AI's less-discussed effects may be the gradual erosion of human cognitive abilities. Rather than focusing on job losses or AI safety, the author examines what happens when people routinely delegate activities such as research, writing, coding, analysis and verification to AI. The concern is not that humans suddenly become incapable, but that skills can slowly atrophy because people no longer practice them. The author calls this phenomenon the “Great Forgetting”—a form of cognitive deskilling that can occur while people remain employed and apparently productive.
The article uses software development as one example. AI coding assistants can generate substantial amounts of code, allowing developers to spend more time describing objectives and reviewing outputs. The author argues, however, that writing and debugging code were themselves forms of structured reasoning and problem-solving. When those activities are increasingly delegated, developers may become less capable of independently constructing solutions or diagnosing subtle problems. The same concern applies to knowledge workers who use AI for research, emails, analysis and professional writing: productivity may increase while the underlying expertise required to perform the work manually gradually weakens.
A second concern is the disappearance of productive struggle. The author argues that humans learn partly through the effort involved in recalling information, resolving ambiguity, comparing conflicting sources and working through difficult problems. AI systems are designed to remove precisely this friction by producing polished answers quickly. The article also warns that AI systems are becoming increasingly complex combinations of models, search, memory, tools and autonomous loops, making it harder for users to understand how an outcome was produced. If people stop practicing the underlying skills while simultaneously becoming dependent on opaque systems, recovering those capabilities during an AI failure could become difficult.
The article ultimately presents three possible futures: continued cognitive decline managed through AI-literacy programmes; a disruptive “reckoning” following a major AI failure; or an intentional redesign of AI around human capability preservation. The author favors the third approach, suggesting ideas such as measuring skill retention, deliberately introducing friction into AI interfaces, maintaining periods of AI-free work and using AI to identify knowledge gaps rather than simply replacing human learning. Importantly, HackerNoon labels the piece “Opinion piece / Thought Leadership” and “AI-assisted,” so its predictions and probability estimates should be read as arguments rather than established scientific forecasts.
