The AI Resistance is Forming. (Should You Join?)

What It Covers

Cal Newport and author Brad Stolberg make the case for a "pro-human resistance" to AI adoption, arguing through three planks that AI erodes meaningful struggle, homogenizes creative output toward a soulless average, and functions like an addictive cognitive virus that spreads from one area of life into all others.

Key Questions Answered

• •The Effort-Meaning Connection: Meaning in work, relationships, and leisure requires struggle by definition. AI's core promise — removing friction — directly undermines the mechanism through which humans derive satisfaction. Stolberg frames this as the "click-of-a-button economy": essays, recipes, dating scripts, all on demand. The problem is not just the output quality but the loss of the fulfilling process that produces genuine satisfaction and a sense of accomplishment.
• •AI as an Average-Maker: Georgetown research on creativity confirms that LLM-generated writing produces punchy sentences, but all punchy sentences sound identical — what researchers call mass homogenization. If you are bad at something, AI raises you to average. If you are already skilled, it pulls you down. Anyone pursuing mastery in a craft should treat that domain as a protected zone, deliberately kept free from AI assistance to preserve differentiation and quality.
• •The Slippery Slope of Partial Use: Stolberg's own experiment with $20/month ChatGPT and Claude subscriptions illustrates the adoption trap. Starting with grammar checks, he found the tools irresistibly suggested rewrites, then conclusions, then structural changes. Newport adds that LLMs are engineered to respond with "that's great, just one small change" — a psychologically addictive feedback loop that makes partial use functionally impossible to maintain over time.
• •Pseudo-Productivity Amplification in Knowledge Work: AI accelerates the worst dynamic in office jobs — visible activity substituting for actual value creation. Newport describes AI-generated emails summarized by AI, turned into AI-written PowerPoints, then re-summarized by another AI. The cost of producing visible work activity drops to zero, creating "AI swirl" analogous to email swirl, while actual output quality and employee meaning both decline simultaneously without improving company revenue ratios.
• •The Ultra-Processed Information Framework: Stolberg proposes treating AI like ultra-processed food — acceptable occasionally for low-stakes tasks, but dangerous as a dietary staple. The practical rule: identify where you want "real food" experiences in your life versus where convenience is acceptable. Critically, remove AI apps from your phone and browser plugins from your computer, because frictionless access is the primary driver of scope creep into protected domains.
• •Intentionality Over Adoption Speed: Newport draws a direct parallel to email and social media — both entered daily life without deliberate frameworks and degraded attention, mental health, and work quality over a decade. The actionable stance: you are not obligated to be a crash-test dummy for new technology. Wait until a specific use case is genuinely easy and demonstrably improves your life by the standard of depth and meaning, not efficiency alone, before adopting it.