Noisy
← this week  ·  Subscribe →

The world is being quietly rearranged by people who write very long documents.

NBER · May 1, 2026
The title they went with Automation, Learning, and Career Dynamics Noisy translates that to

Cheaper automation officially becomes a trap for workers who cannot learn new skills fast enough

Cheaper technology creates the exact economic trap it was supposed to eliminate.
The bet
Since the 1980s, the economic debate over automation has been a simple binary: either robots take the jobs, or robots create better jobs. This paper ends the assumption that the technology itself decides the outcome. The bet is that policymakers will ignore the learning variable and just subsidize the tech, accidentally tipping low-learning regions into permanent skill traps. Watch for local governments offering tax breaks for factory automation without tying them to worker retraining programs.
What happened
New research shows that when automation gets cheaper, it can either make an economy better off or worse, depending on how well workers learn new skills. If workers can learn quickly, cheaper automation helps everyone; if not, it can trap people in low-skill jobs.
Why this matters
The economic consensus has treated cheaper technology as an automatic upgrade for everyone. It turns out that if your workforce does not learn fast enough, cheaper robots just lock them into low-skill purgatory forever. The technology is not the destiny, the training budget is.
Making technology cheaper creates widespread prosperity. Making technology cheaper traps an entire economy in low-skill jobs.
Who wins, who loses
High-learning economies High-learning economies conveniently capture all the welfare gains of automation while everyone else gets stuck.
Low-skill workers Low-skill workers lose the assumption that technological progress will eventually pull them up.
Automation subsidy writers Anyone writing a check to subsidize corporate automation.
The signal
People are ignoring this continuous-time general equilibrium math paper right now. That changes the minute a union representing hundreds of thousands of workers strikes over the transition rate from worker to manager. Labor negotiators will use this exact model in the next contract cycle to demand guaranteed retraining budgets every time a company buys new software. Meanwhile, automation vendors will ignore the learning-rate variable entirely and keep marketing their products as tools that guarantee double-digit revenue growth.
The longer story
The labor economics debate is moving away from the 2010s panic that robots will replace every human worker. Researchers are now focusing strictly on human capital traps. This fits a trend where eight out of ten new academic papers argue automation is neither good nor bad for workers. Algorithms simply multiply the effects of the corporate structures that already exist.
If you insist
Read the original →
The Sendoff
A new economic model states that workers acquire their long-term skills strictly through the daily tasks they are assigned. Folks whose only task is watching a kiosk print receipts are not anticipating any massive intellectual breakthroughs.