Reading List

Most comprehensive survey of economist views on AI's economic impact. 69 leading economists, 52 AI experts, 38 superforecasters, 401 public respondents. Unconditional forecasts cluster near historical trends for GDP, TFP, and LFPR despite 61% expecting moderate/rapid AI progress by 2030. Rapid scenario (14% prob): 3.5% GDP growth, LFPR drops to 59.1% (~10M AI job losses), top 10% holds 75% of wealth. 71.8% of economists support job retraining (+1pp LFPR, +0.2pp GDP). 80% credible intervals show massive uncertainty in rapid scenario.

Repost of influential 2024 essay with new framing. Argues compute is a producer-specific constraint on AI that preserves comparative advantage for human labor, even if AI surpasses humans at every task. Opportunity cost of AI scales with its productivity, so cheaper compute doesn't eliminate the case for human workers. Key risks: energy competition (the 'horse' scenario), inequality from capital concentration, and adjustment frictions. References Korinek & Suh (2024) model showing wages drop but don't collapse at full automation.

Introduces the American AI Jobs Risk Index. 9.3M US jobs vulnerable under median adoption (range 2.7-19.5M), ~6% industry-wide displacement. Information (18.3%), Finance and Insurance (16.5%), Professional Services (15.6%) face steepest risks. Writers/Authors (57%), Computer Programmers (55%), Web Designers (55%) are most vulnerable occupations. Innovation hubs (San Jose 9.9%, DC 9.1%) face highest geographic risk. $757B in annual income at risk (2-15% of total wage income). 4.9M 'tipping point' workers in 33 occupations swing from <10% to >40% displacement.

49% of jobs have at least a quarter of their tasks performed using Claude. High-tenure users show 10% higher success rates, evidence consistent with learning-by-doing. Coding tasks migrating from Claude.ai to automated API workflows (+14% since Aug 2025). Early adopters with high-skill tasks have more successful interactions, identifying a skill-biased adoption channel that may deepen labor market inequality. Average task value on Claude.ai declined from $49.3 to $47.9/hr as adoption broadened to lower-wage tasks.

Formal model showing that when tasks are quality complements (O-ring production), partial automation triggers a 'focus' mechanism that can raise worker income. Task-by-task substitution logic is incomplete because automating one task changes the return to automating others. Automation decisions can require bundled adoption even when quality improves smoothly. Linear exposure indices overstate displacement when tasks are complements. The relevant object is not average task exposure but the structure of bottlenecks.

Applies Gans & Goldfarb's O-ring model of automation to show that AI exposure measures are routinely misinterpreted as displacement forecasts. In jobs with many complementary tasks (consulting, medicine, academia), partial automation triggers a 'focus effect' that raises worker productivity and wages. Low-dimensional jobs (trucking, warehousing) face the real displacement risk because firms have stronger financial incentive to fully automate when few non-automated tasks remain. Whether exposure leads to better or worse outcomes depends on demand elasticity, job dimensionality, and firm incentives -- not average task exposure alone.

Surveys 250 years of technological anxiety across three recurring themes: displacement fears, work dehumanization, and stagnation pessimism. Finds predictions of widespread technological unemployment were 'by and large, wrong' but cautions against trivializing costs to the displaced. Ricardo reversed his own position on machinery in 1821. The Luddite riots were driven more by wages than anti-machinery sentiment. Annual hours worked halved from 2,950 (1870) to 1,500 (1998). Concludes with Amara's Law: we overestimate technology's short-run effect and underestimate its long-run effect.

Jan/Feb 2026 CPS data and Anthropic's new 'Observed Exposure' metric do not suggest any substantial AI labor market impact. Occupational dissimilarity, industry dissimilarity, and all exposure/usage metrics remain flat, within historical ranges, or on pre-existing trends. OpenAI exposure quintiles stable. No relationship between automation/augmentation measures and employment or unemployment changes.

Industrial Revolution novels illuminate what living through epochal technological change feels like. Handloom weavers -- the white-collar workers of their day -- saw real wages fall by half between 1806 and 1820. Recommends Shirley (Bronte), North and South (Gaskell), and A Christmas Carol (Dickens) as guides to understanding technological disruption, labor unrest, and society's response to those who lost out. AI transition may be faster and bigger, but the human dynamics are strikingly familiar.

Stanford grad turned AI startup CTO argues cognitive labor automation is months away, not decades. Reports 4-week engineering tasks completed in 45 minutes with AI. Physical labor automation follows on 5-10 year timeline as AI-compressed R&D accelerates robotics. The identity crisis for displaced knowledge workers may be worse than the economic loss, paralleling Rust Belt deindustrialization at larger scale. Post-abundance world splits into zero-marginal-cost cognitive goods and still-scarce physical/experiential goods.

Anthropic invited Claude users to share how they use AI, what they dream it could make possible, and what they fear it might do. Nearly 81,000 people responded in one week — the largest qualitative study of its kind.

