Who is this for
Career changers, parents advising students, founders rethinking team comp, professionals at year 5–15 wondering whether to deepen or pivot. Equally useful to nomads weighing reskilling vs geographic arbitrage.
How long is your skill good for?
An 80-skill calculator that estimates the working half-life of any given skill at your current proficiency. Numbers blend AI-substitution risk, classical-automation degree, and market saturation. Server-rendered, no JavaScript required, no data stored, results encoded in the URL so you can share them.
Eight ways to use this calculator
What it computes
Years of useful working horizon for the chosen skill at chosen proficiency. Numbers blend three pressures: AI-substitution risk, classical-automation degree, and market saturation. Calibrated to WEF Future of Jobs 2025, McKinsey MGI, and the Anthropic Economic Index.
When to use it
Before committing 6–24 months to a credential. Before pricing a contract. Before hiring or growing a team. At career inflection points: end of a role, return from sabbatical, capital event, relocation.
Where the answer applies
Globally. Half-lives are not country-specific because labour-market frictions slow the front edge but do not change long-run trajectory. Local markets affect when the squeeze hits, not whether.
Why half-life matters more than salary
A high-paying skill on a 3-year half-life pays back differently than a mid-paying skill on a 12-year half-life. Lifetime earnings, optionality, and bargaining power all compound over horizon, not over current wage.
Which adjacents to consider
The result lists three skills in the same cluster with longer horizons. They are pivot targets, not replacements. Pair the short-horizon skill with one adjacent in the next 6–18 months to widen runway.
Whose calibration this is
Three-source weighted blend: WEF survey of 1,000 employers (2023, 2025); McKinsey MGI substitution analyses (2017–2024); Anthropic Economic Index Q4-2024 + Q1-2025. Principal editorial overrides where evidence diverges.
How to act on a short half-life
Three moves in order: (1) Stop deepening the skill itself beyond proficient. (2) Spend 20% of skill-building hours on the named adjacent. (3) Move into a role where the short-horizon skill is one input among many, not the entire job.
Totality lens · 32 points to ponder · 16 user POV + 16 developer POV · this tool
User POV — for the practitioner using this tool
Eight dimensions
A professional in any of the 100 surveyed occupations can in principle compute the half-life of their currently-marketable skill stack and project the years until material obsolescence. Few do; most assume their skills will remain valuable through retirement, an assumption that holds for fewer occupations every year as automation and tooling-shifts compound. The diagnostic surfaces the half-life numerically, which is more useful than vague obsolescence-anxiety.
A typical knowledge-worker skill stack carries a half-life of 3-7 years today, down from 10-15 years two decades ago. Some sub-stacks (deep technical specialism, regulatory expertise, relationship-capital) carry longer half-lives; others (tool-chain proficiency, syntactic-language fluency, certification-currency) carry shorter half-lives. Honest computation produces a number; the number guides re-skilling investment.
Of professionals who run the diagnostic and act on the result, perhaps 65-75 percent successfully shift one or two stack-components per 18-month cycle, maintaining marketability through career-decades. The remaining 25-35 percent either underestimate the rate (skills obsolete faster than they re-skill) or over-react (re-skill prematurely into stacks that themselves obsolete before producing return). Calibration matters more than activity.
What works: computing the half-life honestly using current job-market signals not historical norms; investing 8-15 percent of working hours in next-stack acquisition during currency of present-stack; publishing in the new stack while still earning in the old; building relationships in the new ecosystem before the income shift forces it; treating obsolescence as gradient not cliff.
What does not work: assuming current stack will hold value through retirement; treating credentials as permanent (credentials decay with the underlying skill they signal); over-investing in fashionable-stacks that themselves will obsolete (chasing the latest framework); waiting for the income-shift before re-skilling (the re-skilling window closes at the moment income depends on it).
The most common pitfall is conflating skill-currency with skill-value. A professional who maintains skill-currency in a declining-value stack (e.g., a niche legacy technology that retains certifications but loses contracts) shows green metrics on the diagnostic while actually losing market position. The diagnostic separates currency (am I current in my stack) from value (does my stack still command value).
Counter-intuitively, the most-defensible stacks are usually combinations of two or three half-life-different sub-stacks rather than deep specialism in any one. A long-half-life sub-stack (regulatory + relationship-capital) anchored to a short-half-life sub-stack (current toolchain) produces a composite with longer effective half-life than either component alone. Pure deep specialism is fragile to the half-life of that specialism.
The single highest-leverage move at the half-life stage is to identify the longest-half-life component of the current stack and consciously deepen it while rotating shorter-half-life components. Most professionals do the opposite — they deepen the short-half-life components (because that is what current employer rewards) and let the long-half-life components atrophy.
Eight user intents
Knowledge-workers in the 100 surveyed occupations, mid-career professionals (5-25 years post-credential) considering pivot or re-skill, education-policy planners building curriculum recommendations, employers running workforce-planning. Particularly relevant for occupations in software, finance, healthcare, education, manufacturing, professional services.
The irreducible essence: compute half-life honestly, identify longest and shortest half-life sub-stacks, deepen long sub-stacks while rotating short sub-stacks, allocate 8-15 percent of working hours to next-stack acquisition while still earning in current stack.
Best applied annually for ongoing tracking, immediately after major industry-shift events (new tooling release, new regulation, new automation capability), and during pre-pivot evaluation. Less useful in stable career-mid-decades where rate-of-change is lower; most useful in 5-15 years post-credential when re-skilling capacity is high.
