Noah Smith
· Noahpinion
· April 03, 2026 at 16:50
· ⏱ 12 min read
| Read on Substack ↗
Summary
AI is currently replacing tasks, not jobs, but the structure of work is shifting toward three categories: specialists with strongly bundled tasks, generalist 'salarymen' who fill AI gaps, and small businesspeople leveraging AI leverage. This implies no near-term mass unemployment but significant labor market reorganization, with implications for corporate hiring patterns and skill demand rather than specific investable themes.
•U.S. prime-age employment rates are near all-time highs, and CFO surveys show little evidence of near-term aggregate employment declines due to AI.
•Geoffrey Hinton's prediction that AI would displace radiologists has not materialized; radiologists are in greater demand than ever.
•Software engineers now focus on checking and maintaining AI-written code rather than writing code from scratch.
•Hunlum and Vestergaard (2026) find precise null effects on earnings and hours two years after ChatGPT's launch in Denmark, but note task reorganization into AI oversight and integration.
•Garicano, Li, and Wu (2026) theorize that AI replaces weakly bundled jobs faster than strongly bundled ones, with specialists (e.g., bloggers, radiologists) retaining employment due to task bundling.
•The author forecasts future work divided into specialists, generalist salarymen (modeled on Japanese corporate rotation), and small businesspeople using AI agents to operate with minimal teams.