A short, curated digest — every item from a named research institution, translated into plain English with one line on what it means for you. No hype, no doom, no product news.
Anthropic analysed millions of real AI conversations and found the same model runs far more autonomously in some tools than others. Workers who use AI in more automated ways actually reported more optimism about their pay and job security, not less.
How you choose to work with AI shapes your outcome more than the AI itself — directing it well is a learnable skill.
Stanford researchers tracking millions of payroll records find employment for 22–25-year-olds in AI-exposed jobs has fallen roughly 13–16% since late 2022 — but only in roles where AI is used to fully automate work. Where AI assists people instead, employment held steady or grew.
The dividing line isn't your job title — it's which of your tasks AI takes over versus amplifies. That's measurable.
PwC's 2026 analysis of nearly a billion job ads finds workers with AI skills now command a 62% wage premium — up from 25% just two years ago. Entry-level jobs most exposed to AI are now seven times more likely to also demand senior human skills like leadership and creativity.
Pairing AI fluency with distinctly human skills is currently the best-paid combination in the job market.
METR's latest frontier report shows models can attempt ~12-hour tasks with a coin-flip success rate, but the tasks they complete dependably are closer to 90 minutes. The gap between "can attempt" and "can be trusted with" remains large.
That reliability gap is exactly where human oversight stays valuable — checking AI's work is becoming a career skill in itself.
For four consecutive months in 2026, US employers named AI as the top reason for announced job cuts — around 100,000 AI-attributed cuts in the first half, concentrated in tech and clerical roles. Economists caution some of this is "AI-washing" of cuts that would have happened anyway.
Displacement is real but narrow so far — concentrated in specific task profiles, which is why a task-level view of your own role matters.
Three of the world's leading labor economists categorise AI's effects into five types — from pure automation to creating entirely new tasks. Their warning: AI that merely "levels" expertise erodes the wage premium of experienced workers even without eliminating jobs.
If AI makes your hard-won expertise easy for anyone, the defence is moving up to judgment, relationships, and accountability.
METR measures how long a task AI can finish reliably. Their updated benchmark shows frontier models completing tasks that take skilled humans about 5 hours — and that capability has been doubling roughly every 89 days since 2024.
Whatever AI can't do in your job today, check again in six months — plan your skills for the trajectory, not the snapshot.
The IMF finds employment runs 3.6% lower in regions five years after high AI-skill demand appears — concentrated in occupations where AI substitutes rather than complements. About 1 in 10 advanced-economy vacancies now demands a skill that didn't exist before.
AI fluency alone isn't protective — it pays off when paired with tasks where humans stay essential.
MIT researchers scored 19,000 tasks on five human capabilities — empathy, presence, opinion, creativity, hope. Jobs rich in these grew faster from 2015–2023, hired more in 2024, and carry stronger projections through 2034.
The human capabilities in your work aren't a consolation prize — they're the growth segment of the labor market.
Using Danish administrative data covering the whole economy, researchers find AI chatbot adoption produced essentially zero effect on average earnings and hours through 2024 — even as tasks inside jobs visibly restructured.
There's a window between AI capability and labor-market impact. The time to reposition is inside that window, not after it closes.
The research is about averages. Your job is specific. Paste your job description and see which of your tasks AI takes over — and which are yours to build on.
Map my week — free