AI was great until it was not

Based on 14 sampled comments from this thread. Analyzed 2026-10-20.

Commenters spent less time on the article’s prose than on a harder question: what disappears when an LLM becomes the default first stop for answers. Many worry that agents replace the messy exchanges through which coworkers, students, and researchers learn each other’s reasoning. The cost is not merely social nicety; it is weaker mentorship, fewer accidental insights during routine work, and graduates who can produce results without owning the underlying concepts. Several readers see academia as especially exposed, since students may bypass supervisors while supervisors themselves redirect questions to AI. Others reject the idea that human help was a golden age. They recall colleagues who resented interruptions and argue that an always-available tutor can unblock curious learners far faster than university or workplace hierarchies. Experienced users describe LLMs as a force multiplier: they can finally prototype ideas that formerly died for lack of time or programming capacity. The emerging compromise is that AI works best as a complement for people who retain agency and foundational knowledge, not as an autopilot for the work that teaches them. Skeptics doubt most people can reliably distinguish disposable drudgery from the supposedly trivial tasks that later trigger understanding. A smaller group dismissed the article itself as self-indulgent and short on practical conclusions.

The central dispute is whether LLMs chiefly erode learning and collaboration or liberate people from unhelpful bottlenecks. Critics argue that outsourcing questions and difficult exploration hollows out expertise, community, and the productive mental space of routine work; advocates argue that human gatekeepers were often unavailable or unwilling, while AI gives motivated people cheap, immediate instruction and creative reach.

• Praise: An always-available AI tutor can spare learners the bottleneck of reluctant colleagues and make fast, corrective explanation available when curiosity is still alive.
• Praise: For experienced builders, coding agents turn long-shelved ideas into working prototypes, expanding what one person can realistically create rather than merely saving keystrokes.
• Praise: Treating LLMs as a complement rather than a substitute can reward people who keep developing judgment, leaving them more capable than users who blindly delegate thought.
• Praise: Cheap experimentation lets determined learners test many failed paths quickly, potentially shortening the route from vague intuition to genuine understanding.
• Concern: Making the agent the workplace’s first responder cuts off the informal explanations through which engineers trade context, teach each other, and form a real professional community.
• Concern: Students who let a chatbot carry the hard parts may finish credentials with outputs but without the durable conceptual command that difficult study once forced them to build.
• Concern: Routine tasks are not reliably disposable: their boredom and repetition can create the idle mental space where unexpected connections and original ideas emerge.
• Concern: When AI answers replace supervisors and peers, academia risks becoming more isolated and more dependent on generated material rather than reciprocal intellectual challenge.
• Concern: The convenience of asking a model can make avoiding hard exploration feel rational, gradually slowing human skill growth even when the tool appears to accelerate each task.