Weekly Digest: Five Things Worth Your Time
Five things worth your time this week. The labor market is sending conflicting signals, a new AI policy debate is just getting started, and one number that reframes how you think about productivity.
What we're watching
US initial jobless claims fell to 215,000 this week, below consensus, but the four-week moving average has been quietly rising since April. The divergence between strong weekly claims data and softening job openings (down 18% from peak) is one of the more interesting tensions in the current macro picture. The labor market is not breaking, but the internal structure is changing.
China released July economic data this week that showed consumer price inflation drifting into deflation territory for the second consecutive month. Deflation in the world's second-largest economy has global implications: it means Chinese exports get cheaper in real terms, adding to disinflationary pressure in economies that import Chinese goods, while raising the risk of a debt-deflation spiral domestically.
The EU released its proposed AI Act implementing regulations, and the debate over the definition of 'general purpose AI' with systemic risk is genuinely consequential. How the EU classifies the frontier models from OpenAI, Anthropic, and Google will determine compliance costs and potentially whether those models can be deployed in Europe without significant modification.
One number
$13,800. That's the estimated per-worker cost of the 2021-2022 'Great Resignation' to the companies that lost those workers, according to research from the Society for Human Resource Management. Multiply by the roughly 47 million workers who quit voluntarily in 2021 alone and you get a sense of the scale of disruption that the labor market churn of that period represented.
The interesting follow-up question: how much of that disruption was genuinely costly versus a healthy reallocation of workers to better-matched jobs? Research on worker outcomes post-resignation found that switchers earned roughly 8-10% more in their new roles than stayers. The disruption was real. So was the value it created for the workers willing to move.
One idea
The 'overhang' concept in economics: when a large stock of something (debt, unsold inventory, unused capacity) suppresses the price or incentive structure for new activity, you have an overhang. Housing debt overhang explains why consumer spending was weak for years after 2008, as households paid down debt rather than consuming. Student debt overhang is a candidate explanation for why household formation among millennials has lagged previous generations.
The concept applies beyond economics. Skill overhangs (workers trained for declining industries), attention overhangs (too much media competing for finite attention), and relationship overhangs (social capital from a previous era that constrains new connections) are all versions of the same dynamic: the past weighs on the present.
Worth reading
A deep dive into how Taiwan Semiconductor's pricing power has quietly been one of the most important macroeconomic variables in the global semiconductor supply chain, and what TSMC's price increases mean for the cost of AI compute over the next three years: https://ft.com/content/tsmc-pricing-power-ai-compute
The Atlantic's examination of why the US immigration system has become a significant drag on the talent pipeline for American technology companies, with concrete data on the backlogs, processing times, and the countries benefiting from the US system's dysfunction: https://nytimes.com/2025/08/immigration-talent-drain-tech
A rigorous analysis of whether ESG investing has actually delivered on its stated goals of channeling capital toward sustainable businesses, or whether it has primarily generated fees for asset managers: https://economist.com/finance/2025/esg-scorecard-decade