Survey of ~750 CFOs: <0.4% aggregate job loss expected, productivity paradox (reported gains 3x implied), finance leads at >2% productivity growth. Routine clerical roles declining 0.76pp/yr offset by skilled-technical gains. 85.4% of firms expect to invest in AI in 2026.

Swedish population register data (4.6M job ads, full-population employer-employee records): 22-25yr employment in high-AI occupations fell 5.5% by 2025H1 vs low-exposure occupations within same employers. 50+ workers rose 1.3%. Broad posting decline driven by Riksbank rate hike, not AI. Effect ~2x larger for young women. Replicates US 'canaries' finding (Brynjolfsson et al.) in Nordic register data; contrasts with Finnish null.

Interactive mapping of 350+ occupations by AI exposure and adaptability using GovAI/Brookings research. 6.1M clerical/admin workers most vulnerable (86% women). Consensus: no measurable aggregate displacement yet, but white-collar jobs first in line. Economists stress humility — predictions have historically been poor.

Explores how AI automation extends beyond job displacement to challenge deeper questions about work's role in identity, purpose, and social meaning.

Dutch youth employment in GenAI-vulnerable occupations fell 13% from Q4 2022 to Q3 2025, while other occupations rose 3%. Job postings in vulnerable occupations down 25% vs <10% elsewhere. ICT sector grew 6.5% but youth employment fell 11%. Mirrors US findings from Brynjolfsson et al. (2025).

Individual AI productivity gains (10x) aren't translating to firm-level value. Historical parallel: electrified textile mills saw no output gains for 30 years until factories were redesigned. Proposes 'Institutional Intelligence' framework with 7 pillars for organizational AI adoption.

ATMs didn't reduce bank teller employment due to complementarity, but iPhones/mobile banking caused a 51% collapse (332K to 164K, 2010-2022). Task automation within existing paradigms rarely displaces; paradigm replacement does. AI is still in the 'slotting in' phase.

Evidence on AI labor impact is inconclusive; disruption pace matches prior tech transitions. Under 1/5 of firms using AI per Census BTOS.

New 'observed exposure' metric combining LLM capability with real usage. No systematic unemployment rise, but young worker hiring slowing in exposed occupations.

Eight-month study of 200 employees found 83% said AI increased their workload through greater pace, scope, and hours -- leading to burnout and cognitive fatigue.

First canonical aggregator of research on how AI is impacting jobs at the occupational level, with implications and actions for job seekers.

Millions of women in clerical and customer service roles face AI exposure, echoing the pattern of manufacturing's toll on men during deindustrialization.

Employment declined 1% since late 2022 in the 10% of sectors most exposed to AI. Wages in AI-exposed sectors rising faster, suggesting augmentation and displacement happening simultaneously.

No meaningful relationship between productivity and AI adoption economy-wide, but companies that quantified task-level AI productivity saw a median 30% gain.

Overall employment growing robustly, but 22-25 year olds' employment stagnant since late 2022. AI-exposed jobs driving the decline for young workers.

Two-thirds of 3,500 euro-area firms use AI. Companies with frequent AI use are 4% more likely to hire additional staff. Only 15% cite reducing labour costs as a factor.

Little evidence of AI disruption in labor data. Forward-looking labor market components have improved. AI data center construction driving a pick-up in construction hiring.

Survey of ~6,000 executives: 90%+ report no employment impact from AI over past 3 years. Firms expect only -0.7% employment reduction over next 3 years (US: -1.2%).

No evidence that AI reduces employment in the short run. AI augments worker output -- enabling faster task completion and better decisions -- without displacing labour.

Comprehensive review: 35.9% of US workers used generative AI by December 2025. Adoption accelerating but unevenly distributed across industries and skill levels.

Of 37.1 million workers in the top quartile of AI exposure, 26.5 million are in occupations with above-median adaptive capacity. Not all exposed workers are equally vulnerable.

~10% of US work affected short-term. AI will boost productivity and GDP by 1.5% by 2035. Jobs AI can completely replace saw 0.75% employment fall 2021-2024.

AI can already replace 11.7% of the US workforce ($1.2T in wages). Tech layoffs represent just 2.2% of total wage exposure -- the visible tip of the iceberg.

AI will increase TFP by only 0.53-0.66% over 10 years. Much less transformative than the hype suggests, because AI can only automate a fraction of tasks in most occupations.

37.4% of US workers used AI on the job in the past week. Adoption highest among college-educated (52%) and workers earning $100K+ (58%). Survey of 25,000 workers.

Formal models show AGI-driven demand collapse is theoretically possible if labor share collapses and capital owners are satiated, but conditions are extreme. Benzell et al.'s OLG model shows 'immiserating growth' via capital decumulation when wages fall and savings decline. Imas concludes negative growth is unlikely but demand-side forces will moderate AGI growth expectations. Recommends sovereign wealth funds for broad capital ownership.