Generative AI, Simplicity, and Easiness (gioleppe.github.io)

At Linux Day 2025 the author noticed how Google’s new “AI Overview” — an LLM-generated answer box atop search results — is already reshaping behavior: early reports and preliminary research show it siphons traffic from canonical sources (news sites, Wikipedia), and people are using screenshots of those answers to respond to others instead of reading original material. A personal anecdote (friend B answering a WhatsApp question by copying Google’s Overview) illustrates how effortless access to synthesized answers drives adoption: because users already know how to use search, adding an AI layer makes correct-looking information extremely easy to consume and share. The essay’s core point—drawing on Rich Hickey’s Simple Made Easy—is that “easy” (familiar, low-friction) and “simple” (low internal intertwinement or complexity) are orthogonal. Generative AI products (GitHub Copilot, Claude Code, Cursor, AI Overview) increase ease but do not reduce the intrinsic complexity of software engineering, distributed systems, or creative disciplines. Consequences include skill atrophy, misinformed management decisions, and loss of learning opportunities. For the AI/ML community this means prioritizing tools that enable learning and composability, investing in training, and resisting claims that models simplify deep domain complexity; use GenAI to augment workflows thoughtfully, not to replace foundational knowledge.