130-page paper introduces "bottleneck pressure": work AI can't do gains value as the rest speeds up
soumitrashukla9 · x · 2026-11-17
A 130-page paper by Soumitra Shukla (Burning Glass Institute) introduces "bottleneck pressure" — the marginal value of AI-infeasible tasks rises as AI speeds up the rest of a job.
Key findings:
- Built on 288M US job postings (2018–mid-2026) matched line-by-line to ONET tasks with GPT-4 feasibility ratings; 90% of task mentions fall outside the occupation's ONET list.
- 53% of mention-weighted variation in pressure lies within tasks across occupations, vs just 0.3% using catalog-listed tasks.
- Senior postings face more pressure than junior ones every year, 6.5% more in H1 2026.
- Mean occupation pressure rose 6%, three-quarters from task mix shifting toward higher-pressure occupations.
- Higher pressure predicted faster posting growth after 2022, beyond linear exposure or observed AI use; results hold using Anthropic Economic Index / OpenAI Signals feasibility criteria.
The paper argues pressure should be reported alongside exposure to show where improving AI's blind spots pays off most.
More from AGI Musings
- Top researchers are quietly stopping paper submissions, observer notes — avt_im · 2026-11-17
- "Abundant intelligence should terrify businesses built on keeping it scarce" — _AustinCalvert_ · 2026-11-17
- AI conference authors and reviewers complain unprecedentedly about AI slop — khademinori · 2026-11-17
- Sam Altman: AI already speeds up OpenAI's research, but a self-improving button? 'Definitely not' — ChrisGPT · 2026-11-17
- "Financiers can remain unconvinced longer than you can remain a techno visionary" — shakoistsLog · 2026-11-17
- Kai-Fu Lee pushes CEOs to hire AI agents as Chief of Staff—or even Co-CEO — kaifulee · 2026-11-17