Geographically variable. Half-life accelerates in technology-hub geographies (San Francisco, London, Singapore, Bengaluru) and decelerates in stability-anchored geographies (regulatory-heavy fields in slower-changing regions). The diagnostic accepts geography parameter to weight half-life appropriately for local market.
Skill half-life is misunderstood because professionals experience their own skill stack continuously and gradually, missing the externally-visible step-changes in market valuation. The diagnostic forces externally-anchored measurement (current job-market signals) over internally-anchored measurement (subjective sense of expertise).
Highest-leverage stack-components vary by occupation. For software: relationship-capital + system-design + one current language; tooling-currency rotates. For finance: regulatory + relationship-capital + analytical foundations; product-specific stays current. For healthcare: clinical judgement + regulatory + one specialty current; diagnostic-tooling rotates. For education: pedagogical-craft + subject-mastery; tool-stack rotates.
Trust: hiring managers in target geography speaking honestly (which is rare; many speak diplomatically); BLS or equivalent national occupational forecast data; professional-association longitudinal earnings surveys; honest peer-conversation with professionals 5-10 years ahead of you. Ignore: vendor-marketing of "future-proof skills" (always self-serving); generic "future of work" content (rarely occupation-specific).
Proceed by listing your current stack components honestly, computing half-life of each via diagnostic, identifying the 1-2 components with longest half-life and 1-2 with shortest, allocating learning-time toward deepening long-life components while building next-stack components for short-life ones. Re-run annually with updated job-market signals.
Developer POV — for the architect, maintainer, AI tool, future contributor
Eight dev dimensions
Skill half-life composes from data/atlas-skill-half-life.php (100 occupations × stack-components × half-life-bands × geography-weights), the engine in includes/atlas-skill-half-life-engine.php (computes composite half-life given current stack input + geography), and the obsolescence-narrative composer at includes/atlas-half-life-narrative.php. Single-file render; deterministic; no APIs.
SoftwareApplication on the toolkit page; ItemList of 100 occupations; Occupation schema on per-occupation breakouts. The diagnostic-flow emits HowTo schema with steps. Result-display emits Recommendation schema per stack-component analysis. Structured-data testing validates against schema.org.
Forward to /toolkit/cofounder-fit/, /toolkit/founder-burnout/. Outward to /careers/{role}/, /occupations/{slug}/. Cross-content injector tokens: "skill", "half-life", "obsolescence", "re-skill", "career-pivot". Link weaver hyperlinks the 100 occupation names + major skill categories.
Payload ~20 KB total. Diagnostic JS ~7 KB minified vanilla. Render ~250-400 ms server-side. LCP typically the page hero. CLS near zero. The 100-occupation registry lazy-loads only the occupation the user selects (not all 100 at once).
Stack-component input renders one-component-at-a-time on mobile, multi-input on desktop. Tap-targets ≥48px. Geography selector uses native <select>. Result-display collapses gracefully at narrow viewports. Sticky progress-bar.
Native semantic HTML throughout: <fieldset> + <legend> per stack-component group, <input type="number"> + <label> for years-of-experience inputs, native <select> for geography/occupation. Keyboard-accessible. Focus-visible outline. Color-blind-safe palette for half-life-bands (uses navy/silver/gold not red/yellow/green).
URL: /toolkit/skill-half-life/. Canonical. OG. Twitter. Sitemap. IndexNow on edit. SoftwareApplication schema. Per-occupation breakout pages at /toolkit/skill-half-life/{occupation}/ provide indexable surface for occupation-specific queries (ramping in v149.4+).
To add a new occupation: append to data/atlas-skill-half-life.php with required fields (slug, name, sector, stack_components[], half_life_per_component, geography_weights). Diagnostic auto-picks up. To add a geography: append to geography_weights map. Total ship: ~30 min per occupation including handwritten obsolescence-narrative.
Eight dev intents
Joint. Half-life data refreshed annually (occupational-forecast data updates on annual cycle from BLS, Eurostat, national stats). Stack-component schema reviewed semi-annually for tooling shifts.
Tech: PHP 8.3 + vanilla JS. Helpers: ajg_atlas_half_life_compute(), ajg_atlas_half_life_occupation(), ajg_atlas_half_life_narrative(). No framework dependency. Single-file render. JS bundled inline (~7 KB).
Annual data refresh aligned with BLS Occupational Outlook Handbook update + Eurostat occupational-forecast revision. IndexNow on edit. Lastmod tracks file mtime per ajg_crawl_lastmod_lookup().
Code: data/atlas-skill-half-life.php (100 occupations schema), includes/atlas-skill-half-life-engine.php (compute logic), includes/atlas-half-life-narrative.php (obsolescence prose), toolkit/skill-half-life.php (page).
Why 100 occupations rather than ISCO-08 full set (~436): the 100 are the occupations most relevant to AJG audience (knowledge-workers in tech, finance, healthcare, education, manufacturing, professional services) and produce dense per-occupation handwritten content; the full ISCO set would dilute density.
Critical: atlas-skill-half-life.php, atlas-skill-half-life-engine.php helpers. Optional: per-occupation industry-trajectory deep-dives, per-stack-component case studies. Required for diagnostic: schema + engine.
Same ownership. Half-life data verified against BLS (US), ONS (UK), DESTATIS (Germany), Eurostat (EU), MEXT/MHLW (Japan), national-stats-office equivalents elsewhere. Tooling-currency assessments cross-referenced with developer surveys (Stack Overflow, JetBrains, GitHub) for tech occupations.
To extend with a new geography weight-map: (1) collect occupation-half-life signals from local source; (2) compute weight relative to baseline geography (USA); (3) append to geography_weights in atlas-skill-half-life.php. Total ship: ~2 hours per geography.