits the US within 18 months.

the result: more variation, more accurate prediction data, and fewer blind spots. when you see a score on hardhat, it's informed by what's happening to that profession everywhere on earth — not just your zip code.