Jay Caspian Kang's eight predictions for higher education in The New Yorker opens with a deceptively calm premise: the next decade won't be Armageddon. But read the predictions carefully and a harder truth surfaces. The institutions that survive will be the ones that successfully reposition themselves as AI-integrated credential factories. The ones that don't will be treated as apps that failed to attract users. The metaphor is not accidental.
The Learning Gap and the Middle School Window
A 2026 paper on arXiv titled "AI-Integrated Learning Management System for Middle School" presents longitudinal data showing that middle school is the critical window for building learning routines that persist through high school and beyond. The paper's finding: early AI integration in structured learning environments compounds positively over time. The implication no one wants to say out loud is that students in under-resourced schools who don't encounter AI-integrated learning in that window will carry a structural disadvantage that higher education cannot fully remediate. Kang's prediction that elite universities will double down on selectivity while regional schools face enrollment cliffs is the downstream consequence of exactly this dynamic. Fast Company's piece on unlearning at work asks what professional habits need to be shed to survive the current moment. But unlearning is a luxury. It requires the prior existence of stable, transferable skills worth replacing.
DIY Health AI and the Democratization Mirage
The arXiv paper on the DIYHealth Suite is quietly radical. It argues that most health AI advances rely on hospital-grade devices, creating a two-tier system where the people most in need of health monitoring are least likely to access AI-powered tools. The researchers built a dataset and benchmark specifically for home-based health management. This is the access argument applied to healthcare. It mirrors exactly what the higher education cliff looks like from the other end: systems designed for the well-resourced that generate data on the well-resourced, then claim universality. TurboFund's list of 25 healthtech VCs shows where capital is flowing. It is, overwhelmingly, toward the hospital-grade end of the market. The DIY tier remains the domain of academic papers and grant funding, not Series As.